• Open

    DeFi Protocol Types Part 2: Yield, Bridges & Governance
    Series: Web3 Security Zero se Advance 🛡️ | Article #18 By HackerMD | 30 min read Continue reading on Medium »
    Subdomain Takeover: A Complete Guide from Beginner to Advanced
    A comprehensive deep-dive for bug bounty hunters, penetration testers, and security researchers Continue reading on Medium »
  • Open

    别再用网页翻译看源码了!你的私人翻译神器LibreTranslate,部署避坑指南来了 - 一名程序媛呀
    看GitHub源码或英文文档时,还在用谷歌翻译来回切换?翻译不准、网络抽风、代码被转义?本文分享一款开源、可本地部署的翻译工具LibreTranslate,从安装、使用到与ArgosTranslate的对比选择,帮你搭建流畅阅读外文技术资料的翻译神器。  ( 1 min )
    上周热点回顾(6.15-6.21) - 博客园团队
    热点随笔: · 面试官问:你用 AI 编程半年了,那怎么保证 Claude Code 写出来的代码是对的? (狂师) · 分享一些程序员很棘手但是却又简单的工具 (保持当下) · 我的第一个skill (第十昵称) · Hermes官方桌面版发布了 (程序员晓凡) · 短短三天,英伟达和微软同时宣告  ( 1 min )
    零基础也能做!用 AI 提示词一键生成 Three.js 3D 中国地图 - 微山湖上静悄悄
    前言 想做一个炫酷的 3D 中国地图,但觉得 Three.js 太难?WebGL 太复杂?GeoJSON 是什么鬼? 这篇文章教你用 AI 写提示词,零基础也能快速生成一个可交互的 3D 区划地图。不需要手写一行代码,只需要说人话。 最终效果:可拖拽旋转、缩放、鼠标悬停高亮的 3D 中国省份地图。  ( 3 min )
    我把整个代码库喂给 Claude Code,工具超 50 个就静默丢失,这个坑太阴了 - 码哥字节
    上个月我们做了一个实验:把一个 180K 行的 Spring Boot 单体代码库接入 Claude Code,让它做一次全量架构分析。 结果 Claude 给出了一份比我们技术 lead 更细的依赖关系图,还发现了三处我们自己没注意到的循环依赖。 听起来很美好。但在那之前,我们踩了整整 6 周的坑  ( 4 min )
    代码之外周刊(第179期):人最大的行动阻力,是「想清楚再说」 - 我没有三颗心脏
    代码可以构建世界,但生活中的乐趣远不止于此。这里是我一周的精选。 周刊开源(Github:wmyskxz/weekly),欢迎提交 issue,投稿或推荐精彩内容。 题图 2026 年 2 月,一支由 16 名非洲和国际专家组成的科考队进入安哥拉东部偏远的 Lisima 高原,发现数十个科学界从未记  ( 4 min )
    Roslyn语法的模式匹配之EasySyntax增加模式匹配支持 - xiangji
    一、先看一个模式匹配的Case 该Case是一个使用模式匹配检测回文算法,只有一行代码 这是不是你见过最简洁的检测回文代码 但是这里面用到了不少模式匹配 有逻辑模式、列表模式、var模式和切片模式 模式匹配不仅仅是语法糖,在.net中有很高的地位 所以SourceGenerator非常有必要支持生成  ( 4 min )
    Vercel Eve 快速入门:使用 eve init 构建你的第一个 Agent - 程序猿DD
    上一篇:《一个目录一个 Agent,Vercel Eve 的这套 Agent 架构设计太舒服了!》 我们聊了 Eve 为什么值得关注:它不是再提供一种模型调用写法,而是把 Agent 运行所需的项目结构、持久执行、沙箱、审批、渠道和评测等能力,放进一套 filesystem-first 的工程框架里  ( 3 min )
    手写高性能在线 JSON 工具|Web Worker 工程化打包 + 语法自动修复 + 多语言代码生成实战 - 疯狂SQL
    本文深度剖析纯前端 JSON 工具的完整实现。基于 Web Worker 彻底解决大文件解析卡顿;自研智能语法修复引擎;并分享 esbuild 独立构建 Worker + manifest 映射的工程化最佳实践,附完整性能基准测试数据。  ( 11 min )
    我开源了一个开发者专属的智能 JSON 工具,得到了媳妇高度认可 - 程序员小富
    大家好,我是小富~ 这期给大家分享一个我开发的工具! 最近做了一个开发者专属的 JSON 工具 easy-json!在线体验地址:easyjson.xiaofucode.com,不管是什么奇形怪状的 JSON 字符串,扔进去都能智能解析,先看效果是不是还挺能打的 现在网上这类工具确实不少,我也用过很  ( 2 min )
    AI Native 架构:有限上下文、确定性边界与质量闸门 - AI小老六
    AI Native 的关键不是让模型替代一切,而是重新划定确定性逻辑、模型运行时与质量闸门的边界。 原文链接:AI 小老六 导语 AI Native 架构最容易被讲成一个很大的概念:模型会写代码、会调工具、会看页面、会自己规划任务,于是软件形态都要重来。这个判断方向没错,但如果直接从“未来应用长什么  ( 2 min )
    折腾半小时,终于让AI 能直接帮我写飞书文档了 - 程序员晓凡
    飞书 CLI 实战:让 AI Agent 帮你操作飞书 一、事情是这样的 前阵子在排查公司内网一个 Java 服务的问题,日志刷了一屏幕。问题定位到了,方案也理清了,于是习惯性对 AI 说:"帮我把分析结果整理一下,写成一个飞书文档。" AI 很快就生成了一段漂亮的文字分析。然后我说:"写到飞书里。  ( 2 min )
    [MAF预定义的AIContextProvider-04]Mem0Provider——长期记忆基于的云端解决方案 - Artech
    `ChatHistoryMemoryProvider`利用我们提供的向量数据库,对每次调用产生的消息针对指定的Scope维度进行存储,并将当前消息作为查询文本,结合设定的Scope维度检索历史消息作为上下文的一部分来参与LLM的推理。除了这种需要我们们自己搭建和维护的基于向量数据库的解决方案之外,我...  ( 3 min )
    Claude Fable和Claude Mythos 5同时发布:注意力机制下愈加强大的AI大模型 - 岳小哥AI
    强化学习通过人工打分排序调整模型行为,塑造出不同大模型性格。Claude变得保守严谨,遇到不确定问题倾向于说不清楚。ChatGPT变得健谈,愿意尝试回答各种问题。对齐人类偏好是第三阶段训练的核心,直接决定模型好不好用。  ( 1 min )
  • Open

    SocGholish Favourite Weapon: I Spent the Weekend Building the Same Thing.
    An international law enforcement operation dismantled SocGholish. Instead of reading about it, I rebuilt the entire attack chain. Continue reading on Medium »

  • Open

    Reflected XSS into a JavaScript string with angle brackets HTML encoded | Shiv Kumar | OWASP Top 10…
    Portswigger lab Reflected XSS into a JavaScript string with angle brackets HTML encoded Continue reading on Medium »
    Exploiting AI Agents to Perform Destructive Actions — A Hands-On Lab Walkthrough
    Artificial Intelligence (AI) agents are rapidly becoming part of modern applications, helping users automate tasks, analyze content, and… Continue reading on InfoSec Write-ups »
    How one line of JavaScript let me walk into an admin panel
    So this one started just out of curiosity. Continue reading on Medium »
    “How to Set Up an Android Application Security Testing Lab Fast!!”(for kali linux)
    by Muhammad Murtaza(Penetration Tester/Cybersecurity expert) Continue reading on Medium »
    The 2026 Bug Bounty Starter Pack — 10 Free Tools That Cost Me $0 to Learn
    The Lie That Stops People Before They Start Continue reading on Medium »
    Can You Build a Career on Bugcrowd? I’m Going to Test It. (Day 0)
    🧠 Idea That Hit Me Today… Continue reading on Medium »
    From Revoked Privileges to Resource Creation: A Privilege Persistence Vulnerability in a VDP…
    My name is Montaser Mohsen, a security researcher and bug bounty hunter focused on web application security, broken access control… Continue reading on Medium »
    How I Found My First Bug
    I wasn’t looking for anything groundbreaking. I just asked a question no one had bothered to ask before. Continue reading on Medium »
    Smart Contract Reading Code Audit Karna Seekho
    Series: Web3 Security Zero se Advance 🛡️ | Article #16 By HackerMD | 28 min read Continue reading on Medium »
    Bug Bounty Nightmare: Company Banned Me to Avoid Paying
    I’ve been an avid bug bounty hunter for the past few years, and I’ve found my fair share of bugs from all classifications. One thing I’ve… Continue reading on Medium »
  • Open

    MCRTA Exam Walkthrough — Multi-Cloud Red Teaming on AWS, Azure & GCP
    Certification: MCRTA (Multi-Cloud Red Team Analyst) Issued by: CyberWarFare Labs (CWL) Difficulty: Intermediate Format: Practical… Continue reading on Medium »
    Building WAF-Bypass.dev: A Small Tool for Learning Web Payload Obfuscation
    When you spend time learning web application security, you quickly realize that payloads are only one part of the story. Continue reading on Medium »
    Information Disclosure via Exposed Dreamweaver Metadata Files and Directory Listing
    Continue reading on Medium »
    SQL Injection Authentication Bypass Leading to Unauthorized Admin Access in Human Resource…
    Introduction : Continue reading on Medium »
    THM Weaponization Room
    https://tryhackme.com/room/weaponization Continue reading on Medium »
    Why I’m No Longer on Medium (And Where to Find My Work Now)
    The WireHawk Security Blog is now live at blog.wirehawksecurity.com Continue reading on Medium »
    Insecure Direct Object Reference (IDOR) — TryHackMe IDOR Room
    1. Overview Continue reading on Medium »
  • Open

    一文读懂 OpenAI Codex 源码的原理、架构与未来 - 哥不是小萝莉
    1. 概述 如果你第一次打开 codex-main 这个源码目录,很容易被它的规模吓住:顶层有 npm 包、Rust workspace、SDK、app-server、MCP、插件、技能、沙箱、TUI、云任务、线程存储、模型提供商、登录认证等大量模块。它不像一个传统命令行工具,也不像一个简单的 Ch  ( 12 min )
    Spring事务 - 风倒梧桐树
    全面讲解 Spring 事务原理与实战避坑,包含 ACID、事务管理器、@Transactional 注解详解、传播行为、隔离级别、MVCC,深度分析 AOP 代理底层逻辑,汇总 7 大事务失效场景,附带校招面试真题与 MySQL 日志拓展知识点。  ( 2 min )
    Loop Runtime 架构拆解:别再手动催 Agent,先把工程闭环跑起来 - AI小老六
    拆解 Coding Agent 从 Prompt 到 Loop 的系统化改造:触发、隔离、验证、状态和人工闸门。 原文链接:AI 小老六 导语 过去一段时间,很多人使用 Coding Agent 的方式其实还停留在"远程结对":把需求写清楚,补一段上下文,等它返回结果,再继续追问。这个办法有用,但杠  ( 2 min )
    Fable 5 被关停前 72 小时,到底做出了哪些不像 AI 能做的应用? - 我没有三颗心脏
    Fable 5 被关停前 72 小时,到底做出了哪些不像 AI 能做的应用? 2026 年 6 月 9 日,Anthropic 放出了 Fable 5。 6 月 12 日,它被暂停访问。 中间只有 72 小时。 就这 72 小时,网上冒出一批很奇怪的东西。 比如只看游戏截图,通关了《Pokémon  ( 6 min )
    从"任意文件复制"深挖Java I/O:字符流与字节流的本质抉择 - MalouP
    从文件复制这个小场景深挖Java I/O设计哲学:字符缓冲流适合文本但会损坏二进制数据,字节缓冲流才是万能方案。对比两种流的本质差异、缓冲区大小对性能的影响,以及NIO零拷贝的延伸思考。附完整代码示例和实战选型原则。  ( 2 min )
    GDB调试变量、内存与寄存器查看与修改 - ttkwzyttk
    一、变量查看与修改 在使用 GDB 调试程序时,除了控制程序的运行流程以外,最常见的操作就是查看变量的值、查看参数的值,以及在调试过程中临时修改某些变量的值。通过这些命令,可以观察程序运行时的数据变化,从而判断程序逻辑是否符合预期。 1.1 查看函数参数 使用info args(i args)命令可  ( 5 min )
    Microsoft Agent Framework 1.0 正式接棒,.NET AI 进入 Agent-Native 时代 - 张善友
    本文从技术趋势深度解读视角,解析 MAF 1.0 的五层架构设计、核心理念演化、关键 API 差异、开放协议集成以及 BUILD 2026 最新动态,为 .NET 开发者和 AI 技术从业者提供一份完整的迁移认知地图  ( 5 min )
    数字电路实验BLog - xaioka
    前言 在完成航班配载系统初次实验后,我自认为已经掌握面向对象程序设计的基础思路,然而数字电路逻辑仿真系统实验,让我对软件架构设计产生了全新、系统的认知。 本次实验区别于前次对现实物理载体的简单建模与数值计算,核心任务为解析结构化文本输入,在内存中构建完整电路拓扑模型。系统需要兼容多种基础逻辑门,包含  ( 2 min )
    15天学会AI应用开发(八)使用向量数据库实现RAG功能 - aqi00
    上一篇文章采用字符串匹配的方式来查找知识,这种方式比较呆板不够智能,接下来引入向量数据库,通过向量化实现更精细的知识检索。 一、文本嵌入Embedding模型 向量化一段文本的时候,经常用到all-MiniLM-L6-v2,它是一个轻量级的文本嵌入Embedding模型,专门把文字转成数字向量,给R  ( 1 min )

  • Open

    Deferred Procedure Calls — DPCs : A Real Deep Dive Internals, WinDbg, and Exploitation
    IRQLs → KDPC internals → WinDbg live analysis → vulnerable driver code → ROP-based exploitation → BYOVD → mitigations breakdown Continue reading on Medium »
    AI Red Teaming Will Need Qualified Human Pentester
    AI red teaming will not stay a tool-only activity. CISOs will want qualified people to scope tests, validate findings, and own risk… Continue reading on Medium »
    Building LivingOffTheLand.dev: A Practical Reference for LOLBins
    When learning offensive security, red teaming, and post-exploitation techniques, one concept that appears again and again is “Living off… Continue reading on Medium »
    NRT-Bench as a Narrow-Game Model of Nuclear Plant Cyber-Security — and Why It’s Still a Toy Model…
    Author: Berend Watchus. Independent AI & Cybersecurity Researcher. Continue reading on System Weakness »
    Kerberos — The Three Headed Dog and How to Kill it?
    When I’d just started reading about Kerberos, my brain immediately went to that three-headed dog from Greek mythology. You know, the one… Continue reading on Medium »
  • Open

    From a Simple Input Field to Stored XSS — A Realistic Bug Bounty Journey
    Introduction Continue reading on Medium »
    I Tested a Login Form and Accidentally Found Remote Code Execution
    Category: Web Application Security Vulnerability Chain: SQL Injection → Database Access → Remote Code Execution Severity: Critical Status… Continue reading on Medium »
    Bug Bounty Automation Toolkit 2026 Automate Recon & Find Vulnerabilities Faster | Advanced Ethical…
    Everyone is automating recon in 2026. The tools are faster than ever. The pipelines run while hunters sleep. The results for most of them… Continue reading on Medium »
    Time-Based Blind SQL Injection — Confirming and Exploiting with curl and Ghauri
    A Bug Bounty Writeup | HackerOne Report Continue reading on Medium »
    Poisoned Pipeline in Google’s Gemini-CLI: workflow_run PPE
    How attacker-controlled artifact data flows into a privileged GitHub Actions context and exposes GEMINI_API_KEY — found via Google OSS VRP. Continue reading on Medium »
    Active and Passive Recon, What Is It?
    When I first started learning and looking into bug bounty, I encountered something that confused me and made every program look difficult… Continue reading on Medium »
    User Role Controlled By Request Parameter -Privilege Escalation Walkthrough: Portswigger
    The Objective of this lab was to show that authorization methods should never rely solely on user-controlled inputs. In this assessment, I… Continue reading on Medium »
    My First Bug Found Me (I Wasn’t Even Looking)
    Sometimes the universe just hands you a payout. Continue reading on Medium »
    API Fuzzing for Bug Bounty — Part 2b: Injection, Bypasses & Output Exploitation
    In Part 2a, we broke down the auth and authorization layer, JWT attacks, IDOR chains, and mass assignment. This post picks up at the input… Continue reading on OSINT Team »
    The Token That Trusted Too Much
    How I Learned to Break JWT Authentication and Why You Should Test It Too Continue reading on OSINT Team »
    Burp Suite Scanner Explained — Burp Suite Learning Series #5
    Where Automated Scanning Meets Manual Skill Continue reading on OSINT Team »
    My New Personal Blog
    I’m excited to share my new personal blog with you 🚀 Continue reading on Medium »
  • Open

    1-Click Account Takeover via Open Redirect through Regex Bypass in Domain Validation
    Khan Academy disclosed a bug submitted by farr: https://hackerone.com/reports/3723458
  • Open

    1-Click Account Takeover via Open Redirect through Regex Bypass in Domain Validation
    Khan Academy disclosed a bug submitted by farr: https://hackerone.com/reports/3723458
  • Open

    Python中绘制R的科研级可视化图 - 一眉师傅
    一、重要的桥接Python库 rpy2# pip install rpy2 我的R语言版本为 4.1.3 ,需要rpy2 3.5.X版本去对应 二、确保R环境 确保你电脑已安装 R,并在 R 里装好相关包:# install.packages(c("ggplot2", "ggstatsplot",  ( 3 min )
    lil_tea c++ 2026 style guide - lil_tea
    lil_tea c++ 2026 style guide 本文使用 yoga air + wsl + ubuntu 26.04 + vscode 编写. 把所有 debug 完成, 我保证, codin 是我一辈子的朝圣. 这是我在开发和算竞都使用的代码风格. 部分借鉴自 the cherno, 部  ( 5 min )
    SSRF漏洞:从内网探测到云元数据窃取,黑客是怎么绕过的? - 明.Sir
    上个月帮朋友做了一次红蓝对抗,对方一个看似人畜无害的「图片裁剪」功能,差点让我们把整个阿里云账号的 AccessKey 都掏出来了。 这就是 SSRF(Server-Side Request Forgery,服务端请求伪造) 的威力——你以为是前端传了个URL,服务端老实巴交地去请求,结果那台服务器  ( 3 min )
    一个目录一个 Agent,Vercel Eve 的这套架构设计太舒服了! - 程序猿DD
    2026 年 6 月 17 日,Vercel 发布了开源 Agent 框架 Eve。官方给它的定位很直接:像 Next.js 之于 Web 应用一样,Eve 想给 Agent 提供一套约定明确、可直接进入生产环境的框架。但它真正想解决的,并不是「怎么再写一个 Agent Loop」,而是 Agent  ( 3 min )
    AI Agent 30天速成|Day5 笔记 - 云淡风轻YangG
    AI Agent 全日制30天速成|Day5 教学笔记 今日总学习目标 区分短期记忆/长期记忆,掌握Agent分层记忆架构设计 实现滑动窗口、摘要压缩、向量记忆三种主流记忆方案 基于Day4规划Agent改造,接入持久化对话记忆(Redis存储会话) 解决超长多轮对话Token爆炸、历史遗忘、上下文  ( 4 min )
    K8s 多 Master 重启:流程梳理与问题排查 - 怎么还在写代码
    前言 集群部署在 VMware 创建的三台虚拟机上,每台虚拟机同时承担 Master 角色。因长期未做系统安全更新,近期执行了 dnf upgrade-minimal --security --allowerasing 升级内核与软件包。内核等更新需重启节点才能生效,而三台节点都运行着 etcd,重  ( 8 min )
    GDB调试简介与调试配置 - ttkwzyttk
    一、GDB简介 GDB,全称为GNU Debugger,是GNU项目提供的一款命令行调试工具,主要用于调试C、C++ 等语言编写的程序。在Linux开发中,GDB是非常常用的调试工具,尤其适合分析程序运行过程中的崩溃、段错误、逻辑错误以及变量变化情况。 正常情况下,我们运行一个程序时,只能看到程序最  ( 1 min )
    我把那个迭代了 18 个版本的 SDK 整个掀翻重写了:stock-sdk v2 升级手记 - 程序猿的程
    stock-sdk 历经 18 个 v1 版本、门面类膨胀至 105 个扁平 getXxx 后,于 v2 推倒重来:API 收拢为命名空间,subpath 支持按需 tree-shaking,数据契约与 SdkError 统一,normalizeSymbol 收敛符号格式,CLI/MCP/Playg...  ( 3 min )
    浏览器缓存之【结构化数据库与缓存】: IndexedDB、Cache storage 和 Storage buckets - 橙子家
    本文主要介绍了结构化数据库与缓存中的 IndexedDB、Cache storage 和 Storage buckets,以及期间的区别,供参考。  ( 5 min )
    GPIO同时复用(ADC与TIM) - ifndef
    众所周知,MCU的GPIO可复用为不同的功能.那么是否可以同时使能多个功能呢?可以(当然,部分架构可能不支持) 常见示例: 模拟IIC: GPIO在开漏输出的同时可以读IDR来获得输入 无LPUART的MCU实现低功耗唤醒: 把RXD同时复用为外部中断来唤醒,但会丢失首字节 UART自动波特率识别:  ( 2 min )
    逆向一个被遗忘的DVD游戏格式:从DES加密到Rust模拟器 - AloysHF
    缘起 前阵子翻出一台老的便携 DVD 播放器,发现里面居然有游戏菜单。七个分类,80 多款游戏。查了一下发现这东西叫 Native32,凌阳科技的芯片方案,2005-2011 年间大量用在 DVD 播放器和车载显示器上。 游戏由 Potatoo Multimedia Studio 开发,格式是私有的  ( 4 min )
    [技术复盘] Windows Python 打包实战:Nuitka 环境踩坑总结与 CI 自动化构建全指南 - cup11
    Python 打包成 EXE 并非易事,Nuitka 虽然性能卓越,但也隐藏着不少“暗雷”。本文基于 Meta Assistant 的开发实践,解决了 Nuitka `sys.executable` 未能检测是否已打包、CI 构建失败 `FATAL: Nuitka does not work in ...  ( 2 min )
    HagiCode 中 AI 提交使用的提示词:设计思路与实现拆解 - Newbe36524
    HagiCode 中 AI 提交使用的提示词:设计思路与实现拆解 当你把一堆乱七八糟的改动丢给 AI 让它帮你提交时,背后到底发了一段什么样的提示词给模型?为什么提示词要写成那个样子?这篇文章把 HagiCode 里真正驱动"AI 提交"的提示词拆给你看。 背景 用 AI 辅助开发这事,其实也算是经  ( 3 min )
    Claude Code 跨电脑会话上下文迁移完全指南(附实战案例) - only赟
    换电脑后如何无缝延续 AI 辅助开发?一份手把手的迁移手册,涵盖原理、方案与避坑细节。 目录 一、为什么要迁移会话? 二、先搞懂 Claude Code 把数据存哪了 路径编码规则(非常重要) 三、两种迁移方案 方案 A:相同路径(最简单) 方案 B:路径不同(通用方案) 实战场景 迁移包结构 四、  ( 3 min )
    从 OpenClaw.NET 的 /loop 实现,看 Loop Engineering 如何从概念走向工程实践 - 张善友
    一、引言 Loop Engineering 这个词最近又热起来了。 如果你从去年开始关注 AI 工程化领域的动态,大概已经习惯了这种概念迭代的节奏——Prompt Engineering 还没完全消化,Context Engineering 就登场了;Harness Engineering 的论文刚  ( 4 min )
    3min手搓一个帮助文档,很合理吧! - 神仙别打架
    最近在加班加点的做一个Token工厂, 模式类似与京东:有自营算力产生的模型,也外接第三方旗舰模型。 作为Token聚合分发平台,帮助文档是刚性需求。 Docsify是一个将Markdown文件转换为单页面静态网站(无需任何构建过程)的开源项目,实测下来3min就能构建并搭建一个 帮助文档站点。 1  ( 3 min )
    Adaboost算法原理与计算实例 - 王新文
    Adaptive Boosting(自适应提升)基于Boosting思想实现的一种集成学习;算法核心思想是:通过逐步提高那些被前一步分类错误的样本的权重来训练一个强分类器。 本章会讲解Adaboost算法推导过程,并用一个数值实例带着读者计算,使读者能够完全理解这个训练弱分类器至强分类器构建的过程。  ( 2 min )
    15天学会AI应用开发(七)有了大模型为什么还要引入RAG - aqi00
    ​前面的文章依次介绍了如何截断历史会话的对话记录,包括按照记录数量截断、按照Token长度截断,以及浓缩为摘要截断等等,其中摘要操作又分为三大类: 1、使用第三方的摘要库对文本摘要; 2、使用在线大模型对文本摘要; 3、使用离线大模型对文本摘要; 以上对会话记录的各种处理操作,统称为AI应用的上下文  ( 1 min )
    从 豆包 到 Codex CLI:一名普通开发者的 AI 工具进化路线 - IT王师傅
    一、前言 作为一名普通开发人员,我对 AI 工具的上手其实算比较晚。今年五月以前,基本还只是把 AI 当作聊天机器人使用。有时听同事提到一些关键词都听不懂,甚至不知道该如何提问,心里多少有些惭愧。直到最近因为一个契机,沉下心认真摸索了一段时间,才真正感受到 AI 工具的魅力和强大,也切身体会到它对生  ( 2 min )
    [MAF预定义的AIContextProvider-03]ChatHistoryMemoryProvider——赋予Agent从经验中学习的能力 - Artech
    LLM具有固化的知识,而且针对LLM的调用是完全无状态,永远只做一锤子买卖。但是交给Agent的任务基本上不可能一蹴而就,而且还希望Agent具有学习进化的能力。所以你会发现,很多的Harness手段的目的就是为了弥合两者之间的鸿沟。解决这个问题的基本的前提是:需要赋予Agent记忆。  ( 5 min )
    Siri要接入AI了,苹果手机上一句话让GPT写文案、DeepSeek写代码的时刻来了 - 岳小哥AI
    AI创造是大规模模式重组:文本靠自回归预测,图像靠扩散从噪声还原,代码靠Code LLM学习千万仓库。DeepSeekV4编程成本1美元,Suno文字生成歌曲。AI不知何为美但见过足够美,所以能生成美。  ( 1 min )

  • Open

    Demystifying Vulnerability Assessment: A SOC Analyst’s Perspective
    Continue reading on Medium »
    Atacar para defender
    Uma breve história sobre busca ativa por vulnerabilidades e como isso chegou no domínio cibernético Continue reading on Medium »
    Cazando Infraestructura C2.
    Continue reading on Medium »
    Hunting for Module Stomping Targets
    Swapping Guesswork for Process-Specific Precision Continue reading on Medium »
    Input Manipulation & Prompt Injection | TryHackMe
    Understand the basics of LLM Prompt Injection attacks. Continue reading on Medium »
    LLM Output Handling and Privacy Risks | TryHackMe
    Learn how LLMs handle their output and the privacy risks behind it. Continue reading on Medium »
    MobileHackingLabs — IOT Connect writeup
    So our goal for this lab is to exploit a vulnerability in the application in a broadcast receiver that allows us to activate the master… Continue reading on Medium »
  • Open

    What is Bug Bounty Hunting and Why Is It So Popular in 2026?
    From hobbyist hackers earning beer money to elite researchers pocketing millions, here is why ethical hacking has become a billion-dollar… Continue reading on Medium »
    SSRF Explained Like You Are Five (For Bug Bounty Beginners)
    One of the highest paying bugs, finally explained in plain words. Continue reading on Medium »
    I Tracked My First 90 Days of Bug Bounty Income. Here Are the Real Numbers.
    No hype. Just the spreadsheet. Continue reading on Medium »
    The One Thing Nobody Checks in Password Reset (And Why It Pays)
    Everyone tests the token. Nobody tests the after. Continue reading on Medium »
    How I Found a Server-Side Entitlement Issue That Allowed Free Users to Access Premium Features
    Continue reading on Medium »
    $900 Session Flaw: Deprovisioned Users Retain Access After Permission Removal
    Hi Everyone! Recently, while testing a SaaS platform (let’s call it ExampleCenter), I came across a very interesting access control issue… Continue reading on Medium »
    Complete Guide to Authentication Vulnerabilities
    Master authentication vulnerability exploitation and mitigation in Bug Bounty, featuring technical impact analysis and PoC examples. Continue reading on Medium »
  • Open

    AI Agent 30天速成|Day4 教学笔记 - 云淡风轻YangG
    AI Agent 全日制30天速成|Day4 教学笔记 今日总学习目标 理解Agent规划、任务拆解核心思想,掌握ReAct、Plan-Solve标准推理框架 基于前3天代码,实现自主任务拆解Agent(复杂问题自动拆分多子任务) 实现多工具串行/并行调度、任务状态管理、失败子任务重试机制 整合RA  ( 4 min )
    CUDA软件实现跨线程块同步 - RainbowC0
    本文针对CUDA设备端跨线程块同步需求,在计算能力低于9.0(如RTX 4090 8.9、V100 7.0)无法使用集群同步的情况下,通过分析协作组grid.sync()源码,揭示了其基于屏障变量原子自增与符号位翻转的软件实现机制。借鉴该原理,设计并实现了一种自定义的跨块同步函数sync_ctas,...  ( 3 min )
    LCA问题的高效解法:欧拉序 + RMQ 详解 - LiQirui
    \(\text{0x01}\). 问题定义 \(\text{Problem}\space \text{Definition}\). 最近公共祖先简称\(\text{LCA}\)。两个节点的最近公共祖先,就是这两个点的公共祖先里面,离根最远的那个。 为了方便,我们记某点集 \(S=\{v_1,v_2,  ( 4 min )
    没有网线,如何实现主机间互传数据 - only赟
    Windows 局域网文件共享实战:解决"账户被禁用"与"网络访问拒绝"问题 本文记录了一次 Windows 局域网文件共享的完整排错过程,从最基础的网络连通性检查,到本地安全策略配置,再到 NTFS 权限设置,最终实现两台电脑之间的文件传输。文中所有 IP 地址、用户名、主机名均为示例,实际操作时  ( 2 min )
    OpenClaw.NET 祭出 TokenJuice:Agent 时代的"Token 瘦身引擎",让 LLM 上下文不再膨胀 - 张善友
    TokenJuice 的解决方案也足够优雅——不引入额外的 LLM 调用,不牺牲确定性,不增加外部依赖,用静态规则 + 流水线处理,在 < 5ms 内完成 50%-95% 的压缩。Fail-open 的设计哲学、三层规则配置体系、NativeAOT 兼容性,都体现了对生产环境的深刻理解。  ( 4 min )
    深度学习进阶(三十一)FlashAttention:IO 感知的精确注意力 - 哥布林学者
    上一篇我们把现代大模型的五个核心模块拼回了 LLaMA 这个完整案例中,可以看到注意力机制仍然是计算最密集的部分。 而这个密集程度在序列变长时,会变得越来越恐怖: 标准自注意力的计算复杂度和空间复杂度都是 \(O(n^2)\):序列长度翻倍,计算量翻四倍,内存占用也翻四倍。 而在之前,我们用 KV  ( 4 min )
    Windows x64 构建 liboqs-java教程 - 云烟飘渺PM
    Windows 下构建 liboqs-java,实现 PQC 算法的调用 liboqs-java 本身不是一个纯 Java 库,它底层依赖 liboqs 这个 C 语言实现的后量子密码学库。也就是说,我们最终虽然是在 Java / Spring Boot 项目里调用 PQC 算法,但真正执行算法逻辑  ( 5 min )
    AI时代下,技术真的“平权”了吗?我们还要怎样“死磕”细节? - 杜劲松
    AI时代下,技术真的“平权”了吗?我们还要怎样“死磕”细节? 前言 最近这两年,各种AI编码助手和代码大模型呈爆发式增长。只要能写清楚Prompt,无论是生成一个带有Tailwind CSS的高保真数据大屏,还是快速搭建一个包含基础频控逻辑的短信平台MVP,甚至是写一段复杂的复杂不规则框客流统计算法  ( 1 min )
    CVE-2021-4034-pwnkit提权 - Doll_Marker
    漏洞核心原理一 CVE-2021-4034 (PwnKit) 的核心原理在于 pkexec 程序对命令行参数处理的一个越界读取(Out-of-bounds Read)漏洞,以及由此引发的环境变量写入。 参数处理越界 在 Linux 中,一个程序的 main 函数接收两个重要的参数:argc(参数数量  ( 4 min )
    同事做 PPT 比你快 5 倍,不是因为他努力,是因为他用了这套工具流 - 码哥字节
    上周一,我们组的会议纪要还要靠小陈手工整理,开完一个小时的早会,她要再花 40 分钟敲完发出来。同一天,她旁边的同事开完会直接站起来去喝咖啡,3 分钟后纪要已经在群里了。 不是谁效率高低的问题,是用没用对工具的问题。 我在后端做了 10 年,最近两个季度把职场 AI 工具摸了个遍。今天把我们实际跑通  ( 2 min )
    PHP 泛型之殇 泛型 RFC 提案被拒绝 - JaguarJack
    PHP 泛型之殇 泛型 RFC 提案被拒绝 PHP 开发者眼中的泛型现实 PHP 很可能不会迎来泛型。这并非什么新鲜事。一些核心开发者撰写了详尽的技术博客,阐述为什么运行时擦除式泛型不是个好主意,以及他们为何对当前的 RFC 投下反对票。 然而,PHP 社区中还有另一群人,他们非常支持这个 RFC,  ( 2 min )
    Pi Agent 对接实现:消息解析、重试与取消 - Newbe36524
    Pi Agent 对接实现:消息解析、重试与取消 接一个 CLI 形态的 AI agent,绕不开三件事:怎么把它私有的事件流翻译成稳定消息、失败之后到底谁负责重试、用户点取消时进程怎么干净地停。其实这三件事说穿了,不过是"分清职责"罢了,只是真做起来,才知道水有多深。 背景 最近我在做一个 AI  ( 3 min )
    AI大模型"幻觉"从何而来?解密GPT-4、DeepSeek一本正经胡说八道的真相 - 岳小哥AI
    DeepSeek能帮你解数学题、改合同、读文献,但它给出的答案里可能充斥着信誓旦旦但子虚乌有的事实和引用,这源于其推理能力的根本局限。  ( 1 min )
  • Open

    Breaking Down Two Simple Vulnerabilities That Exposed A School’s Admission Records
    No content preview
    BITSCTF 2026 Writeups | OSINT And Steganography / Forensics Challenges
    No content preview
    “Bug Bounty Bootcamp #48: OAuth + XSS ”
    The Ultimate Account Takeover One-Two Punch Continue reading on InfoSec Write-ups »
    Phone Numbers and Emails to Hidden Subdomains: The OSINT Acquisition Pipeline That Uncovered a…
    No content preview
    Brave Blue Team Lab (CyberDefenders)
    No content preview
    CRTA Exam Writeup — Passed | CyberWarFare Labs
    No content preview
    Web-RTA Exam Writeup — Passed | CyberWarFare Labs
    No content preview
    VulnHub — sunset: dawn | Full Walkthrough
    No content preview
  • Open

    Breaking Down Two Simple Vulnerabilities That Exposed A School’s Admission Records
    No content preview
    BITSCTF 2026 Writeups | OSINT And Steganography / Forensics Challenges
    No content preview
    “Bug Bounty Bootcamp #48: OAuth + XSS ”
    The Ultimate Account Takeover One-Two Punch Continue reading on InfoSec Write-ups »
    Phone Numbers and Emails to Hidden Subdomains: The OSINT Acquisition Pipeline That Uncovered a…
    No content preview
    Brave Blue Team Lab (CyberDefenders)
    No content preview
    CRTA Exam Writeup — Passed | CyberWarFare Labs
    No content preview
    Web-RTA Exam Writeup — Passed | CyberWarFare Labs
    No content preview
    VulnHub — sunset: dawn | Full Walkthrough
    No content preview
  • Open

    “Get Verification Code” Is Missing in iOS 18 and iOS 26; Here’s Where It Went
    If you have an Apple device running iOS 18 or iOS 26 and gone looking for the old Get Verification Code option under Settings → [user name] → Sign-In & Security, you’ve probably noticed it’s no longer there. A quick search turns up forum threads, support comments, and even GitHub issues all reaching the same […]  ( 6 min )

  • Open

    OCI 明明分配了 200G 系统盘,为什么 df 只看到 30G? - AlfredZhao
    在 OCI 上创建测试主机时,明明给启动盘分配了 200G,但进入系统后执行 df -h,根目录却只有 30G 左右。这种情况并不是磁盘没有分配成功,而是分区和 LVM 还没有把剩余空间用起来。 01 | 先看 200G 到底在不在 笔者环境中,df -h 看到的是文件系统大小: /dev/mapp  ( 2 min )
    AI Agent 全日制30天速成|Day3 笔记 - 云淡风轻YangG
    AI Agent 全日制30天速成|Day3 教学笔记 今日总学习目标 掌握向量嵌入(Embedding)原理、文本向量化、向量相似度计算 搭建简易本地向量库(内存版FAISS),实现文本入库、检索、删除 打通基础RAG检索链路:文档分片→向量化存储→用户提问召回片段→注入Prompt 结合Day2  ( 4 min )
    OpenClaw.NET 重大更新:Goal 机制登场,让 AI Agent 不再"半途而废" - 张善友
    OpenClaw.NET 的 Goal 机制代表了一种重要的工程思路:与其试图训练模型"不要偷懒",不如在运行时层面为 Agent 装上"导航系统"。  ( 5 min )
    【原创】如何利用网卡TSN硬件特性实现EtherCAT 确定性发帧与 DC 同步? - 沐多
    本文以 Rockchip RK3576/3588 + stmmac + IgH 为例,重点解析 TSN 在 EtherCAT 主站中的应用。针对 EtherCAT 发送抖动受 OS调度制约的痛点,讲清如何用 TSN 的 EST 门控与 Launch Time 把每帧发送时刻下沉到网卡 PHC硬件,实...  ( 8 min )
    【Agentic RL / 强化学习 / OPD】OpenClaw-RL 源码阅读笔记 --- (6)--- Rollout - 罗西的思考
    【Agentic RL / 强化学习 / OPD】OpenClaw-RL 源码阅读笔记 (6) Rollout 目录【Agentic RL / 强化学习 / OPD】OpenClaw-RL 源码阅读笔记 (6) Rollout0x00 概要0x01 Rollout基础1.1 概念1.1.1 标准 R  ( 6 min )
    CVPR 2026 | 全新强化学习框架 BeautyGRPO:重塑真实人像 - vivo互联网技术
    本工作提出全新强化学习框架BeautyGRPO。实验证明,BeautyGRPO 在真实场景的肌肤纹理重建与整体审美对齐上,全面超越了现有的专精修图方法与通用编辑大模型。  ( 2 min )
    House of apple2手法及部分源码解析 - firefly_star
    在2.34移除了hook函数之后,堆利用就少了一个大的攻击方向了。而House of apple这个手法就给我们提供了新的攻击方向:IOFILE结构体。虽然在house of orange就有所利用,之后也有一些利用了相关结构体的手法,但都没有House of apple条件简单。 House of  ( 8 min )
    Openai Codex 重大更新 已支持接入任意开源大模型 - JaguarJack
    Openai Codex 重大更新 已支持接入任意开源大模型 如果你还认为 Codex(OpenAI 推出的 AI 编程助手)只能搭配 GPT 系列模型使用,那你可能错过了近期最重要的一次更新。 Codex CLI 现在已经全面支持 OSS 模式(Open-Source Mode),任何兼容 Ope  ( 3 min )
    vi 删除指定范围的行,不用再反复按 dd - AlfredZhao
    目前各种强大的现代编辑工具可能已经让很多人忘了vi的用法,但笔者最近在配合调试客户环境的代码时,就遇到经常要直接在那黑乎乎的putty窗口下直接使用vi来操作,反复去删除指定多行代码的场景很是痛苦,加上远程网络时而还会不稳,连续的dd很不方便且容易误操作。所以就倒逼自己回来复习了下vi删除行的快捷操  ( 2 min )
    一套 Spec-First 的 AI 编程工作流 - 袋鼠云数栈前端
    本文由袋鼠云数栈 UED 团队分享。文章详细拆解了如何通过 Spec-First 工作流与 AGENTS.md 基础设施,让 AI 从“猜需求”转向“按方案精准编码”,教你用清晰的意图表达彻底释放 AI 的编程潜力。  ( 3 min )
    C#摸鱼实录——IoC与DI案例详解 - 假设狐狸有信箱
    IoC(控制反转)与DI(依赖注入) 开一个新的模块哈,在这个模块里面,我们主要讲一个东西如何使用,尽量不纠结概念,简单过过 之前老是被人说,是不是过于偏向于学院派了,所以从现在开始,我们将只关注能不能用 这个模块里面,我想讲的,大多数是在实际项目中常用的东西,例如一些NuGet 包,一个语法,或者  ( 8 min )
    JuiceFS PB 级数据同步优化:断点续传、安全与带宽控制 - JuiceFS
    在数据迁移、跨云同步与对象存储备份等场景中,juicefs sync 常用于执行大规模数据同步任务。当数据规模达到 TB 到 PB 级、对象数量达到数百万甚至数十亿级时,单次任务执行周期通常会延长到数小时甚至数天。 在这个过程中,系统运行过程中通常会逐步暴露出以下几类问题: 任务在网络抖动、进程异常
    信创全平台交叉编译工具链(全套C++17) - 书生执笔画浮沉
    这套工具链原本一直想藏着掖着作为个人的“私服”使用。 但转念一想,随着公司人员的流动,这些底层基建早晚也会流传出去。 与其等别人缝缝补补地发出来,不如自己做个彻底的整理和开源,也算是为国产化平台的快速发展做个贡献吧。 名词速查: loongarch64_OA:指代 Loongarch64 Ol  ( 2 min )
    .NET 程序在装有企业防水墙的电脑上运行报错:内存不足 - zhuweisky
    昨天有个客户反馈,在一台Win10电脑上运行 傲瑞通(OrayTalk),结果点击联系人时,右边的聊天窗口始终出不来。现象如下图所示: 一. 问题排查 于是,我们远程到客户的这台电脑上,发现这台电脑有装企业防水墙(驱动级自动文件加密),并且傲瑞通的日志内容有如下报错记录: System.Drawin  ( 1 min )
    别再「整理笔记」了——Karpathy 们已经在用 AI 养一个会自己长大的知识库 - 小龙虾不说话
    一、那个叫做"知识库"的数字坟场 先说一个你可能亲历过的场景—— 公司花了三个月搭了个知识库。Confluence 也好,飞书文档也好,SharePoint 也好。大家热血沸腾地上传制度文件、项目复盘、销售话术、产品 PRD。搜索框静静躺在页面顶部,仿佛只要它在,知识就在。半年后没人更新了。一年后搜  ( 2 min )
    5年,一个程序员是如何把私有化在线客服系统做到第一名的 - 升讯威在线客服系统
    一转眼 5 年过去了。今天对我来说同样是具有里程碑意义的日子,一个大客户经过长达近 3 个月的评估,测试,终于签单了。过程十分辛苦,甚至一度让我有点想放弃。  ( 1 min )
    并发性能优化复盘 - 胡云Troy
    上周服务治理系统的同事(注意不是客户)发现一个问题,他调用的 api 接口不稳定。有时候正常,有时候报错,关键是大部分时候都是报错。 更关键的是这个接口是我开发的。 好吧,理所当然的他来找我。第一时间“男人的第六感”告诉我是后端服务处理时间太长了,导致客户端在超时时间内未处理完成。要不把 timeo  ( 2 min )
    Android 实时音视频接入教程:媒体补充增强信息(SEI) - RTC实战笔记
    本文主要介绍了「媒体补充增强信息(SEI)」的基础接入流程,并结合官方文档梳理了环境准备、核心 API 调用、媒体处理和结果验证等关键环节。  ( 3 min )
    Ubuntu系统搭建企业级GitLab高可用主备代码仓库完整指南 - Johny_Zhao
    一、架构选型与规划 1.1 高可用主备架构方案 企业级GitLab高可用部署主要有以下几种主流方案: 方案一:DRBD + Pacemaker + Corosync(主备模式) 该架构为Active/Passive(主/备)模式,只有主节点运行GitLab服务并挂载存储,备节点通过DRBD实时同步数  ( 5 min )
    Electron 应用如何上架微软商店:从 MSIX 打包到商店提交 - Newbe36524
    Electron 应用如何上架微软商店:从 MSIX 打包到商店提交 其实 Electron 说到底,不过是个普普通通的 Win32 桌面应用罢了,可微软商店它只认 MSIX。这篇文章,就借着我们 HagiCode Desktop 实打实跑通的那套构建配置,把「注册开发者账号 → 打 MSIX 包  ( 3 min )
    Burp Suite+MCP+DeepSeek - Xiao-Hei
    工具准备: 1、Burp Suite v2025+(版本若较低可查看扩展商店中是否支持下载MCP server这个插件) 2、Cherry Studio 3、一条能用的API key 这里咱使用的是Burp Suite v2026 API咱使用的是https://platform.deepseek.  ( 1 min )
    gt-checksum v4.0.0 新功能解读系列文章(3):反向回滚 SQL——修复可审计、可回退 - GreatSQL
    gt-checksum v4.0.0 新功能解读(3):反向回滚 SQL ✅ 一键开启 genRollSQL=ON,修复 SQL 自动生成反向回滚操作 ✅ DELETE→INSERT、INSERT→DELETE、TRUNCATE 三种映射策略,覆盖全场景 ✅ 回滚 SQL 只写文件不在线...  ( 4 min )
    Hermes Agent Skill Runtime 架构拆解:让 AI Agent 不再从零开始 - AI小老六
    拆解 Hermes 如何把执行轨迹沉淀为技能、记忆和自修复闭环,让 Agent 真正积累经验。 原文链接:AI 小老六 导语 很多 Agent 产品有一个尴尬的问题:它们看起来每天都在工作,实际上每天都从零开始。 用户让它处理第 1 个复杂任务时,它会试错;第 20 次遇到类似任务时,它还在试错;到  ( 2 min )
    MiniMax-M3 开源实测:部署、推理与基准测试全记录 - GPUStack
    MiniMax-M3 是 MiniMax 最新开源的原生多模态大模型,约428B 总参数/23B激活参数,原生支持1M上下文。本文基于 GPUStack 与VLLM,演示从镜像与权重准备、模型部署、对话实测到基准测试的完整流程,并实测了基于 EAGLE3 的投机解码加速。  ( 3 min )
    NineData 社区版 V5.1.0 正式发布!AI Agent 能力全面释放,新增 48 种数据复制和对比链路 - NineData
    NineData 社区版 V5.1.0 这次更新,核心还是围绕 AI 数据库运维能力继续升级。ChatDBA 新增支持类 PostgreSQL、Oracle 和 SQL Server 数据库的全链路性能诊断,SQL 智能优化继续扩到十多种数据库,SQL 执行与任务 Skill 及对应 OpenAPI...  ( 2 min )
    记一次微服务架构下的HTTP请求头“大小写”丢失排查之旅 - 杜劲松
    在最近的微服务排障过程中,业务方反馈了一个诡异的问题:客户端发起请求时,明确携带了驼峰写法的请求头(如 appKey: asd),但请求经过反向代理和网关,到达后端具体的 Spring Boot 业务服务时,业务代码里取出来的请求头全变成了小写(appkey: asd)。 面对这种链路较长的问题,最  ( 3 min )
    测试必学:推荐4个测试用例设计最值得掌握的Agent Skill,几乎覆盖所有用例设计场景! - 狂师
    做软件测试的朋友都清楚,测试用例设计有多么重要。它直接决定了测试覆盖是否全面,也直接影响了后续整体测试流程的质量。 传统模式下,大家手动一条一条编写用例,不仅耗时费力,还经常因为考虑不周,漏掉异常场景、边界条件,给线上质量埋下隐患。 如今借助 AI + Agent Skill 能力,我们可以把用例设  ( 3 min )
    如何编写一个SpringBoot项目告警推送的Starter - 虚无境
    前言 最近有一点时间了,于是便开始做以前自己想做但是没有完成的事情。之前我其实就一直想写一个通用一点的告警推送组件,把项目里的异常信息、慢请求、状态码异常、JVM 指标,甚至数据库慢 SQL 这些内容统一收集起来,然后直接推送到飞书、钉钉、企业微信这类 IM 工具里。 这样做有两个好处,一个是出了问  ( 5 min )
    [MAF预定义的AIContextProvider-02]AgentSkillsProvider——将Agent Skills引入MAF - Artech
    Skills针对Agent的重要性是不言而喻的。从本质上讲,Agent Skills就是随着用户与LLM对话的推进,动态加载被称为Skill作为提示词的一种机制。在大部分实现中,Skill的内容会被封装成角色为Tool的消息被添加到对话历史中,因为这样可以借助针对对话历史的压缩实现对老旧Skill的...  ( 10 min )
    28. Agent 执行到一半想暂停?用 interrupt 给它设个“关卡“! - 老陈说编程
    Agent 跑起来之后,大多数时候我们希望它一气呵成把活干完。但总有些场景不太一样——比如 Agent 要调用一个会删文件的工具,你总得让人确认一下再动手吧?LangGraph 的 interrupt 机制就是干这个的:在指定节点前或节点后暂停执行,等人给了信号再接着跑。 动画视频在《28. Age  ( 2 min )
    读懂计算机视觉CV、语言感知(ASR/TTS)、多模态,就能理解AI是如何“看到”与“听到”世界的 - 岳小哥AI
    当Transformer架构让AI能‘同时看到’整段文字,理解其中的讽刺与隐喻时,语言处理已不再只是识别,而是真正的语义理解。  ( 1 min )
  • Open

    Active Directory Attacks — NTLM Relay Attack
    Continue reading on Medium »
    Information Gathering — Web Edition (Skill Assessment — Walkthrough)
    To complete the skills assessment, answer the questions below. You will need to apply a variety of skills learned in this module… Continue reading on Medium »
    Lightweight Asymmetric Encryption for C2 Implants: A Red Teamer’s Guide from XOR to Rabin
    A practical walkthrough of encryption choices for red team implants — from XOR and AES-CTR to Rabin key encapsulation — with a… Continue reading on Medium »
    Nmap Through the Lens of Networking — Part 2
    OS Detection, NSE, Timing, Evasion, and How Experienced Practitioners Actually Think Continue reading on MeetCyber »
    IP address of vulnerable metasploit IP
    Ping (checking connection with the vulnerable machine) Continue reading on Medium »
    Mobile Application Security | TryHackMe
    Learn mobile application pentesting through static analysis, MobSF, and the OWASP Mobile Top 10. Continue reading on Medium »
    Silent Monitor | TryHackMe
    Enumerate a running internal service, exploit a vulnerable web application, pivot through the system, and crack your way to root. Continue reading on Medium »
  • Open

    The AI Bug Bounty Gold Rush: Why 2026 Is the Best Time to Start (Before Everyone Else Does)
    A field guide for the paranoid, the curious, and the slightly unhinged Continue reading on Medium »
    When a Debug Log Hands Out Every Password You Type
    A forgotten debug statement and a default password are all it takes. Continue reading on Medium »
    From Debugging to Hacking: Using Chrome DevTools Like a Bug Hunter
    Most bug bounty writeups treat Chrome DevTools like a passive utility — a way to inspect network requests, peek at console logs, and maybe… Continue reading on InfoSec Write-ups »
    Breaking Down Two Simple Vulnerabilities That Exposed A School’s Admission Records
    By: Kavin Jindal (@Klevr) Continue reading on InfoSec Write-ups »
    How to Detect SSH Brute-Force Attacks Using Wazuh
    Remote administration is an essential part of managing Linux systems, and SSH is one of the most widely used protocols for this purpose… Continue reading on Medium »
    “Bug Bounty Bootcamp #49: Recon Is Not Stalking”
    How to Find Every Subdomain, Asset, and Hidden Folder Your Target Forgot About Continue reading on InfoSec Write-ups »
    Building a Practical CTI RAG Assistant: An AI-Powered Threat Intelligence Workflow
    Threat intelligence teams collect huge amounts of data every day — CVE feeds, malware reports, IOC feeds, vendor research, and incident… Continue reading on Medium »
    Authentication Bypass via Insecure redirect_url Parameter Leading to Account Takeover
    Introduction Continue reading on Medium »
    The MFA Bypass That Wasn’t an MFA Problem: A Lesson in Broken API Authorization
    A bug bounty story about frontend trust, missing backend enforcement, and why attackers never use your UI. Continue reading on Cyber Security Write-ups »
    I Found a Bug That Looks Harmless But Can Still Get You Paid
    Here is everything I know about HTML Injection, how I test for it, how people bypass filters, and what it actually scores on CVSS Continue reading on OSINT Team »
  • Open

    HTTP/2 sessions never clean up after GOAWAY on invalid protocol errors
    Node.js disclosed a bug submitted by pimterry: https://hackerone.com/reports/3658225
    Permission Model Bypass via `process.report.writeReport()` Path Misvalidation
    Node.js disclosed a bug submitted by suul: https://hackerone.com/reports/3692858
    Reflected XSS in AI Chat Bot Greetings at help.shopify.com via Markdown Image Rendering
    Shopify disclosed a bug submitted by saltymermaid: https://hackerone.com/reports/2509022 - Bounty: $1600
  • Open

    HTTP/2 sessions never clean up after GOAWAY on invalid protocol errors
    Node.js disclosed a bug submitted by pimterry: https://hackerone.com/reports/3658225
    Permission Model Bypass via `process.report.writeReport()` Path Misvalidation
    Node.js disclosed a bug submitted by suul: https://hackerone.com/reports/3692858
    Reflected XSS in AI Chat Bot Greetings at help.shopify.com via Markdown Image Rendering
    Shopify disclosed a bug submitted by saltymermaid: https://hackerone.com/reports/2509022 - Bounty: $1600
  • Open

    Downloading iOS 26 iCloud Backups
    Elcomsoft Phone Breaker 11.2 adds the ability to download iCloud backups created on devices running iOS and iPadOS 26 and, by extension, iOS/iPadOS 27 beta. With this release, Elcomsoft Phone Breaker becomes the first and only third-party tool capable of pulling these backups from Apple’s cloud. That might read like a routine compatibility update. It isn’t. […]  ( 6 min )
  • Open

    Slort — RFI via PHP allow_url_include + Writable Scheduled Task Binary to Administrator | OffSec PG…
    No content preview
    “Bug Bounty Bootcamp #47: Account Takeover 101 — How to Steal Everyone’s Account (Legally)”
    You don’t need to be a hacker in a hoodie. Just a missing IDOR, a leaky invite link, or a mass-assignable “role” field — and suddenly… Continue reading on InfoSec Write-ups »
    Build an IDOR Vulnerability Lab: Why WHERE Clauses Don’t Protect Your API.
    No content preview
    BEARCAT CTF 2026 WRITEUPS
    No content preview
    I almost ordered a product for free. (Business Logic Vulnerability)
    No content preview
    Building a Hackbot for Bug Bounties — Auth Testing Subagent Setup
    No content preview
    “Bug Bounty Bootcamp #46: Not Allowed From Your IP?”
    — How to Spoof, Brute-Force, and Mass-Assign Your Way Past Authentication Walls” Continue reading on InfoSec Write-ups »
    TryHackMe — Blog CTF | Full Write-Up
    No content preview
    VulnHub — Shenron: 1 | Full Walkthrough
    No content preview
    I Pentested a Real CRM System and Found 4 Critical Vulnerabilities — Here’s the Full Attack Chain
    No content preview
  • Open

    Slort — RFI via PHP allow_url_include + Writable Scheduled Task Binary to Administrator | OffSec PG…
    No content preview
    “Bug Bounty Bootcamp #47: Account Takeover 101 — How to Steal Everyone’s Account (Legally)”
    You don’t need to be a hacker in a hoodie. Just a missing IDOR, a leaky invite link, or a mass-assignable “role” field — and suddenly… Continue reading on InfoSec Write-ups »
    Build an IDOR Vulnerability Lab: Why WHERE Clauses Don’t Protect Your API.
    No content preview
    BEARCAT CTF 2026 WRITEUPS
    No content preview
    I almost ordered a product for free. (Business Logic Vulnerability)
    No content preview
    Building a Hackbot for Bug Bounties — Auth Testing Subagent Setup
    No content preview
    “Bug Bounty Bootcamp #46: Not Allowed From Your IP?”
    — How to Spoof, Brute-Force, and Mass-Assign Your Way Past Authentication Walls” Continue reading on InfoSec Write-ups »
    TryHackMe — Blog CTF | Full Write-Up
    No content preview
    VulnHub — Shenron: 1 | Full Walkthrough
    No content preview
    I Pentested a Real CRM System and Found 4 Critical Vulnerabilities — Here’s the Full Attack Chain
    No content preview

  • Open

    2FA Bypass: The Complete Guide to Breaking Two-Factor Authentication (And How to Fix It)
    Why I Started Caring About 2FA Bugs Continue reading on OSINT Team »
    Authenticating a PayPal Notification Is Not the Same as Trusting What It Says (CVE-2026–9189)
    CVE-2026–9189. Public on cve.org since 2026–05–29. CVSS 5.3 (CWE-345). Found and reported by me, Muni Nitish Kumar Yaddala, through… Continue reading on Medium »
    SSTImap & SSTI Prevention: Letting the Tool Do the Work
    We did it manually. Now let’s see what happens when you hand it to a tool. Last part of the SSTI series. Continue reading on System Weakness »
    Unauthenticated IDOR on NASA’s GitLab Instance — From Recon to Bypass
    Target: gitlab.smce.nasa.gov Program: NASA VDP (Bugcrowd) Severity: P3 — Broken Access Control / IDOR Status: Resolved Continue reading on Medium »
    “Bug Bounty Bootcamp #48: OAuth + XSS ”
    The Ultimate Account Takeover One-Two Punch Continue reading on InfoSec Write-ups »
    SSTI Exploitation in Twig: Same Idea, Different Language
    Jinja2 was Python. Twig is PHP. The vulnerability is the same — the path looks completely different. Part 3 of the SSTI series. Continue reading on System Weakness »
    The 5 AM Club Is a Lie. Here’s What Actually Works for Bug Hunters.
    Stop trying to wake up early. Start hunting when you actually think. Continue reading on Medium »
    How a Resend Invitation Feature Led to Unauthorized Account Creation
    Hello fellow bug bounty hunters, hope you guys are doing well and finding lots of bugs. Continue reading on Medium »
    Guía Compelta de Vulnerabilidades de Autenticación
    Descubre cómo explotar y mitigar vulnerabilidades de autenticación críticas en Bug Bounty con ejemplos técnicos y análisis de PoC. Continue reading on Medium »
  • Open

    瞬时极限性WebApp实验室:无限接近如何被“看见” - 郝hai
    在微积分的世界里,最深刻的思想往往隐藏在最短暂的瞬间。当时间趋近于零、距离无限缩小、分割不断细化时,一个充满连续性与变化性的数学宇宙逐渐显现。导数诞生于割线向切线逼近的瞬间,积分形成于无数微小面积不断累积的过程,而极限则是连接有限与无限、离散与连续的核心桥梁。然而,这些过程往往被压缩成静态公式,难以  ( 3 min )
    OpenClaw.NET MetaSKILLs 系统深度解析:AI Agent 正在学会「自己给自己写技能」 - 张善友
    接上一篇文章OpenClaw.NET 上线 MetaSkills :软件工程第一性原理的工业级实践, 本篇文章是 MetaSkills 系统深度解析 —— 当 AI 不仅能执行任务,还能编排任务、创造任务,这意味着什么? 一个让工程师崩溃的早晨 想象一下这个场景: 周一早上 9 点,你打开公司内部的  ( 6 min )
    浏览器缓存之【基础键值存储】:Local storage 和 Session storage - 橙子家
    〇、前言 虽然在日常口语中我们常把浏览器中 Application -> Storage 模块中的多种类型统称为“浏览器缓存”,但在计算机科学与 Web 开发的专业定义中,实际上包含了“存储”与“缓存”两个截然不同但又紧密相关的概念体系。 其中本文将着重介绍的 Local storage 和 Ses  ( 4 min )
    高精度算法 - myLv
    一、为什么需要高精度运算 普通的整数类型(int、long long)都有固定的取值范围,例如64位 long long 最大只能存储约 9×10¹⁸ 的数值。当我们需要处理几十位、上百位甚至更长的大整数时(比如大数阶乘、密码学计算、超长数值运算等场景),内置类型会发生溢出,无法正确存储和计算。高精  ( 4 min )
    每月 20 美元的 Claude Pro 会员到底能消耗多少 Token 及额度 - 曾左
    统计时间:2026 年 6 月 3 日下午 15 点至 19 点。 一、先说结论 推测: (1)在五小时计费周期内,最大可消耗 18 美元的额度和 4000 万 Token(使用 Sonnet 模型)。 (2)每周有 10 个五小时计费周期,最大可消耗 180 美元的额度和 4 亿 Token(使用  ( 2 min )
    Electron 如何调用 Windows 原生 API - Newbe36524
    Electron 如何调用 Windows 原生 API 在 Electron 应用里调用 Windows 原生 API,就像想看海却只能看地图。不过折腾了一阵,总算摸索出几条路,写下这篇文章算是留个纪念,也给后来者指个方向。 背景 做 Electron 桌面应用的时候,难免要和操作系统打打交道。在  ( 3 min )
    OpenStack 1.7.2 & Ceph 9.2.1 运维命令 - 人生的哲理
    OpenStack 1.7.2 & Ceph 9.2.1 运维命令,加载认证文件,实例管理,服务状态检查,计算服务 (Nova),网络服务 (Neutron),存储服务 (Cinder),集群整体资源概况,计算节点维护,RabbitMQ 连通性测试,Keystone 项目管理,RAID 检查,软件 ...  ( 6 min )
    更换 Kingbase V9 License 踩坑记 - Theo·Chan
    更换 Kingbase V9 License 踩坑记 本地测试环境使用KDTS迁移数据时候遇到连接用户过多错误,原因是初始化Kingbase实例时候指定的开发版license限制最大连接数10。 我用官网的更换 license 文件 - License管理 | KingbaseES方法更换失败: [  ( 2 min )
    UI Output Protocol 架构拆解:Markdown、HTML 和 UI DSL 如何分工 - AI小老六
    拆解 AI 产品输出从文本到工作台的协议分层:Markdown 写文档,HTML 承载页面,UI DSL 接住操作。 原文链接:AI 小老六 导语 最近不少 AI 产品开始把回答做得越来越"像页面":有卡片、有筛选器、有图表,也有可点击操作。于是一个问题被反复拿出来讨论:​HTML 会不会替代 Ma  ( 2 min )
    同样是库文件,嵌入式静态库和动态库差异到底在哪? - Sharemaker
    本文对比分析了嵌入式开发中静态库与动态库的核心差异,从内存/磁盘占用、编译更新灵活性、模块化加载及部署维护四个维度展开。静态库编译时完整嵌入可执行文件,导致资源冗余但部署简单;动态库通过共享机制节省内存,支持独立更新和按需加载,更适合资源受限且需频繁迭代的场景。文章结合TWS耳机等典型案例,提出混合...  ( 4 min )
    一分钟学会 C++ 标准模板库智能指针 - 天恩软件
    什么是 C++ 智能指针 C++11 标准模板库(STL)引入了现代 C++ 中管理动态内存的核心工具——智能指针。它们位于<memory>头文件中。 智能指针旨在解决 C++ 裸指针带来的手动内存管理问题(如内存泄漏、悬垂指针、异常安全等)。智能指针的行为类似于普通指针,但能够自动释放所拥有的  ( 1 min )
    Vue 模板编译:HTML 是怎么变成 JS 的? - 微山湖上静悄悄
    你写的模板,浏览器不认识 用 Vue 写代码的时候,你一定写过这样的模板: <template> <div id="app"> <p>{{ message }}</p> <button @click="handleClick">点击</button> </div> </template> 看起来很直  ( 2 min )
    C#实现控制台交互式操作 - yi念之间
    前言 上一篇文章《C#实现控制台多区域输出》中,我们介绍了如何利用 Console 实现类似 Agent CLI 的多区域动态界面。 如果说多区域布局解决的是: 界面如何展示的问题 那么本文要讨论的则是另外一个问题: 用户如何与界面交互 相信体验过Claude Code、OpenCode、Herme  ( 3 min )
    [MAF预定义的AIContextProvider-01]TextSearchProvider——RAG在MAF中的实现 - Artech
    作为最核心的AIAgent,`ChatClientAgent`构建了一个管道与LLM交互。为了让管道的输出更符合我们的需求,有两个主要的途径:输入增强(Input Enhancement)和输出增强(Output Enhancement),前者通过通过改变输入让LLM返回更高质量的内容,后者则直接对...  ( 4 min )
    世界杯一粒进球被吹掉,背后可能有多少 AI? - 小七-七牛开发者
    世界杯进球被吹?背后是VAR+AI的精密协作:摄像机追踪29个身体点、球内传感器锁定传球瞬间、AI自动生成越位线与3D动画,再经VAR复核、主裁终裁。AI负责“测得准”,裁判负责“判得明”——技术让判罚更透明,却未消除规则语境下的争议。  ( 1 min )
    短短三天,英伟达和微软同时宣告:真正的AI电脑来了! - 岳小哥AI
    Agent和具身智能,一个扎根在虚拟世界里替你跑任务,一个走进物理世界替你动手操作。这篇文章,我们来聊聊这两条AI执行能力进化的核心路径。  ( 1 min )
  • Open

    Footprinting Lab — Hard (Walkthrough)
    The third server is an MX and management server for the internal network. Continue reading on Medium »
    MobileHackingLabs — Secure Notes writeup
    This was another pretty nice lab I took on involving an exposed content provider that we’re supposed to brute-force for a pin. In the end… Continue reading on Medium »
  • Open

    Authenticated Elasticsearch Painless script execution via Query.search.sort_query on hackerone.com/graphql
    HackerOne disclosed a bug submitted by brumbelow: https://hackerone.com/reports/3694007 - Bounty: $7000
    verify-release rebuilds from the tarball under verification, enabling pre-check command execution and false OK for a malicious curl release tarball
    curl disclosed a bug submitted by argareksapatii: https://hackerone.com/reports/3802645
  • Open

    Authenticated Elasticsearch Painless script execution via Query.search.sort_query on hackerone.com/graphql
    HackerOne disclosed a bug submitted by brumbelow: https://hackerone.com/reports/3694007 - Bounty: $7000
    verify-release rebuilds from the tarball under verification, enabling pre-check command execution and false OK for a malicious curl release tarball
    curl disclosed a bug submitted by argareksapatii: https://hackerone.com/reports/3802645

  • Open

    OpenClaw.NET 上线 MetaSkills :软件工程第一性原理的工业级实践 - 张善友
    当系统平稳运行时,你是发号施令的架构师,驾驭 LLM 和云原生框架在宏观上狂奔。但当服务器在深夜崩溃,当 AI 给出的解释全是指鹿为马的幻觉时,你必须有能力瞬间推开抽象的胡言乱语,化身成底层机器的同类——读十六进制内存快照,抓 TCP 报文头部,在无数死锁的线程栈里,徒手掐死那只 bug。 这,才...  ( 2 min )
    C# .NET 周刊 |2026 年 4 月 4 期 - InCerry
    C# .NET 周刊 |2026 年 4 月 4 期 2026-04-26 dotnet_week_26_4_4 国内文章 写 EF Core 查询,90% 的人第一步就错了:刚子教你避开所有坑 https://www.cnblogs.com/shenchuanchao/p/19889187/ef-  ( 3 min )
    Electron 桌面应用如何接入 Microsoft Store 订阅与永久许可证 - Newbe36524
    Electron 桌面应用如何接入 Microsoft Store 订阅与永久许可证 当你的 Electron 应用要进 Microsoft Store 卖订阅和永久授权,WinRT 那一套商业 API 到底怎么干净地接到业务里?这事儿说起来也算一桩旧梦,我们在 HagiCode Desktop 里  ( 4 min )
    【Azure AI Search】 stopword 是什么,为什么它会影响搜索结果? - 编码者卢布
    问题描述 在 Azure AI Search 中搜索 in brief 时,结果数量有时会比预期多很多。仔细查看返回结果文本,会发现有些文档和 brief 的关系很弱,反而只是命中了 in 这类高频词。 这不是数据问题,也不是 Azure AI Search 的 bug。关键在于一个很容易被忽略的地  ( 2 min )
    Solon v4.0 正式发布,高考记忆版 - 带刺的坐椅
    摘要: Solon v4.0 正式发布,标志着这一 Java 轻量级框架进入新阶段。版本核心围绕"做减法",清理历史弃用项以保持简洁,同时优化 AI 生态:将 skill 概念统一为 talent,新增沙盒隔离等能力,并升级协议支持。此外,推动第三方插件回归官方维护,提升生态可持续性。升级兼容性良好...  ( 2 min )
    【Agentic RL / 强化学习框架】Miles 项目技术分析---(2)--- 关键技术 - 罗西的思考
    【Agentic RL / 强化学习框架】Miles 项目技术分析 (2) 关键技术 目录【Agentic RL / 强化学习框架】Miles 项目技术分析 (2) 关键技术0x00 概要0x01 agentic_tool_call1.1 问题1.2 解决方案1.3 框架自动化的主要流水线1.4 深  ( 17 min )
    Flink 实时数仓开发实战:像后端那样 CI/CD - tonyabasy
    概览 第一篇我们解决了"怎么写"——一条 flink run 跑起完整的 Multi-Statement SQL 脚本。这一篇解决"怎么管":让 Flink SQL 作业的研发流程具备和 Java 后端同样的工程能力——可检测、可追溯、可回滚、自动化。 本文将深入 Flink SQL Validat  ( 3 min )
    Vibe Coding--初识AI编程 - NE_STOP
    走进AI编程 AI是如何理解和生成代码的? token化:AI不像人一样一个字一个字的读代码,而是把文本切成一个个小块,叫做token。 例如,Hello World 会被切成 Hello 和 World 两个 Token。中文的 你好世界 可能被切成 你好 和 世界 两个 Token,也可能被切成  ( 4 min )
    GEPA 架构拆解:让 Prompt 和 Skill 优化不靠玄学 - AI小老六
    拆解 GEPA 如何用轨迹反馈、Pareto 前沿和模块合并,让 Prompt 优化更稳、更可审计。 原文链接:AI 小老六 Agent 系统里的 Prompt 很少是一次写对的。更常见的情况是,线上 case 出错以后,人去翻日志、看工具调用、读模型输出,再手工改一版 Prompt 或 Skill  ( 3 min )
    [开源] Memory Checker:极致轻量的 Windows 托盘内存监测工具,告别内存焦虑 - cup11
    16GB 内存又爆了,任务管理器又在关键时刻死活打不开?渲染、测试时总是担心 OOM 导致系统崩溃?Memory Checker 是一款专为 Windows 打造的极致轻量开源托盘控件。仅 15MB 安装包,通过实时变色的托盘图标,让你对内存占用“一眼定心”,告别死机焦虑。  ( 1 min )
    码哥用扣子搭了一套自动跟进工作流,每天省2小时这不是夸张 - 码哥字节
    上周三下午,一个做销售的朋友发消息问我:"有没有办法让客户填完表单之后,自动把信息记到表格里,同时给我发个钉钉通知?现在是我一个个手动拷贝的,有时候忘了,客户那边还会投诉说没有跟进。" 我问他一天这类信息大概多少条。他说:"少说也有二三十条,每条处理加确认大概要 2-3 分钟,算下来一天光这个就要花  ( 2 min )
    ZP7 和 ZP8 - jwang106
    ZP7 和 ZP8 这两个关键质量控制点(Zählpunkt)的核心区别。 结合汽车生产的实际流程,我们可以对您的总结做进一步的补充和深化: ZP7:总装工序的“内部自检” 您的定义:车辆已完成组装并刷写软件,但尚未通过下线测试(EOL)。 执行主体:通常由总装车间的现场生产人员执行。 核心目的:侧  ( 1 min )
    掏出手机就能搭个 WebDAV 同步服务器?这操作有点香 - 一名程序媛呀
    还在为不同软件间的配置同步头疼吗?本文带你用一部安卓手机,通过 Termux 和一款轻量级神器 gowebdav ,零成本搭建私人 WebDAV 服务器。无需公网 IP,不用注册第三方服务,十分钟就能让 Tampermonkey、KOReader 等软件实现无缝同步。内含详细避坑指南和方案对比,让你...  ( 2 min )
    给博客加入黑客帝国背景字幕 - 北七星
    〇、效果如本页的背景所示,黑客帝国的字幕 一、本功能需要开通博客园的JS权限 二、在博客设置 页脚HTML代码 贴入如下代码 <!--黑客帝国背景,移动端不加入--> <div><canvas id="matrixBG" ></canvas></div> <script src="https://f  ( 2 min )
    程序员和软件还有前途吗 —— 从 NocoBase 收入再翻倍谈起 - NocoBase
    原文链接:https://www.nocobase.com/cn/blog/future-of-software-programmers-revenue-doubled 前情 半年前,我们在 NocoBase 发布 2.0 的时候写过第二篇总结文章:《没有 AI,没有融资,一个开源项目的真实收入》,  ( 2 min )
    如何使用 Python 设置 Excel 单元格数字格式 - LAYONTHEGROUND
    在自动化报表生成与数据处理场景中,数值的展示方式直接影响文档的可读性与专业性。千分位分隔、货币符号、百分比显示、日期格式化等效果,均通过 Excel 的数字格式机制实现。本文将介绍如何通过 Python 代码精确控制 Excel 单元格的数字显示格式,示例基于 Free Spire.XLS for  ( 2 min )
    Event-Driven Agent 实战:Prometheus 告警 → LLM → Tool Calling → 自动恢复 - it排球君
    前言 本文主要描述Event-Driven开发中的ReAct模式,并且使用一个demo,彻底搞懂怎么在实际工作中使用Event-Driven模式 话不多说,我们开始 代码结构 代码地址 . ├── agent.py # EventDrivenAgent 主逻辑,负责接收事件、调用 LLM、执行工具  ( 2 min )
    周一上线|瑞幸把咖啡做进 CLI,Fable 5 短暂登场,Stonk Rider 骑上 K 线图 - 小七-七牛开发者
    一边,模型和开发者工具还在继续往真实工作流里走:Gemini 3.5 Live Translate 开始做近实时语音翻译,Kimi-K2.7-Code 开源,DiffusionGemma 探索更快的文本生成,Codex 和 Chrome DevTools 也在继续给 Agent 补浏览器能力。 另一...  ( 3 min )
    把 Agent 效果从 “感觉” 变成 “可验证” - AaronLin
    TLDR:用 A/B 盲测和独立评估,让约束迭代可验证 问题 约束迭代最大的问题,是写了不代表有效。很多时候我们只是觉得 Agent 变好了,但缺少稳定的验证方法 做法 我的流程分为四步: 修改约束文档,形成改后版本 按官方最佳实践 review,先排除明显不合理的写法 派 Subagent 做 A  ( 1 min )
    【YFIOs】用C#开发硬件之设备上云 - 刘洪峰AIoT
    将YF3300-ESP32S3设备数据上传到叶帆物联网平台需要使用 YFLink 协议通过 MQTT 进行通信。本章以完整项目 YeFanIoTTest 为例,介绍从设备初始化到数据上传的完整实现。  ( 46 min )
    上周热点回顾(6.8-6.14) - 博客园团队
    热点随笔: · 折腾了一下午,终于让Codex用上了DeepSeek (程序员晓凡) · 小米开源编程助手 MIMO Code 简介和简单使用测试 (橙子家) · 独立开发者最容易低估的,不是开发成本,而是维护成本 (Solo社区) · C#实现控制台多区域输出 (yi念之间) · 做 AI 工具出  ( 1 min )
    一文读懂AI落地的三驾马车:大语言模型、RAG、多模态AI - 岳小哥AI
    你有没有遇到过这种情况:跑去问AI一个问题,它回答得滔滔不绝,逻辑清晰,听起来像模像样,但你就是有一种"哪里不对劲"的感觉。要么数据过时,要么答案与你的实际情况毫不相干,要么语气信心满满,内容却一本正经地偏了题。 这种"AI大聪明"的尴尬,困扰过太多人。 其实问题的根源不在于AI笨,而在于技术架构本  ( 1 min )
  • Open

    Footprinting Lab — Medium (Walkthrough)
    This second server is a server that everyone on the internal network has access to. In our discussion with our client, we pointed out that… Continue reading on Medium »
    Active Directory Attacks — PRE2K Attack
    Continue reading on Medium »
    ESC1 — Weaponizing Misconfigured Certificate Templates for Domain Admin Privilege Escalation
    The dangerous ADCS vulnerability that turns misconfigured certificate templates into a Domain Admin takeover path. Continue reading on Medium »
    My First Steps into Hardware Security: Exploring USB HID Attacks with an ATtiny85 Digispark
    As someone interested in cybersecurity and red teaming, I’ve spent a lot of time learning about web security, networking, and CTF… Continue reading on Medium »
    Malleable Redirector — A Malleable C2 Companion
    Malleable Redirector is a project created to aid Red Teams in generating meaningful redirectors configs from a Malleable C2 profile. Continue reading on InTheCyber Posts »
    ChameleonCert: Tool for Spoofed Code Signing Certificates in Red Team Operations
    Code signing is one of the most trusted trust signals in the Windows ecosystem. When a binary carries a valid Authenticode signature… Continue reading on Medium »
  • Open

    From Medium to Critical: Chaining CVE-2025–20362 & CVE-2025–20333 in Cisco ASA/FTD Across 20+…
    السلام عليكم ورحمة الله وبركاته Continue reading on Medium »
    How to Mirror Your iPhone Screen on Kali Linux Using UxPlay
    Introduction Continue reading on Medium »
    Fish | Proving Grounds | OSCP Preparation
    Box: Fish Community Rating: Hard Continue reading on Medium »
    Top Platforms to Earn Money From Bug Hunting in 2026
    💻 Wait… You Can Actually Get Paid to Hack? Continue reading on Medium »
    How Parameter Tampering Exposed 4,700 Agents and Tons of PII
    While testing on a target platform, I hit that classic wall where everything seems like i’ve tested before. Continue reading on Medium »
    Improper Access Control Leading to Account Deletion Without 2FA
    So I’d already found a hidden vulnerability on [REDACTED] before (wrote about it here: Hidden Page with 2FA Bypass). After that success,I… Continue reading on Medium »
    READING The JavaScript Gave Me a Cross-Tenant Write + SSRF
    Hey Guys and Welcome Back! 👋 Continue reading on Medium »
    How to Find More Subdomains Than Everyone Else
    Recon is a numbers game. Here is how to win it. Continue reading on Medium »
    5 Beginner Mistakes That Get Your Bug Bounty Reports Rejected
    Avoid these and your acceptance rate jumps overnight. Continue reading on Medium »
  • Open

    WordPress PBN Plugin Drops Dual Webshells via Database Injection
    During a recent incident response engagement, our team uncovered a multi-stage WordPress infection that goes beyond the usual file-based malware. The attacker combined a fake plugin, a remote command-and-control server, and two PHP web shells stored directly inside the WordPress database. The campaign is operated by a Turkish-speaking threat actor and is built around a classic SEO monetization scheme: hidden backlink injection for a Private Blog Network (PBN), most likely tied to the gambling and adult affiliate niche. Continue reading WordPress PBN Plugin Drops Dual Webshells via Database Injection at Sucuri Blog.  ( 7 min )
  • Open

    Vulnerability Report: Buffer Overflow in Path Sanitization
    curl disclosed a bug submitted by newstuff321: https://hackerone.com/reports/3804525
    Unauthenticated file deletion via deleteFileMessage DDP method allows permanent destruction of any uploaded file
    Rocket.Chat disclosed a bug submitted by eldudareeno: https://hackerone.com/reports/3611837
    Malicious Conflux Endpoint Can Leave Stale Global OOO Queue Accounting After Teardown
    Tor disclosed a bug submitted by aptupdate: https://hackerone.com/reports/3701692 - Bounty: $100
  • Open

    Vulnerability Report: Buffer Overflow in Path Sanitization
    curl disclosed a bug submitted by newstuff321: https://hackerone.com/reports/3804525
    Unauthenticated file deletion via deleteFileMessage DDP method allows permanent destruction of any uploaded file
    Rocket.Chat disclosed a bug submitted by eldudareeno: https://hackerone.com/reports/3611837
    Malicious Conflux Endpoint Can Leave Stale Global OOO Queue Accounting After Teardown
    Tor disclosed a bug submitted by aptupdate: https://hackerone.com/reports/3701692 - Bounty: $100
  • Open

    Host & Network Penetration Testing: Network-Based Attacks CTF 1 — eJPT (INE)
    No content preview
    The Intelligent Shield. OpenCTI
    No content preview
    The Art of Taking Notes
    No content preview
    Mastery Hunt: Hidden API Endpoints — A Deep Dive into API Bug Bounty Recon & Exploitation
    No content preview
    CAT Reloaded CTF — CATF 2025  — DFIR Challenges
    No content preview
    IEEE Victoris 4.0 — CTF 2025 — Finals DFIR Challenges
    No content preview
    The Crime Blue Team Lab (CyberDefenders)
    No content preview
    Bug Bounty Bootcamp #45: Token?
    You found a password reset that leaks the magic token in the API response. Or worse — the devs left an endpoint that just gives you… Continue reading on InfoSec Write-ups »
    TryHackMe — Checkmate | Full Walkthrough
    No content preview
    TryHackMe — Break Out The Cage | Full Write-Up
    No content preview
  • Open

    Host & Network Penetration Testing: Network-Based Attacks CTF 1 — eJPT (INE)
    No content preview
    The Intelligent Shield. OpenCTI
    No content preview
    The Art of Taking Notes
    No content preview
    Mastery Hunt: Hidden API Endpoints — A Deep Dive into API Bug Bounty Recon & Exploitation
    No content preview
    CAT Reloaded CTF — CATF 2025  — DFIR Challenges
    No content preview
    IEEE Victoris 4.0 — CTF 2025 — Finals DFIR Challenges
    No content preview
    The Crime Blue Team Lab (CyberDefenders)
    No content preview
    Bug Bounty Bootcamp #45: Token?
    You found a password reset that leaks the magic token in the API response. Or worse — the devs left an endpoint that just gives you… Continue reading on InfoSec Write-ups »
    TryHackMe — Checkmate | Full Walkthrough
    No content preview
    TryHackMe — Break Out The Cage | Full Write-Up
    No content preview

  • Open

    Unauthenticated reading of every file via livechat auth and predicting MongoDB ObjectId()
    Rocket.Chat disclosed a bug submitted by aikido_security: https://hackerone.com/reports/3687142
    Reflected Cross-Site Scripting (XSS) found on IBM.com domain
    IBM disclosed a bug submitted by entrovyx: https://hackerone.com/reports/3664261
    Incomplete Suppression of Transfer-Encoding: chunked Header in HTTP/2 After Redirect From HTTP/1.1
    curl disclosed a bug submitted by unknowperson0212: https://hackerone.com/reports/3793495
    Secure cookies leaked to HTTP origins through HTTPS forwarding proxy
    curl disclosed a bug submitted by daviey: https://hackerone.com/reports/3803415
    UI Consent Bypass via Comma Injection in `addAutoApproveTarget` User-Approval Dialog and Persistence Layer Disagree on Target Scope, Yielding Authen
    PortSwigger Web Security disclosed a bug submitted by hacker-kartel: https://hackerone.com/reports/3717354
  • Open

    Unauthenticated reading of every file via livechat auth and predicting MongoDB ObjectId()
    Rocket.Chat disclosed a bug submitted by aikido_security: https://hackerone.com/reports/3687142
    Reflected Cross-Site Scripting (XSS) found on IBM.com domain
    IBM disclosed a bug submitted by entrovyx: https://hackerone.com/reports/3664261
    Incomplete Suppression of Transfer-Encoding: chunked Header in HTTP/2 After Redirect From HTTP/1.1
    curl disclosed a bug submitted by unknowperson0212: https://hackerone.com/reports/3793495
    Secure cookies leaked to HTTP origins through HTTPS forwarding proxy
    curl disclosed a bug submitted by daviey: https://hackerone.com/reports/3803415
    UI Consent Bypass via Comma Injection in `addAutoApproveTarget` User-Approval Dialog and Persistence Layer Disagree on Target Scope, Yielding Authen
    PortSwigger Web Security disclosed a bug submitted by hacker-kartel: https://hackerone.com/reports/3717354
  • Open

    Footprinting Lab — Easy (Walkthrough)
    We were commissioned by the company Inlanefreight Ltd to test three different servers in their internal network. The company uses many… Continue reading on Medium »
    Active Directory Attacks — Silver Ticket Attack
    A Silver Ticket attack is a Kerberos abuse technique where an attacker forges a valid service ticket (TGS) to authenticate directly to a… Continue reading on Medium »
    Active Directory Attacks — Golden Ticket Attack
    An attacker performing a Golden Ticket attack takes advantage of the trust model of Kerberos authentication, where the Key Distribution… Continue reading on Medium »
    Windows PrivEsc: Hunting Stored Credentials
    Part 3 of the Windows Privilege Escalation series Continue reading on Medium »
    The Cybersecurity Skill That Helped Me Progress Faster Than Certifications, Courses, or Tools
    One of the Most Valuable Things I’ve Built in Cybersecurity Isn’t a Tool, It’s a Routine. Continue reading on Medium »
    Red Team Testing: Simulating Real Attacks to Strengthen Cybersecurity
    Cyber threats are becoming more sophisticated, making it increasingly difficult for organisations to identify and address security… Continue reading on Medium »
    From Login Attempts to Attack Detection: Understanding Authentication Events
    ⛔Before you begin reading: This article is available as a Medium Member story. Continue reading on Medium »
    From Domain Hashes to Remote Access: Demonstrating Pass-the-Hash in an Active Directory Lab
    Introduction Continue reading on Medium »
  • Open

    Building a Hackbot for Bug Bounties — Auth Testing Subagent Setup
    If you have been keeping up with the current state of Bug Bounties on X, you probably heard that some hunters are making small fortunes… Continue reading on InfoSec Write-ups »
    When the warehouse has no lock
    Crypto exchanges spend fortunes on the front door. Hardware security modules for the hot wallets. Bug bounties on the matching engine… Continue reading on Medium »
    GEMINI + KALI MCP SERVER FOR BUG BOUNTY
    परिवर्तनमेव विकासस्य मार्गः। Continue reading on Medium »
    How I Bought a Product for ₹1 and Forged an Admin Session: A Live Web CTF Walkthrough
    In application security, a common phrase is: “Security tools look for known patterns, but human testers look for logical mistakes.” Continue reading on Medium »
    OTP Bypass: How Hackers Break “Unbreakable” OTP Verification
    And How You Can Find These Bugs Too Continue reading on OSINT Team »
    Nikto Tool: A Beginner’s Guide to Web Server Scanning
    If you’re starting your journey in penetration testing or bug bounty hunting, you’ve probably heard about a tool called Nikto. It’s one of… Continue reading on OSINT Team »
    “Bug Bounty Bootcamp #46: Not Allowed From Your IP?”
    — How to Spoof, Brute-Force, and Mass-Assign Your Way Past Authentication Walls” Continue reading on InfoSec Write-ups »
    How I “Hacked” a Company’s Instagram Account (And Got Rewarded for an Out-of-Scope Bug!)
    LinkedIn:- https://www.linkedin.com/in/vansh-rathore-cybersecurity?utm_source=share_via&utm_content=profile&utm_medium=member_android Continue reading on Medium »
    How I Escalated Privileges in a File-Sharing Platform by Changing One Word
    Written by Kerolos Iskander | Jr. Penetration Tester Continue reading on Medium »
  • Open

    My Instructor Said “You Can’t Get a Shell.” I Got Root. — Full Web Pentest Exam Write-Up
    No content preview
    Connectors CTF 2025 — DFIR Challenges
    No content preview
    Silent Breach Lab Writeup (CyberDefenders)
    No content preview
  • Open

    My Instructor Said “You Can’t Get a Shell.” I Got Root. — Full Web Pentest Exam Write-Up
    No content preview
    Connectors CTF 2025 — DFIR Challenges
    No content preview
    Silent Breach Lab Writeup (CyberDefenders)
    No content preview
  • Open

    《动手学深度学习》第一章笔记:机器学习问题到底怎么分类 - 养肥胖虎
    动手学深度学习第一章笔记:机器学习问题到底怎么分类 好家伙, 这次我把《动手学深度学习》第一章读完了。 先说一下,我为什么会突然翻这本书。 更直接的原因是,前面我在做一个从零学习 AI 的小项目。 写一个小项目去用,去学习ai(所谓干中学) 写到训练循环时,代码里开始出现这些东西: _, loss  ( 3 min )
    【Azure AI Search】Index的字段使用默认Analyzer(standard.lucene) 和 en.microsoft 有什么不同? - 编码者卢布
    问题描述 在 Azure AI Search 里,英文检索有时会卡在一个很小的词形差异上:文档里是 brief,搜索 briefs 却搜不到。 搜索 briefs,无法命中只包含 brief 的文档。类似地,audit 和 auditing 也可能因为一个复数形式导致结果不同。 文档明明在,关键词也  ( 2 min )
    【Agentic RL / 强化学习框架】Miles 项目技术分析---(1)--- 总体 - 罗西的思考
    【Agentic RL / 强化学习框架】Miles 项目技术分析 (1) 总体 目录【Agentic RL / 强化学习框架】Miles 项目技术分析 (1) 总体0x00 概要0x01 基础1.1 Agentic RL 的需求与难点1.1.1 传统 RLHF vs Agentic RL 范式对比  ( 12 min )
    SKILL 第一定律:说点 AI 不知道的 - cup11
    大模型每次会话都是一次重启。AI 在几轮资料查阅与修复过后,就忘记了自己“原本的认知”。它无法分辨哪些是自己信手拈来的常识,哪些是经过百般尝试才纠正过来的偏见,因而失去了对“认知死角”的强调,进而产出一份又一份低质量的“手册”式的 SKILL。我们亟需先“拓印”AI 流动的认知,提供对比参照的依据,...  ( 1 min )
    Flask框架01初始了解 - 乔的港口
    Flask概念 flask是一个轻量级别的python web程序,主要的特点就是简单灵活,能够快速的上手构建一个web应用,同时具备扩展到复杂项目的能力 Flask 是基于 Python 开发的微型 Web 框架,被称作微框架,核心原因是本身不绑定任何专用工具与依赖库,原生不内置数据库抽象层、表单  ( 2 min )
    Codex 实践系列 Vol.02:让 Codex 读懂开源项目 Typer - 小七-七牛开发者
    这次用 Codex 读 Typer,最重要的一点是:面对一个新项目,第一步先别急着让它写代码。比较稳妥的做法,是先让 Codex 读目录、找入口、解释核心文件,再沿着一个具体功能追下去,最后通过测试理解项目如何验证行为。  ( 3 min )
    面试官问:你用 AI 编程半年了,那怎么保证 Claude Code 写出来的代码是对的? - 狂师
    前两天刷知乎的时候,看到一个很有意思的帖子,标题叫:《面试官问:你用 AI 编程半年了,那怎么保证 Claude Code 写出来的代码是对的?》 翻完评论区,我没有找到特别满意的答案,反而越发有感触。 借着这个话题,今天来简单聊聊。 不知道正在看文章的你,还记不记得自己第一次体验 Vibe Cod  ( 2 min )
    Agent Workflow Runtime 架构拆解:把 Agent Loop 从提示词搬进代码,长任务才真正稳了 - AI小老六
    拆解 Workflow Runtime 如何用代码接管 Agent Loop,让长任务更稳定、可复盘、可复用。 原文链接:AI 小老六 导语 过去一两年,很多人都在想同一个问题:Agent 为什么一到长任务就开始飘? 单轮问答里,模型很聪明。给它一个目标、几条约束、几个工具,它经常能给出不错的结果。  ( 3 min )
    魔珐星云SDK实战测评:重构数字人交互的底层逻辑 - smileNicky
    魔珐星云SDK实战测评:重构数字人交互的底层逻辑 2025年,AI Agent 彻底爆发。从 Cursor 重新定义编程,到通义灵码、Copilot 占据开发者心智,再到大模型(Qwen、DeepSeek、豆包)不断刷新能力边界——整个行业都在追逐一个目标:让机器响应像人一样自然、即时。 在数字人赛  ( 6 min )
    嵌入式应用开发笔记之web端设备控制台 - pie_thn
    目前正在学习嵌入式应用开发,非科班,非系统学习,半路出家型选手,但是有一定Linux基础,手头正好有一个嵌入式开发设备,硬件资源如下: 主要参数 配置 处理器 单核792MHz Cortex® A7处理器 内存 DDR3 512MB 存储 eMMC 4GB/8GB RS232 1路调试串口 RS48  ( 2 min )
    终于!我造了个「AI代驾」级别的AutoBuilder,一键干掉80%的重复CRUD工作 - weipt
    一个能让后端开发效率翻倍的“自动构建工坊”长什么样? 先说点掏心窝子的话做后端开发的朋友都知道,每个项目启动时,我们都在重复做同一件事: 建表 → 写实体类 → 写Repository → 写Service → 写Controller → 写Swagger注解 → 写数据库文档 → ... 一套流程  ( 1 min )
    从零实现富文本编辑器#14-编辑器历史变更管理与状态回溯 - WindRunnerMax
    先前我们基于React实现了视图层的适配,以此实现React组件生态的复用,降低了开发成本。接下来我们需要讨论的是,编辑器的操作管理并且支持回溯,通常来说可以称之为Redo/Undo功能,而在协同编辑场景下,本地和远程变更同步的实现会更加复杂。 开源地址: https://github.com/Wi  ( 4 min )
    装了 30 个 Skills 之后,我才搞清楚哪些是在白浪费 context - 码哥字节
    三个月前,我把从各路博客扒来的 Skills 一股脑装进 ~/.claude/skills/,总共 34 个。 效果怎样?说实话,Claude Code 确实变聪明了一些——但也开始变慢,有时候明明只是问个代码问题,它会莫名其妙地触发一堆不相关的 Skill,token 飞速消耗。最夸张的一次,一个  ( 3 min )
    AI大模型八股面试题 ,覆盖Agent、RAG、LLM等面试题(600张图解+25万字答案解析) - 小林coding
    大家好,我是小林。 如果你在准备 2026 的 AI 岗面试,先把这个网站收藏了: 小林面试笔记:https://xiaolinnote.com 专注 AI Agent 开发方向的面试题网站,图解 Agent + RAG + LLM 面试题,让一部分人先跑赢 AI Agent 开发面试。 为什  ( 3 min )
    Java反射机制——运行时"透视"类的秘密 - MalouP
    Spring、MyBatis如何在运行时动态创建对象和调用方法?答案就是Java反射。本文从实际困惑出发,用代码实战讲解反射核心API,揭秘框架底层原理,并分析反射的性能代价与使用建议。  ( 2 min )
    PKM vs RAG vs Wiki vs Memory Systems:现代知识系统全面解析 - nextdata
    一份现代知识系统的全景地图 PKM、RAG、Wiki、AI 记忆系统,以及如今实用的 AI 辅助工作流,常常被放在一起讨论,仿佛它们解决的是同一个问题。事实并非如此。它们都与知识有关,但运作在不同层面: PKM 帮助人类思考。 Wiki 帮助团队保存共享知识。 RAG 帮助机器检索外部知识。 记忆系  ( 3 min )
    拆解AI Agent的"神经系统":规划、记忆、工具、ReAct,四个概念让你真正读懂Agent - Ruby_Lu
    很多人第一次接触AI Agent,会有一种类似"读医学教材"的困惑:单词都认识,但读完不知道在讲什么。 "规划"、"记忆"、"工具调用"、"ReAct框架"……这些词经常一起出现,但彼此的关系说不清楚,看起来像是四套不同系统硬拼在一起。 这篇文章,我想把这四个东西的逻辑关系说透。它们不是并列的,更不  ( 2 min )
    Modbus Studio (免费的Modbus主从机软件) - PlayerPencil
    Modbus Studio (免费的Modbus主从机软件) 1. 软件用途 Modbus Studio 是一个用于 Modbus RTU / Modbus TCP 调试的桌面工具,主要用于: 作为 Client 连接真实设备并读取、写入点位。 作为 Server 模拟从站,供其他主站软件或设备读取  ( 5 min )
    面试官:你们项目里的线程池是怎么用的?怎么管理的? - 程序员Seven
    线程池这个问题,平时写业务时好像没什么存在感,很多代码里随手就是一个: ExecutorService executor = Executors.newFixedThreadPool(10); 看起来也能跑,任务也能异步执行,线上一开始也不一定会出问题。 但如果面试官问一句:你们项目里的线程池是怎么  ( 4 min )
    [MAF预定义ChatClient中间件-09]MessageInjectingChatClient-赋予工具消息注入的能力 - Artech
    当工具在执行过程借助注入对话历史的消息来描述当前的情况,以辅助LLM后续能够更加精准的推理,这是非常有价值的。比如工具在执行过程中发现验证的风控风险,可以注入一条`Assistant`消息模拟LLM的回复来提示用户风险的存在。  ( 3 min )
    一个公开的存储桶,让我摸进了公司的内网:云存储安全攻防实战 - 明.Sir
    上周一个客户找到我,说他们收到云厂商的安全告警——某个存储桶在凌晨3点被大量下载,疑似数据泄露。客户很慌,问我:"我们的桶设了权限的,怎么还会被拖?" 我笑着说:"你设的权限,怕不是'公开可读'那种权限吧?" 一查,果不其然。一个存放日志归档的S3兼容桶,Bucket Policy配置成了Princ  ( 3 min )
    和AI一起搞事情#7. 给游戏NPC接入Hermes? - 风雨中的小七
    有一阵没做游戏了,咱接着回来做中医游戏,这期咱们聊聊怎么给游戏NPC装个"智能大脑",顺便看看开发过程中Hook这个老朋友的新玩法。项目代码在这里[tcm_odyssey]  ( 1 min )
    从零开始:在 Windows 服务器上部署 Node.js 项目(小白实战教程) - 码农刚子
    本文记录作者第一次将 Vibe Coding 开发的 Next.js + NestJS 全栈项目部署到生产环境的完整过程。由于 Vercel 不适合托管 Node.js 后端,作者选择阿里云 Windows Server 作为服务器,使用 PM2 守护 NestJS 进程,配合 Caddy 反向代理...  ( 4 min )
    一文读懂AI应用技术:自然语言处理、语音识别/合成、可解释AI - 岳小哥AI
    AI语音助手,目前逐渐开始成为主流手机品牌的标准功能。你有没有想过:在你对手机说"帮我定个明天早上八点的闹钟",手机是怎么听懂的?  ( 1 min )
  • Open

    RL economics, morally charged terms, and "distillation"
    After a number of Twitter discussions, and repeating myself a lot in these discussions, it is time to write a short note on the economics of advancing LLM capabilities through RL, about principles of propaganda and coining new words, and about my stubborn refusal to use the term "distillation" except in a specific narrow sense. How do models advance when human-curated data has run out? It's been a while since we ran out of human data to train LLMs on. We are training on copies of the internet, large piles of (originally pirated, then purchased-and-scanned-and-wholesale ingested) books, and whatever other data sources we can obtain. This leads to a certain performance plateau, as we haven't quite figured out how to make the models more data-efficient in training. The advancements we have …  ( 13 min )
  • Open

    From GNSS Fundamentals to Static GPS Spoofing
    Introduction to GPS GNSS stands for Global Navigation Satellite System, which is a satellite-based navigation system that provides geolocation for users in real-world situations. Different groups/countries invented different navigation systems such as GPS, GLONASS, etc. GNSS generally cannot fail, but if one fails, GNSS receivers can pick up signals from other systems. We are going […] The post From GNSS Fundamentals to Static GPS Spoofing appeared first on Payatu.  ( 65 min )

  • Open

    Hermes官方桌面版发布了 - 程序员晓凡
    大家对 Hermes Agent 应该不陌生了。 Nous Research 开源的 AI 智能体项目,MIT 协议,GitHub 上 star 已经突破了 18 万。OpenRouter 应用排行榜上也是断层第一。 GitHub:https://github.com/NousResearch/he  ( 2 min )
    拆解Harness Engineering和Loop Engineering - boydfd
    从 Prompt 到 Loop:理清 AI Agent 工程的概念演进 前言 如果你最近关注 AI Agent 领域,一定被各种新术语轰炸过:Prompt Engineering、Context Engineering、Harness Engineering、Loop Engineering……这些  ( 12 min )
    15天学会AI应用开发(六)使用离线大模型对文本生成摘要 - aqi00
    ​上一篇文章末尾使用了在线大模型压缩文本生成摘要,但该方式会消耗Token,不便初学者长期使用,更好的办法是采用离线大模型来生成文本摘要。 一、离线大模型的种类 国内常用的离线大模型有阿里Qwen、智谱GLM、深度求索DeepSeek、百度文心等等,以千问的文本大模型为例,又有Qwen1.5-1.8  ( 1 min )
    码哥实测:写了20行SKILL.md,Claude的代码质量提升了一倍 - 码哥字节
    上周,一个朋友在群里发了一张截图:GitHub 上 anthropics/skills 仓库,24 小时涨了 900 颗星。 他的原话是:「一个文件夹项目,13.6 万星?这是认真的吗?」 我点开一看,确实——这个仓库的核心内容就是一堆 SKILL.md 文件,外加几个 Python 脚本。没有炫酷  ( 6 min )
    AI-Coding:2026世界杯实时看板, 支持AI聊天/竞猜/预测等 - SkySeraph
    2026 FIFA 世界杯开赛,周末一边看一边VIBE搭了一个纯静态、无后端的实时数据看板,开源在 GitHub。 在线体验: https://skyseraph.github.io/world-cup-2026/ 先看一下效果 它能做什么 3D 地球仪:48 支参赛队标注在地球上,按小组着色,点击  ( 2 min )
    Jasmine.Format - 一个高性能、线程安全的 .NET HTML 生成库 - 丁双磊
    引言 在 .NET 开发中,生成 HTML 内容是一个常见的需求。无论是构建邮件模板、生成报表、还是创建动态网页内容,我们都需要一种简洁、安全且高效的方式来构建 HTML。今天,我想向大家介绍 Jasmine.Format —— 一个专为高性能场景设计的 .NET HTML 生成库。 为什么需要 J  ( 3 min )
  • Open

    Sapphire Ticket: Ataque, Detección y Respuesta
    Autor: Luis Daniel Martínez Sánchez, Karina Bautista Bautista Ámbito: Active Directory · Kerberos · Purple Team · SIEM Continue reading on Medium »
    Living Off Restricted Interfaces: Data Exfiltration via Flipper Zero (and I’m not really a…
    I had a Flipper Zero on my desk and decided to see if it could play file ferry. Short version: it can. Long version is a small Windows app… Continue reading on Medium »
    Understanding Linux File Permissions: Securing Files and Directories
    In the previous article, we learned how to navigate the Linux file system. Continue reading on Medium »
  • Open

    Breaking Access Control: Logic Manipulation & Role Escalation (Part 2)
    After exploring path manipulation, we move into the core of web vulnerabilities: Business Logic Manipulation and Role Escalation. This is… Continue reading on Medium »
    Information Disclosure in Error Messages — A New Series Begins lab 1
    Lab 1. The first lab in my Information Disclosure series. And it teaches you something every hacker learns early: a website will tell you… Continue reading on Medium »
    Breaking Access Control:Path Manipulation (Part 1)
    In web development, we often see the “Security by Obscurity” fallacy — the belief that hiding a URL or keeping it off the UI is sufficient… Continue reading on Medium »
    Why Bridges Are a Hacker’s Favorite ATM
    The anatomy of billion dollar bridge hacks, the vulnerabilities attackers exploit, and what every security researcher must learn from them. Continue reading on Medium »
    I Wasted 3 Days Intercepting a Flutter App. Here’s What Actually Works.
    Three days. That’s how long it took me to get Burp Suite seeing traffic from a Flutter app during a security assessment. Continue reading on Medium »
    Part 3/3: Exploiting phpinfo() — Turning Information into Compromise
    Finding a phpinfo() file is just the beginning. The real value comes from analyzing its contents and using that data to advance your… Continue reading on Medium »
    Easy Bounties With Open Redirects
    Hi Hackers, Continue reading on Medium »
    Bug Hunter Bug Bounty Toolkit Powered by Claude and Free AI Provider
    Continue reading on Cyber Security Write-ups »
    Bug Hunter’s Freeze!!!!! Loop
    A Simple Framework to Stop Overthinking and Start Finding Vulnerabilities. Continue reading on Medium »
    BugHunter Review 2026: AI Bug Bounty Toolkit Powered by Claude (Free and Paid Options)
    For decades, bug bounty hunting has been a manual, time-intensive process: researchers manually enumerate subdomains, run vulnerability… Continue reading on Medium »
    Why Your UIKit Border Colors Break in Dark Mode (And How to Fix It)
    The hidden gotcha with CGColor that every iOS developer should know Continue reading on Medium »
    The Best Bug Is the One You Actually Report
    Stop chasing perfection. Start sending reports. Continue reading on Medium »
  • Open

    Burp Suite Professional: browser-powered crawl can write attacker-controlled files through file input handling
    PortSwigger Web Security disclosed a bug submitted by kawakatz: https://hackerone.com/reports/3712279 - Bounty: $5000
  • Open

    Burp Suite Professional: browser-powered crawl can write attacker-controlled files through file input handling
    PortSwigger Web Security disclosed a bug submitted by kawakatz: https://hackerone.com/reports/3712279 - Bounty: $5000

  • Open

    还在被框架绑架?一文看懂“六边形架构”,让你的核心业务稳如泰山! - Swizard
    引言:那些年我们写过的“面条代码” 痛点场景: 你一定经历过这样的噩梦:系统最初用 MySQL 存储数据,后来为了性能要迁移到 MongoDB。结果你发现,业务代码里密密麻麻全是对 MySQL 驱动的直接调用。或者,老板突发奇想,要求把原本的 Web 页面功能,原封不动地搬到一个新的命令行工具(CL  ( 2 min )
    独立开发者最容易低估的,不是开发成本,而是维护成本 - Solo社区
    很多独立开发者都有过类似经历:某个周末突然来了灵感,觉得一个小工具很值得做,于是连续几天高强度写代码、调页面、接接口、改交互,终于把第一个版本推上线。那一刻很有成就感,产品链接可以打开,核心功能能跑通,截图也能发出去,甚至还有朋友在评论区说“这个不错”“挺有用的”。 但一个月之后,情况往往会变得微妙  ( 1 min )
    【EF Core】继承策略——TPC - 东邪独孤
    在开始主题之前,老周分享另一个知识,碰巧这知识点也是 EF Core 的,是前些天一位新手程序猿问的,他那是一个小项目,因为小,所以采用 Code First 的方案。不过程序有两个版本,一个是用 SQLite 数据库,一个用 SQL Server。然后有些实体他设定了 CHECK 约束。众所周知,  ( 7 min )
    Unity项目适配华为鸿蒙系统的原生库加载问题排查与解决 - charlee44
    深入剖析了在 Huawei P30 (鸿蒙 4.0)上解决 Unity 原生库 RELRO 装载异常的完整排查路径,通过优化链接器标志、策略性使用静态库及规范符号导出,为鸿蒙系统下的底层兼容性问题提供了切实可行的解决方案。  ( 5 min )
    Zenith.NET 开发札记:把 .NET 图形 API 推向现代 RHI - o王先生o
    Zenith.NET 最近做了一轮比较大的 RHI 重构。它不是一次普通的 API 改名,也不是单纯整理代码,而是把整个图形抽象层从早期“更容易上手的封装”,往更现代、更贴近 DirectX 12 / Vulkan / Metal 的底层模型推进。 这轮重构的重点,是新版引入了哪些能力、为什么要做  ( 3 min )
    C# .NET 周刊 |2026 年 4 月 3 期 - InCerry
    C# .NET 周刊 |2026 年 4 月 3 期 2026-04-19 dotnet_week_26_4_3 国内文章 C# 15 类型系统改进:Union Types https://www.cnblogs.com/hez2010/p/19891530/union-types-in-cshar  ( 5 min )
    C++ ASCII 3D无尽跑酷游戏 - LiQirui
    TIP: 本文代码可以通过Dev-C++ 5.11编译,不用安装外部库 Github仓库:https://github.com/LiQirui-git-hub/ASCIIPakour 介绍: 这是一个3D游戏,仅支持Windows(暂时,未来有概率改变) 同时,画面不是用新窗口渲染,是命令行内渲染  ( 7 min )
    用SymPy自动因式分解:从面积拼图到代数恒等式 - wang_yb
    大家好,今天要解决一个痛点是关于 因式分解公式 的。 直接说问题: 用 Manim 展示 $ x2 + 5x + 6 = (x+2)(x+3) $ 的“十字相乘”面积模型,你需要先想好怎样把大矩形拆成四块$ (x^2)、(2x)、(3x)、(6) $,再手动计算每一块的边长和位置。 换成 $ x^2  ( 4 min )
    Claude 编程经验 - GoCodingInMyWay
    Claude Code,或者说 AI 智能体编程,聊一下个人的一些感受与心得。 感受 ⛵ 与 AI 智能体协作编程,可以说像是在航海,一起驶向计划的目的港。 我虽是那个舵手,但能放手让 AI 来掌舵了。不过,难免偏航,不时需要人来纠正一下航向。 我并没有甩手全权交给 AI,想来是几个方面吧: 一是  ( 2 min )
    26. Agent 记不住业务数据?用 Store 给它加个“笔记本“! - 老陈说编程
    Agent 通过 Checkpointer 记住对话上下文不是什么难事,但要是想让它存点“业务数据”——比如用户偏好、任务进度、历史操作记录这类东西——光靠 Checkpointer 就有点力不从心了。对话线程之间彼此隔离,换个线程就像失忆一样,之前积累的信息全用不上。Store 机制就是来解决这个  ( 2 min )
    [免费试用] iNeuOS_Vision视觉分析在SOP标准工作操作规程的应用 - iNeuOS工业互联网系统
    传统SOP依赖人工监督、人工核验,存在监管滞后、人工成本高、漏检误检等痛点。依托iNeuOS_Vision机器视觉技术,可实现SOP操作的自动化识别、实时监测、流程校验与异常告警,适配各类工业标准化作业场景,助力工业作业全流程可视化、智能化管控,提升生产标准化水平与作业效率。  ( 1 min )
    一文速览 HarmonyOS 6.1.1 推出的十个新特性 - aqi00
    ​HarmonyOS 6.1.1 已于 2026 年 5 月下旬正式发布,该版本在 HarmonyOS 6.1 基础上增强了若干特性,让鸿蒙系统变得更流畅更好用,下面结合《鸿蒙HarmonyOS 6应用开发:从零基础到App上线》一书对 HarmonyOS 6.1.1 新特性中的常用部分逐一讲解。  ( 2 min )
    Speed Tools:一套低侵入的 Android 插件化 + 动态换肤 + 字体切换框架 - future_li
    Speed Tools:一套低侵入的 Android 插件化 + 动态换肤 + 字体切换框架 作者:一航 GitHub:jasonliyihang/speed_tools 博客首发于 CSDN,本文基于 2026 年最新代码重构整理。 一、前言 几年前我在 CSDN 写过一篇 [《android 插  ( 3 min )
    一文读懂AI基础技术:机器学习、深度学习、计算机视觉 - 岳小哥AI
    从数据中寻找规律的机器学习,到模仿人脑思考的深度学习,再到让机器看懂世界的计算机视觉,这三项技术共同构建了人工智能最底层也最核心的逻辑框架。  ( 1 min )
  • Open

    How I Was Able to Create an Account on Any Third Party Application Using an Acronis-owned Email…
    Hello fellow bug bounty hunters, hope you guys are doing well and hunting lots of bugs. Continue reading on Medium »
    Google API Key Exposure: Risks, Discovery, and Impact
    Demonstrating Security Impact: Escalating Exposed Google API Keys to Access Sensitive Services Continue reading on Medium »
    Finding an Open Redirect Pattern on a Shopify Storefront: A Bug Bounty Recon Walkthrough
    While working on a web application security internship focused on bug bounty hunting, I picked the GRW Trading FZE program on YesWeHack… Continue reading on Medium »
    Blind Extraction of Password Hashes via an Unauthenticated GraphQL Count Oracle
    High Severity Vulnerability with $XXX Bounty Continue reading on Medium »
    Easy SSRF from Wayback Machine
    Link Continue reading on Medium »
    How Nmap Helped Me Discover a $4,000 Vulnerability on an E-Commerce Platform
    From Simple Recon to a $4,000 Bug: My Nmap Discovery Story Continue reading on Medium »
    Threat Hunting with WAF - LetsDefend
    This write-up is based on a training scenario from LetsDefend and is shared for educational purposes only. Continue reading on Medium »
    How I Found a Critical Vulnerability in a Fortune 500 Company — and Got Paid $15,000 for It
    A beginner’s deep dive into bug bounty hunting: mindset, methodology, and the moment everything clicked. Continue reading on Medium »
    How a Simple MFA Test Turned Into an Interesting Meta Authentication Research Case
    Like many security researchers, I occasionally spend time reviewing authentication flows on platforms that I personally use. Most of the… Continue reading on Medium »
    Nibbles | Proving Grounds | OSCP Preparation
    Start off with a nmap scan of the target: Continue reading on Medium »
    Exposed Sensitive Logs and SQL Queries Revealing User Data
    One evening, I was exploring the internet like a digital detective, using a Shodan to hunt for hidden clues. Suddenly, I found an open… Continue reading on Medium »
    Essential Guide to Information Disclosure
    Master information disclosure vulnerabilities: identify attack vectors, exploit leaks, and mitigate web security risks. Continue reading on Medium »
  • Open

    Scenario 1: Red Team vs Blue Team: How to Execute a Kerberoasting Attack and Detect It Using Wazuh…
    Welcome to the first practical scenario of my Active Directory Cyber Range series built entirely on VMware ESXi. Continue reading on Product Coalition »
    Threat Hunting with WAF - LetsDefend
    This write-up is based on a training scenario from LetsDefend and is shared for educational purposes only. Continue reading on Medium »
    Linux File System Navigation: A Beginner’s Guide
    In the previous article, we built our cybersecurity lab. Continue reading on Medium »
  • Open

    Duplicate chunked Transfer-Encoding lets a malicious origin smuggle a response across reused HTTP proxy connections
    curl disclosed a bug submitted by violet12331: https://hackerone.com/reports/3795615
  • Open

    Duplicate chunked Transfer-Encoding lets a malicious origin smuggle a response across reused HTTP proxy connections
    curl disclosed a bug submitted by violet12331: https://hackerone.com/reports/3795615

  • Open

    What is Brie — Bug Capture Tool?
    Brie is a browser extension that helps developers and QA teams capture and report bugs with full context , automatically. Instead of… Continue reading on Medium »
    SSTI: When the Server Runs What You Type
    You entered a name. The server treated it as code. Learning how to identify and fingerprint Server-Side Template Injection vulnerabilities. Continue reading on InfoSec Write-ups »
    How I Found My First Bug in Bug Bounty
    Yes believe me your first real bug is probably closer than you think. Continue reading on Medium »
    JWT Attacks: Algorithm Confusion Vulnerability
    Hi Hackers, Continue reading on Medium »
    7 Swift Concurrency Bugs That Shipped to Production
    And the Patterns That Prevent Them Continue reading on Medium »
  • Open

    领域驱动 vs 本体驱动:DDD 代码建模与 Ontology 语义建模的对比分析 - 刀法如飞
    DDD 与 Ontology 对比分析:代码建模与语义建模的异同 探讨领域驱动设计(DDD)与本体论建模(Ontology)之间的本质差异,搞清其背后的理论体系和运行机制。 一、双维建模:逻辑深度与语义广度 复杂业务系统的建模方法与开发方式可以分为两条路线: DDD 范式:以应用代码开发为主,利用充  ( 4 min )
    再也不用数括号了!安利一个JSON Path可视化查找神器 - taozsay
    一句话认识这个工具 在做数据采集或调用第三方API时,经常需要从嵌套复杂的JSON中提取指定数据。今天给大家安利一个免费在线工具——JSON Path Finder,地址是 https://formatlist.com/json-path-finder。 先直接看效果:把JSON数据粘贴进去,它会自  ( 2 min )
    开源免费的桌面自动化神器,AI 一句话生成工作流:AutoFlow Studio - 可可西里***
    它能做什么? 简单说:让你的电脑自动帮你干活。 批量点击、填表、截图 → 告别重复劳动 定时执行任务 → 每天自动签到、自动导出报表 图像识别 + 自动化 → 找到屏幕上的按钮自动点 AI 一句话生成 → "在坐标500,300点击,输入hello按回车" 直接变成自动化流程 为什么值得关注? 1.  ( 2 min )
    实时操作系统(RTOS)的核心认知基石 - 沐多
    实时操作系统内核的核心矛盾:它必须在"快"和"准"之间做出不可调和的抉择。快是吞吐量的追求,准是确定性的承诺。一个通用操作系统优化的是平均响应,一个RTOS优化的是最坏情况——这不仅是技术路线的分叉,更是工程哲学的根本对立。  ( 4 min )
    Claude手搓的IntelliJ Git扩展插件上线 - Cn#工作室
    背景 在日常Java开发工作中,迭代着迭代着本地就有一堆分支,批量删除的话有一行命令,如: git branch | grep 'release-' | xargs git branch -D 但我总记不住,也不想去住,每次操作都要翻小本本。 一直要开发一个插件,无奈不会啊。我之前是.NET专业户,  ( 1 min )
    AI编程系列02:合并知识功能,给 AI 问数和 RAG 场景打基础 - AlfredZhao
    按照之前文章《氛围编程实战系列:先规划清楚学习路径》这个规划路径,我们今天来开发第二个功能:合并知识功能。 总结这篇文章的初期阶段,其实让笔者非常的困惑。因为经历了太多思想碰撞。最终认为如果要持续学习一件新事物,还必须要从现实出发,不要为了学习而学习,而要为了解决一个实际问题而学习,哪怕这个问题开始  ( 2 min )
    别把 Product Hunt 当成冷启动:独立开发者真正要找的不是流量,而是对的人 - Solo社区
    很多独立开发者在产品刚做出来的时候,都会下意识想到 Product Hunt 或 Hacker News。 这并不奇怪。对一个没有用户、没有品牌、没有媒体资源的小团队来说,把产品发到一个已经聚集了大量产品爱好者和技术人群的平台上,听起来几乎是最合理的选择。那里有人愿意点开新产品,有人愿意评论,有人会  ( 2 min )
    混沌工程实战:基于 Toxiproxy 验证短信网关的超时兜底与频控链路 - 杜劲松
    ⚠️ 安全警告:混沌工程工具威力巨大。Toxiproxy 仅限于开发、测试(SIT/UAT)或特定的混沌工程演练环境中使用。严禁将其部署在生产环境的真实业务链路中,以免造成不可挽回的生产资损与事故。 在微服务架构和系统可观测性建设中,验证各链路的容错与兜底机制(Error-handling fail  ( 2 min )
    别只盯着GitBook了!这个文档神器让你的笔记秒变网站 - 一名程序媛呀
    还在为团队文档维护头疼吗?本文带你认识一个极简却强大的文档神器docsify,无需编译、一个页面搞定一切。从安装配置到避坑指南,手把手教你搭建一个高颜值、易维护的知识库网站,让你的技术文档管理体验瞬间起飞。  ( 2 min )
    成为 AI 智能体工程师的 10 个步骤 - nextdata
    AI领域的角色正在随着行业发展而转变。了解进入这个领域需要什么。 引言 在过去十年中,AI领域经历了快速发展,尤其是在ChatGPT推出之后。如今,工程师们很少再讨论经典机器学习,整个对话都被大语言模型(LLM)所主导。而现在,焦点正在从LLM转向AI Agent。 Agent正在改变软件的运作方式  ( 1 min )
    流式输出(Streaming)原理与踩坑经验 - 第十昵称
    流式输出(Streaming)原理与踩坑经验 本人在日常开发中,遇到流式输出相关的问题,一般都需要靠大模型协助定位问题,归其根本是因为我对流式输出的原理认识不足。所以本篇文章记录我学习流式输出的原理,以及在实际开发中遇到的问题。 整体流程: 大模型生成 token ↓ 打包成 chunk(一个或多个  ( 5 min )
    CSharp 后端服务器如何做到:一边发请求一边看代码覆盖率 - ahfuzhang
    作者:张富春(ahfuzhang),转载时请注明作者和引用链接,谢谢! cnblogs博客 zhihu Github 公众号:一本正经的瞎扯 背景 最近在测试一个服务器的时候,遇到一个难题:我如何才能构造出多种请求,以便尽可能的覆盖到所有分支? 写单元测试固然是个办法,但是服务器依赖 mysql 和  ( 2 min )
    凌霞 618 年中大促,全线半价,叠加满减! - kaika1
    2026 年中大促正式开启。6 月 16 日至 6 月 30 日,凌霞软件旗下 Halo 建站系统与 1Panel 服务器管理面板全线产品一律 5 折优惠,新购、订阅续期、版本升级、版本变更四类订单均享折扣,无需输码、无需凑单。同期开放两轮大额满减券,可与半价折扣叠加使用,最高额外再省 ¥800。  ( 1 min )
    SkyWalking / Jaeger / ARMS 已经有了,还需要统一可观测平台吗 - IT运维监控
    很多团队已经有 APM 了。 Java 服务接了 SkyWalking,云上应用用了 ARMS,部分云原生团队接了 Jaeger,新项目开始按 OpenTelemetry 做链路追踪。慢接口、错误调用、服务拓扑、Trace 瀑布图,也都能看。 所以当有人再提“统一可观测平台”时,团队会有一个很自然的  ( 2 min )
    42 倍小文件性能提升、85% 吞吐增长:多云 AI 场景下的 JuiceFS 存储实践 - JuiceFS
    星辰征途是一家聚焦 AI 搜索与电商场景多模态 AIGC 应用的初创公司,成立两年多,业务主要面向海外市场。公司目前的主要产品包括:Gensmo(gensmo.com) 聚焦时尚穿搭,提供虚拟试穿、造型推荐和商品搜索;ZooClaw(zooclaw.ai) 面向更广泛的生活与工作场景,提供 AI A  ( 3 min )
    为什么你的收藏越积越多,却越来越没用? - 张居斜
    在 AI 时代,获取知识已经不是瓶颈了。真正的瓶颈是——你积累了那么多东西,但它们从来没有被"编译"过。 信息囤积 ≠ 知识积累 看看你自己的数字生活:微信收藏了几百篇文章,Obsidian 里存了上千条笔记,浏览器书签栏挤得连图标都看不清,聊天记录里散落着各种"稍后看"的截图。 这些东西加起来可能  ( 2 min )
    数据库审计不是记流水账:先锁定高危动作与关键对象,再谈数据集与工具落地 - 刘子毅
    很多团队一谈到数据库审计,第一反应就是“把所有 SQL 都记下来”。仿佛日志越全,安全感越足。可真到出事的时候,却发现自己面对的只是海量文本,根本不知道从何查起。真正有价值的数据库审计,必须能够清晰回答四个问题: 谁做了高风险操作? 针对哪个库、哪张表、哪些数据做的? 这个动作是否越权、异常或违反流  ( 2 min )
    How LLMs Actually Work(翻译) - royalrover
    本文基于 0xkato 的英文文章 "How LLMs Actually Work" 改写整理,用于中文读者学习参考。 原文:https://www.0xkato.xyz/how-llms-actually-work 标签:Machine Learning · Transformers · LLM  ( 4 min )
    用了 lark-cli 一个月,AI 操作飞书再也没出过岔子——说说真实体验 - 码哥字节
    上周我在 Claude Code 里说了一句:「帮我把今天的站会纪要发到研发群」。 Claude 很流畅地组织好了内容,然后调用 lark-cli 发送。在它真的发出去之前,终端里打出了一段 dry-run 预览——消息体、目标群组 ID、发送时间,全部清清楚楚列在屏幕上,等我确认。 我看了一眼,发  ( 4 min )
    虚拟 DOM —— 中间层的智慧 - 微山湖上静悄悄
    操作真实 DOM 有多贵? 先看一段代码: // 把一个 <div> 的背景色改成红色 document.getElementById('box').style.backgroundColor = 'red' 你觉得这一行代码的执行成本是多少? 答案远比你想象的复杂: 1. JS 引擎找到 DOM  ( 3 min )
    Claude Code 配置第三方模型后,内置工具到底用的谁的? - C_C_菜园
    Claude Code 配置第三方模型后,内置工具到底用的谁的? 核心问题:Claude Code 配了智谱(Z.AI)的 key 后,WebSearch、图片分析、web_fetch 这些能力,是用的 Claude/Anthropic 原生的,还是被替换成了 GLM 的? 结论:全部用的是智谱的  ( 3 min )
    Spring Cloud Gateway 的 SpEL 表达式注入漏洞(CVE-2022-22947) - Sol_9
    Spring Cloud Gateway 的 SpEL 表达式注入漏洞(CVE-2022-22947) 受影响版本 以下版本的 Spring Cloud Gateway 存在此漏洞: 3.1.0 3.0.0 至 3.0.6 以及其他更早的、已停止维护的版本 环境搭建 由于我是在公司电脑(刚入职),所  ( 2 min )
    Google AX 控制面拆解:分布式 Agent 如何把断点恢复、审计策略和执行调度收进同一条链路 - AI小老六
    Google AX 的重点不在“再造一个 Agent 框架”,而在补上 控制面、状态恢复、故障隔离、权限审计 这些真正决定生产可用性的工程能力。  ( 2 min )
    [MAF预定义ChatClient中间件-08]OpenTelemetryChatClient-实现链路跟踪和性能监控 - Artech
    赋予部署的应用和服务可观测性已经是一个基本的需求,在这方面,`OpenTelemetry`无疑已经称为了事实上的标准。`OpenTelemetryChatClient`是一个预定义的`IChatClient`中间件,它利用重写的`GetResponseAsync`和`GetResponseStrea...  ( 2 min )
  • Open

    Blind vs Out-of-Band Vulnerabilities: Understanding the Difference Every Security Researcher Should…
    Hanafeyz Continue reading on Medium »
    Why Your Subdomain Takeover Reports Keep Getting Closed as N/A (And How to Fix It).
    Subdomain Takeover is often described as one of the easiest vulnerabilities to find in Bug Bounty programs. Continue reading on Medium »
    Setting Up Your Cybersecurity Lab
    So far in this series, we’ve learned how networks work. Continue reading on Medium »
  • Open

    Incomplete Fix for CVE-2026-21637: OCSPRequest and resumeSession Events Crash Node.js TLS Server via Unhandled Synchronous Exceptions
    Node.js disclosed a bug submitted by shinchan_69: https://hackerone.com/reports/3781015
  • Open

    Incomplete Fix for CVE-2026-21637: OCSPRequest and resumeSession Events Crash Node.js TLS Server via Unhandled Synchronous Exceptions
    Node.js disclosed a bug submitted by shinchan_69: https://hackerone.com/reports/3781015
  • Open

    IEEE Victoris 4.0 — CTF 2025 — Quals DFIR Challenges
    No content preview
    DVWA Cheat Sheet (Low & Medium)
    No content preview
    How I Built a SOAR Automation in Microsoft Sentinel That Responds to Attacks Without a Single Click
    No content preview
    Six levels, one lesson: LLMs cannot keep a secret
    No content preview
    Recovering a Forgotten Password in a Self-Hosted n8n Docker Deployment
    Learn how to recover complete access to a self-hosted n8n Docker deployment when password reset emails fail. Continue reading on InfoSec Write-ups »
    Header Manipulation: Bypasses, Probing, and the Security Audit Nobody Does
    No content preview
    Beyond the Patch: Understanding the SonicWall SSL-VPN MFA Bypass Exposure
    No content preview
    I Simulated an SSH Brute-Force Attack on My Ubuntu Server — Here’s How Fail2Ban Stopped It
    Building a simple attack lab to understand how Fail2Ban detects and blocks repeated SSH login attempts. Continue reading on InfoSec Write-ups »
  • Open

    IEEE Victoris 4.0 — CTF 2025 — Quals DFIR Challenges
    No content preview
    DVWA Cheat Sheet (Low & Medium)
    No content preview
    How I Built a SOAR Automation in Microsoft Sentinel That Responds to Attacks Without a Single Click
    No content preview
    Six levels, one lesson: LLMs cannot keep a secret
    No content preview
    Recovering a Forgotten Password in a Self-Hosted n8n Docker Deployment
    Learn how to recover complete access to a self-hosted n8n Docker deployment when password reset emails fail. Continue reading on InfoSec Write-ups »
    Header Manipulation: Bypasses, Probing, and the Security Audit Nobody Does
    No content preview
    Beyond the Patch: Understanding the SonicWall SSL-VPN MFA Bypass Exposure
    No content preview
    I Simulated an SSH Brute-Force Attack on My Ubuntu Server — Here’s How Fail2Ban Stopped It
    Building a simple attack lab to understand how Fail2Ban detects and blocks repeated SSH login attempts. Continue reading on InfoSec Write-ups »

  • Open

    Top Platforms to Earn from Hacking in 2026
    Hi, I’m Vipul 👋 — the human behind TheHackersLog Let me tell you something most people get wrong about bug bounty hunting. Continue reading on OSINT Team »
    CVE-2025–0282: How a One-Line Bug in Ivanti’s VPN Let Attackers Walk Through the Front Door
    A balanced breakdown of the root cause, the exploit chain, and what defenders — and pentesters — should take away. Continue reading on Medium »
    “We Encrypt With AES-256” Is Not a Security Architecture
    The questions nobody asks — and the attacks that follow when they don’t Continue reading on Medium »
    API Fuzzing for Security Testing: Part 2A: Breaking Authentication & Authorization
    JWT Attacks, IDOR/BOLA, Mass Assignment, and Authorization Abuse in Real-World APIs Continue reading on OSINT Team »
    Discovering a Security Flaw in a Government Portal Used by 3 Lakh+ Students
    A few weeks ago, while interacting with the Bihar State Education Finance Corporation (BSEFCL) portal, I came across something that seemed… Continue reading on Medium »
    Blind SSRF: You Can’t See the Response. You Don’t Need To.
    The server is still making requests. You just have to find a different way to know that. Part 3 of the SSRF series. Continue reading on System Weakness »
    Claude BugHunter AI: The Ultimate AI Tool for Bug Hunting
    How AI Is Transforming Modern Bug Bounty Hunting and Penetration Testing Continue reading on Medium »
    My First Step into Cybersecurity: Why I Chose Penetration Testing
    Hello! 👋 Continue reading on Medium »
    Guía Fundamental Sobre Information Disclosure
    Domina las vulnerabilidades de divulgación de información: descubre vectores, explota fugas y mitiga riesgos de seguridad web. Continue reading on Medium »
  • Open

    Active Directory Attacks — NTDS.dit Extraction
    NTDS.dit (New Technology Directory Services Directory Information Tree) is the core database of Active Directory Domain Services, storing… Continue reading on Medium »
    Windows PrivEsc: Unquoted Service Paths Explained
    Part 2 of the Windows Privilege Escalation series Continue reading on Medium »
    Windows PrivEsc: Unquoted Service Paths Explained
    Part 2 of the Windows Privilege Escalation series Continue reading on Medium »
    Insider-Threat-I: Simulating an Advanced Malicious Actor and Engineering the Splunk Telemetry…
    Every security analyst remembers the first time they realized that the systems we trust to protect enterprise infrastructure are, by… Continue reading on Medium »
    Blacktop — BugThrive CTF Lab Writeup
    Introduction Continue reading on Medium »
  • Open

    AI Security: explanation to Exploitation || Part 1
    No content preview
    Chaining Stored XSS and CSRF in Typemill CMS: A Deep Dive into Attribute Injection
    No content preview
    How I Built a Burp Extension Efficiently with Claude
    No content preview
    Ramnit Blue Team Lab (CyberDefenders)
    No content preview
    Making A SQLi Lab Is Not Difficult, Build One With Me.
    No content preview
    Analyzing CVE-2026-32743: PX4 MAVLink Buffer Overflow DoS
    No content preview
    Hacking a Fortune 500 Finance Company via Envoy Proxy Misconfiguration
    No content preview
    Hacking Into A Server Through FTP
    Yes, There are Still Servers That Use It Continue reading on InfoSec Write-ups »
    Android App Penetration Testing: From APK Decompilation to Runtime Exploitation [Tools and Labs]
    No content preview
    Building Another Vulnerable Lab — SSRF.
    No content preview
  • Open

    AI Security: explanation to Exploitation || Part 1
    No content preview
    Chaining Stored XSS and CSRF in Typemill CMS: A Deep Dive into Attribute Injection
    No content preview
    How I Built a Burp Extension Efficiently with Claude
    No content preview
    Ramnit Blue Team Lab (CyberDefenders)
    No content preview
    Making A SQLi Lab Is Not Difficult, Build One With Me.
    No content preview
    Analyzing CVE-2026-32743: PX4 MAVLink Buffer Overflow DoS
    No content preview
    Hacking a Fortune 500 Finance Company via Envoy Proxy Misconfiguration
    No content preview
    Hacking Into A Server Through FTP
    Yes, There are Still Servers That Use It Continue reading on InfoSec Write-ups »
    Android App Penetration Testing: From APK Decompilation to Runtime Exploitation [Tools and Labs]
    No content preview
    Building Another Vulnerable Lab — SSRF.
    No content preview
  • Open

    Command Injection via Unsanitized Bundling Options in `aws-cdk-lib/aws-lambda-nodejs`
    AWS VDP disclosed a bug submitted by inkerton: https://hackerone.com/reports/3558713
    Firecracker Out-of-bounds Read/Write Local Privilege Escalation Vulnerability
    AWS VDP disclosed a bug submitted by terrynini38514: https://hackerone.com/reports/3738654
    CRLF Injection via Custom HTTP Headers
    curl disclosed a bug submitted by bugthiru: https://hackerone.com/reports/3741744
    heap-use-after-free in state.referer when CURLOPT_REFERER replaced or cleared after perform
    curl disclosed a bug submitted by fg0x0: https://hackerone.com/reports/3774279
    RCE + PAT Exfiltration via pull_request_target in privacy-configuration/auto-respond-pr.yml Direct Supply Chain to All DDG Browsers
    DuckDuckGo disclosed a bug submitted by 6r1ff1n: https://hackerone.com/reports/3619288
    RCE + Supply Chain Attack via pull_request_target in content-scope-scripts/semver-label.yml Affects All DuckDuckGo Browsers
    DuckDuckGo disclosed a bug submitted by 6r1ff1n: https://hackerone.com/reports/3619287
    SSRF via Improper Redirect Validation in Rocket.Chat oEmbed Function
    Rocket.Chat disclosed a bug submitted by button142857: https://hackerone.com/reports/3383079
    SSRF via improper validation after DNS name resolution in the link-preview feature
    Rocket.Chat disclosed a bug submitted by button142857: https://hackerone.com/reports/3393664
  • Open

    Command Injection via Unsanitized Bundling Options in `aws-cdk-lib/aws-lambda-nodejs`
    AWS VDP disclosed a bug submitted by inkerton: https://hackerone.com/reports/3558713
    Firecracker Out-of-bounds Read/Write Local Privilege Escalation Vulnerability
    AWS VDP disclosed a bug submitted by terrynini38514: https://hackerone.com/reports/3738654
    CRLF Injection via Custom HTTP Headers
    curl disclosed a bug submitted by bugthiru: https://hackerone.com/reports/3741744
    heap-use-after-free in state.referer when CURLOPT_REFERER replaced or cleared after perform
    curl disclosed a bug submitted by fg0x0: https://hackerone.com/reports/3774279
    RCE + PAT Exfiltration via pull_request_target in privacy-configuration/auto-respond-pr.yml Direct Supply Chain to All DDG Browsers
    DuckDuckGo disclosed a bug submitted by 6r1ff1n: https://hackerone.com/reports/3619288
    RCE + Supply Chain Attack via pull_request_target in content-scope-scripts/semver-label.yml Affects All DuckDuckGo Browsers
    DuckDuckGo disclosed a bug submitted by 6r1ff1n: https://hackerone.com/reports/3619287
    SSRF via Improper Redirect Validation in Rocket.Chat oEmbed Function
    Rocket.Chat disclosed a bug submitted by button142857: https://hackerone.com/reports/3383079
    SSRF via improper validation after DNS name resolution in the link-preview feature
    Rocket.Chat disclosed a bug submitted by button142857: https://hackerone.com/reports/3393664
  • Open

    小米开源编程助手 MIMO Code 简介和简单使用测试 - 橙子家
    〇、前言 2026年6月11日,小米 MiMo 团队正式发布并开源 MiMo Code V0.1.0。 官方宣称:MIMO Code 是面向开发者的新一代 AI 编程助手,支持无限上下文,帮助你更高效地理解、构建与协作。 那么它到底是一个怎样的编程助手,本文将进行一个简单的介绍,供参考。 官网地址:  ( 3 min )
    如何更优雅地提供 MCP Resources - 编写人生
    本文使用AI辅助编写。 在设计 MCP Server 时,很多人会先想到 tools。比如连接 SQL Server 后,可以提供 list_tables、describe_table、execute_query、get_procedure_definition 等工具。这样确实可以让 AI 查询数  ( 6 min )
    软考系统规划与管理师到底是干嘛的?用“大厂物业经理”的逻辑带你了解软考系规 - 二价亚铁
    系统规划与管理师 到底是干嘛的?很多考友尤其是做运维的兄弟,经常在“高项”和“系规”之间纠结。今天咱们就用最直白的大白话,把这个系统规划与管理师拆解开看。建议点赞+收藏,下次报名时可以翻出来了解下。 一、 为什么大家都对“系规”有误解? 很多朋友一听到“系统规划与管理师”,第一反应就是:这不就是个高  ( 2 min )
    【强化学习框架】Uni-Agent 深度技术分析(2)--- 关键技术 - 罗西的思考
    【强化学习框架】Uni-Agent 深度技术分析(2) 关键技术 目录【强化学习框架】Uni-Agent 深度技术分析(2) 关键技术0x00 概要0x01 修改扩展点1.1 verl 扩展点全景1.2 关键扩展点详解E1: AgentLoopBase——继承契约E3: 注册机制——外部注入 vs  ( 11 min )
    微调LLM前你需要了解的一些概念-- 反向传播解析 - royalrover
    01 · 先建立直觉:训练 = 一个反复打分与改作业的循环 假设你在教一个学生做数学题。流程是这样的:题目给他做(前向)、老师打分(loss)、老师指出每一步错在哪、错多少(反向求梯度)、学生根据这些反馈修改自己的解题习惯(优化器更新参数)。 神经网络的训练本质上就是这个循环,只是"学生"换成了一堆  ( 4 min )
    线上 Redis 突然“爆”了,怎么办? - 佛祖让我来巡山
    凌晨三点,手机疯狂报警——Redis CPU 100%,内存快满了,业务接口一个个超时…… 别慌,这篇文章不堆术语,我会像聊天一样,带你一步步排查问题、稳住现场、顺便把优化思路讲清楚。 一、先搞清楚“爆”是什么意思? “机器爆了”通常指以下几种情况,不同情况处理方式不一样: 现象(你能看到什么) 最  ( 3 min )
    PortSwigger SQL注入LAB11 - C2H5OH
    PortSwigger SQL注入LAB12 今天我们来做PortSwigger SQL注入的LAB12,这道LAB跟之前的几道LAB不太一样,但是更加贴近现实了,那么我们现在开始吧: 【本篇目标】1. 理解并掌握基于可视化错误的SQL注入2. 通过SQL注入来获取所需administrator的密  ( 1 min )
    机房断电搞崩服务器 | 人大金仓 V8 全量备份跨实例完整恢复实录 - Theo·Chan
    把全量备份的 Kingbase v8 实例备份文件恢复到另外一个实例中 背景介绍: 公司的机房断电!!!虽然离谱但是它就是断电了。导致安装了麒麟V10操作系统的服务器宕机且尝试了几种方式后均无法恢复启动,好在磁盘是完好的。因为是测试服务器,经过评估后选择重装系统。服务器上仅有的比较重要一些的资产是部  ( 2 min )
    MagicWorld 实现长时交互视频世界建模 - vivo互联网技术
    MagicWorld 针对当前视频世界模型在长时间交互中易出现运动不合理与场景崩坏的问题,提出了一种面向长时稳定性的交互式建模框架。该方法通过引入基于光流的运动约束提升动态真实性,利用历史检索机制增强跨时间一致性,并通过多步聚合的训练策略优化整体交互序列质量,从而有效缓解误差累积问题。整体上,Mag...  ( 1 min )
    用匿名内部类优雅地计算方法执行时间 - MalouP
    写代码时总要在每个方法里复制粘贴计时逻辑?本文用匿名内部类实现了一个优雅的耗时计算模板,从"笨办法"到Lambda表达式,一步步优化代码结构,适合想提升代码设计能力的Java学习者。  ( 2 min )
    SAP中FI和MM的核心集成—物料移动自动生成凭证 - tageerxing
    一、相关表格介绍 1、 T156:移动类型主表 Movement Type (Inventory Management) 你可以把它理解为一张基础信息登记表。 主要作用:定义了所有可用的移动类型代码,并存储其最基本、最通用的描述信息。它是 T156SC 中 BWART 字段的检查表,保证了数据的合法  ( 6 min )
    Apache Doris全链路性能优化实战指南:从架构设计到生产落地 - 刘子毅
    引言 在数字化转型加速推进的今天,实时数据仓库已成为企业数据驱动决策的核心基础设施。Apache Doris作为一款高性能、易运维的开源MPP分析型数据库,凭借其极简架构、亚秒级查询响应和一站式分析能力,被广泛应用于用户行为分析、实时报表、数据集市等核心业务场景。然而,随着企业数据量的爆发式增长和查  ( 3 min )
    AI-提效模板之--SKILL.md - 威兰达
    Skills 是人与 AI 协作的桥梁(在这里等于 xx.md文件): 1. 人类把领域知识、流程经验、环境约束写进 Skill, AI 读懂后去执行。 Skills 就是这座桥——人类专注定义怎么做,AI 负责真正去做。 2. 让任务执行变得准确、稳定、可复用。 3. Skills 以 MarkD  ( 1 min )
    写爬虫时用了代理还被封?Python 代理的那些隐藏坑,我替你踩明白了 - 一名程序媛呀
    写爬虫或调用API时,IP被封是家常便饭。很多人以为加上代理就万事大吉,结果反而更慢、报错更多。这篇文章从实战出发,一次性讲透http、https、socks代理的区别和用法,教你如何高效遍历代理列表,并扒开那些官方文档不会告诉你的常见坑。读完你会知道,代理用不好,比不用还危险。  ( 2 min )
    OpenVINO™ C# API 3.3 全新发布!正式接入 OpenVINO GenAI,C# 本地大模型开发全面启航! - 椒颜皮皮虾
    OpenVINO™ C# API 3.3 全新发布!这次升级的重点,是把 OpenVINO GenAI 正式带进 C#/.NET 生态。开发者可以在 C# 项目中直接调用本地 LLM 文本生成、Whisper 语音识别、VLM 图文问答等能力,不再必须绕 Python 服务或外部脚本。 3.3 继...  ( 5 min )
    [MAF预定义ChatClient中间件-07]PerServiceCallChatHistoryPersistingChatClient——基于ReAct循环的一步一存档 - Artech
    在默认的情况下,`ChatHistoryProvider`基于**调用**对产生的请求和消息进行存档。如果一次调用涉及多轮ReAct循环,意味着每次调用可能会很多条消息,但是它们只会在ReAct循环结束之后才会被存档一次。如果最后存单失败,意味着这些消息将全部丢失,所以有时候我们ReAct循环的每次...  ( 3 min )
    AI Native 竞争力:真正稀缺的不是会用 AI,而是把事往前推的人 - AI小老六
    AI 时代,拉开差距的不是工具熟练度,而是在不确定里启动、验证和接力的能力。 原文链接:AI 小老六 我现在越来越不相信一种说法:只要把 AI 工具 用熟,人就会自然变强。 工具当然重要,但它解决的是“怎么做得更快”,没有自动解决“什么值得做”“该从哪里下手”“做砸以后怎么办”。真正拉开差距的,往往  ( 2 min )

  • Open

    做 AI 工具出海一个月,我赚到了 0 美元,但学到了这些 - 码农刚子
    本文描述了刚子使用AI开发两个出海工具、一个月后收入为零的完整经历,深刻揭示了成功故事背后的诸多陷阱:开发仅占全部工作量的10%,推广运营才是真正的关键;同质化竞争异常激烈,免费用户转付费的转化率极低;真正能赚钱的方式其实是卖课卖社群。刚子最后建议开发者务必先验证需求、主动避开红海赛道、提前想好明确...  ( 2 min )
    黄河流域pwn的wp(缺的比较多) - firefly_star
    前言 这比赛我也是被ai乱杀了,pwn第二轮上题写了会我就没打了。有点太恐怖了,后面我打星途杯去了。现在回头看其实pwn题出的很好,也是感谢ziran申出这么多题和分享自己的学习经历了,ziran是申!!!这里贴一下官方wp,我这只有部分题的题解(我太菜了),要看全解可以直接去ziran申那里看看黄  ( 6 min )
    C# .NET 周刊|2026年4月2期 - InCerry
    各位亲爱的周刊订阅者,久等了!上月因不可抗力因素,不得已停更了一个月。现已恢复更新,会抓紧赶上进度。感谢大家的守候,久等了! 国内文章 去中心化的内网聊天软件Lachat是如何实现的? https://www.cnblogs.com/zhuweisky/p/19836680 本文介绍了一款名为Lac  ( 3 min )
    [Full Clock 技术复盘] 二、SvelteKit 实战避坑指南:PWA、SSR 样式断裂、持久化防抖 - cup11
    作为第二篇技术复盘,本文通过弃用 vite-plugin-pwa,改用 $service-worker 解决了 PWA 报错 Uncaught (in promise) non-precached-url: non-precached-url :: [{"url":"index.html"}];添加...  ( 1 min )
    批量改图片DPI的Python脚本 - Laurentianelle
    批量改图片DPI的Python脚本 | 写论文终于不用一张张改了 做学术论文、投期刊的朋友应该都懂,图片DPI这玩意儿平时不起眼,投稿的时候真能急死人。大部分期刊都要求300dpi以上,有的甚至要600dpi。但平时ArcGIS出图、截图啥的,默认都是72或者96dpi,每次投稿前都得一张张改,烦得  ( 1 min )
    【Azure AI Search】 searchMode=any 和 searchMode=all 有什么区别? - 编码者卢布
    问题描述 在 Azure AI Search 中查询同一组关键词时,经常会遇到一个现象:searchMode=any 返回很多结果,改成 searchMode=all 后结果数量明显下降,甚至只剩很少几条。 例如查询下面这组关键词: audit brief report compliance 如果使  ( 2 min )
    Snowflake Summit 26 见闻实感:Goodbye Data, Hello AI - ApacheSeaTunnel
    感谢 Snowflake 这次邀请我以 Snowflake 雪花大使身份参加 Snowflake Summit。这次大会给我的触动比预想中更大,大家知道,我一直做从事数据行业,早期在 Teradata,后来在 IBM,后来在企业里 Lenovo、中金、万达等管大数据,后来成为 Apache Soft...  ( 3 min )
    微调LLM前你需要了解的一些概念-- 基于 Qwen3 配置文件的实践 - royalrover
    本文基于如下 Qwen3 配置文件进行讲解: { "architectures": [ "Qwen3ForSequenceClassification" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 151  ( 7 min )
    Raft算法处理细节 - NE_STOP
    Raft算法 Raft算法 是通过日志复制管理来达到集群节点一致性算法,这个日志复制管理发生在节点中的Leader和Followers之间。Leader节点负责管理日志复制过程,以实现各个节点间数据的一致性。 角色、人气及角色转变 Raft中,节点有三种角色: Leader:唯一复制客户端请求的节点  ( 1 min )
    洛谷P14637 [NOIP2025] 树的价值超详细题解与碎碎念 - 司马只因锥
    前言 本文去掉代码全文共 4800 字。 这篇博客并非完全意义上的题解,主要是对我学习这道题的思路回顾与总结,基于以下两篇题解融合和补充(所以记号不是我原创的,我觉得尤其是第二篇题解定义的名词非常形象易懂直接抄了),以及一些碎碎念。 写了很多细节问题,都是我在学习这个题时思考过的,所以这篇文章非常长  ( 6 min )
    当 leader 被隔离: etcd 网络分区深度分析 - 胡云Troy
    etcd-raft 节点在 follow,leader 和 candidate 状态流转。状态转移图如下: 图片摘自 https://raft.github.io/raft.pdf 正常情况下,各节点在自己的角色里好好干活。但是如果出现异常,比如网络分区后,各个节点会做什么呢? 本文主要讨论网络分区  ( 5 min )
    创建型模式:对象的诞生艺术 - 江鸟Dev
    创建型模式关心的核心问题只有一个:对象怎么来的? 把"怎么创建对象"这件事封装起来,让使用者不需要知道细节,拿来就用。  ( 9 min )
    Python PDF 解析入门:提取信息、表格与元数据 - LAYONTHEGROUND
    PDF是日常工作中最常用的文档格式之一。借助 Python,我们可以批量提取 PDF 中的文本、图片、表格等数据,从而实现自动化处理。本文将介绍如何使用 Free Spire.PDF for Python 这个免费库来完成常见的 PDF 读取任务。 1. 环境准备 1.1 安装免费 Python P  ( 2 min )
    MySQL数据库CPU与内存监控全攻略:从系统到内核的立体化观测体系 - 刘子毅
    引言:无监控,不运维 在关键信息基础设施安全保障体系中,实时、精准的资源监控是发现性能瓶颈、预防故障的第一道防线。MySQL数据库作为业务数据的承载核心,其CPU和内存的消耗直接决定了服务的吞吐量与稳定性。很多性能故障(如上一篇文章分析的CPU间歇性飙高)若能被及时监控和预警,就完全可以在恶化前被扼  ( 2 min )
    如何用 SLO 报表管理核心服务可用性 - IT运维监控
    很多团队说自己在做稳定性治理。真正落到日常工作里,往往还是事故复盘:系统出故障,开会,写原因、影响和改进项;过一阵子,又出故障,再开会。这当然比什么都不记录要好,但它还不是治理。它只是把事故写进文档。 稳定性治理要回答的是更难、也更持续的问题:核心服务这个月到底可用多久,哪些接口消耗了最多不可用时间  ( 2 min )
    数据分析基础:numpy学习 - 聪明猪^(* ̄(oo) ̄)^
    numpy学习 前言 本节内容目录清晰,可以依据目录进行查找相应的方法完成对应操作 导入: import numpy as np 1. 出现原因 为什么需要Numpy? 它是python专门设计的一个工具库,适合高性能的数据处理应用场景。 如下例子:可以清楚对比出numpy数组和list数组计算同一  ( 12 min )
    手动翻群 40 分钟变 5 分钟,我用 wx-cli + Claude Skill 搭了个本地总结器 - 码哥字节
    你有没有这种感受:早上开完站会,拿起手机一看,项目群、运营群、技术交流群——三个群加起来 400 条消息,最重要的那条结论被淹没在表情包和「收到」里面,根本找不着。 我之前每天要花将近 40 分钟翻群——不是认真看,就是怕漏掉什么重要的事。直到我把 wx-cli 和 Claude Code Skil  ( 3 min )
    LangChain 入门 Memory 会话记忆 - VipSoft
    目录短期记忆-基于内存完整代码长期记忆-基于数据库完整代码记忆管理示例代码 https://docs.langchain.com/oss/python/langchain/short-term-memory Agent 的记忆(Memory)分两类: 短期记忆(short-term memory):  ( 5 min )
    生成式AI助力小型企业:2026年完整指南 - nextdata
    探索生成式AI如何帮助小型企业提升生产力、自动化任务、改善客户体验、降低成本,并在2026年及未来驱动业务增长。 引言 生成式AI正在改变小型企业的运营方式。曾经被认为只有大型企业才能使用的技术,如今已变得更加经济实惠且易于使用,使初创公司和小型企业也能利用其能力实现增长和效率提升。 从创建营销内容  ( 2 min )
    如何集成和使用 impeccable - Newbe36524
    如何集成和使用 impeccable 其实也没啥,就是在 HagiCode 项目里集成个设计工具,勉强算是个完整方案吧——包括仓库结构、多语言工作流、内容维护这些,都是在实践里慢慢摸索出来的。 背景 说起软件开发里的 UI/UX 设计质量,其实还挺重要的,毕竟能直接决定产品的用户体验和商业成功。只是  ( 3 min )
    ABC460F 题解 - Natho_nA
    原题链接 闲话 赛时看到 F 马上就想到点分树,只剩十分多钟口胡了一下就跑了。 赛后看题解发现全是线段树分治做的,去原题 P2056 学习了一下点分树做法。发现赛时的口胡离正解还差得远。 思路 首先做一个重链剖分,进而可以以 \(O(\log n)\) 的时间求出任意两点间的距离。 把点分树建出来,  ( 5 min )
    IDEA 2026.1最新版安装教程 - 半页の时光
    废话不多说,先上 IDEA 2026.1 版本安装成功的截图,如下图,可以看到已经成功安装到 2099 年辣,舒服的很! 卸载老版本 接下来,我就将通过图文的方式, 来详细讲解如何安装 IDEA 2026.1 版本至 2099 年。首先,如果小伙伴的电脑上有安装老版本的 IDEA , 需要将其彻底卸  ( 1 min )
    前端岗位歧视:做得最多,凭什么最不被看见? - 微山湖上静悄悄
    项目上线了,群里一片欢腾。 产品经理发了朋友圈,后端同事被疯狂 @ 夸接口稳,老板周会上点名表扬了架构设计。 唯独前端,没人提。 就好像这个项目里,前端从来没存在过。 但登录页是谁写的?表单校验是谁调的?那个丝滑的加载动画是谁做的?兼容各种浏览器又是谁加班到凌晨? 都是前端。 做了 6 年前端,太熟  ( 1 min )
    SolonCode(编码智能体)支持鸿蒙 PC - 带刺的坐椅
    SolonCode,基于"Java 运行时 + Web 交互"的架构设计,天然具备跨平台能力。在鸿蒙 PC 发布之初,SolonCode 即可运行。  ( 2 min )
    C#实现控制台多区域输出 - yi念之间
    前言 近一年以来,AI Agent的发展速度非常快。 如果经常使用一些Agent CLI工具,例如 Claude Code、Gemini CLI、OpenCode 等产品,会发现它们有一个共同特点: 虽然运行在终端之中,但已经完全不是传统命令行程序的样子。 在执行任务过程中,它们通常会同时展示: A  ( 3 min )
    [MAF预定义ChatClient中间件-06]利用ImageGeneratingChatClient开发专业图片生成Agent - Artech
    我们目前已经有相当专业的图片生成的模型,它可以利用我们提供的文本提示来生成高质量的图片,但是由于我们对文字的驾驭能力不够,写不出迎合LLM的提示词。ImageGeneratingChatClient中间件结合我们注册的ImageGenerator将两者结合在一起:我们通过与Agent对话的方式说出我...  ( 6 min )
  • Open

    I Found a Critical OAuth Vulnerability in Open edX — Here’s How It Happened
    CVE-2026–53636 | CVSS 7.5 High | Affecting 45M+ users worldwide Continue reading on Medium »
    SSRF Exploitation: From Reading Internal Pages to Dumping System Files
    We already proved the server takes requests from us. Now let’s see how far that actually goes. Part 2 of the SSRF series. Continue reading on InfoSec Write-ups »
    Ghost Sender: Why Your Email Gateway Isn’t Protecting Exchange Online
    Continue reading on Medium »
    I Spent 30 Days Learning Bug Bounties. Here Is What Nobody Tells You.
    The honest diary of a beginner, week by week. Continue reading on Medium »
    Building a Real Bug Bounty Toolkit on Windows 11 in 2026
    Every command, every tool, every setting. Copy, paste, you are running. Continue reading on Medium »
    Advanced Subdomain Enumeration for Bug Bounty Hunters in 2026
    Most hunters run subfinder and call it recon. The hunters finding critical bugs run six sources, permutate every result, chase… Continue reading on Medium »
    Exposing Nearly 2,000 User Records Through a SQL Injection
    How a seemingly simple SQL Injection led to the exposure of nearly 2,000 user records and highlighted the dangers of plaintext password… Continue reading on Medium »
    Find Every Vulnerability with AI Hacklabs + WSTG
    The Future of AI-Powered Web Application Security Testing Continue reading on Medium »
    Privilege Escalation via Session Fixation Leading to Session Hijacking
    Abstract Continue reading on Medium »
  • Open

    curl-ipv4-percent-normalization-SSRF
    curl disclosed a bug submitted by monk17: https://hackerone.com/reports/3791168
    Trailing-Dot Hostname in Redirect Silently Strips Client Certificate and Auth Credentials
    curl disclosed a bug submitted by azraelxuemo: https://hackerone.com/reports/3791191
    curl/libcurl vulnerable to TLS truncation attacks
    curl disclosed a bug submitted by nyymi: https://hackerone.com/reports/1826392
  • Open

    curl-ipv4-percent-normalization-SSRF
    curl disclosed a bug submitted by monk17: https://hackerone.com/reports/3791168
    Trailing-Dot Hostname in Redirect Silently Strips Client Certificate and Auth Credentials
    curl disclosed a bug submitted by azraelxuemo: https://hackerone.com/reports/3791191
    curl/libcurl vulnerable to TLS truncation attacks
    curl disclosed a bug submitted by nyymi: https://hackerone.com/reports/1826392
  • Open

    Using a Nonlinear Crystal KTP to create a SFG/SHG solution
    I’m doing an experiment with KTP and two lasers (1052nm and 1550nm), this is the first time I try to create some more advanced optic project. Initially I was thinking it was just shining the laser light in the KTP crystal and everything was fine. Then I discovered that I need to use two mirrors … Continue reading Using a Nonlinear Crystal KTP to create a SFG/SHG solution →  ( 11 min )
  • Open

    Como o MDM da Apple Realmente Funciona – e Por Que Ele Pode Ser Contornado
    Você já se perguntou o que acontece nos bastidores quando um Mac corporativo “sabe” que pertence a uma empresa antes mesmo de você criar… Continue reading on Medium »
    Webverse-Pro GeoJearyy
    lab write-up Continue reading on Medium »
    ️‍♂️ Cache Deception: The Silent Bug That Tricks CDNs Into Leaking Private Data
    Continue reading on Medium »
    The Self-Healing Network Attack Nobody Saw Coming
    Everyone is still wardriving like it is 2017. Meanwhile, something else is already out there. And it does not sleep. It does not forget… Continue reading on Medium »
    Ghost Networks + Silent C2: The Invisible Attack Stack of 2026
    Two things are converging right now. And when they collide, the entire offensive security landscape does not just change. It becomes… Continue reading on Medium »
    The Digital Nuclear Bomb: Understanding Zip Bombs
    By Soumodeep Das Continue reading on Medium »
    From Service Account to SYSTEM: Abusing SeImpersonatePrivilege on Windows
    A practical guide to Potato attacks for penetration testers Continue reading on Medium »
    Local File Inclusion (LFI) via page Parameter Leading to Source Code Disclosure
    Vulnerability Summary Continue reading on Medium »
    Critical — Authenticated Remote Code Execution (RCE) via Unrestricted File Upload
    Executive Summary Continue reading on Medium »
  • Open

    Agentic AI Security: The Hidden Attack Surface Beyond Prompt Injection
    Artificial intelligence is not just responding to prompts and generating text anymore. The age of agentic artificial intelligence has arrived. With it comes the ability to think, make decisions, plan and act independently. What Is Agentic AI? While regular AI models just respond to inputs with outputs, an agentic AI system uses APIs, databases, cloud […] The post Agentic AI Security: The Hidden Attack Surface Beyond Prompt Injection appeared first on Payatu.  ( 69 min )
    Binwalk Path Traversal Vulnerability: Turning Firmware Analysis into Code Execution
    The Bug That Turns Security Tools Against Researchers I found a bug in binwalk that lets an attacker write files anywhere on your computer. If you are a security researcher, you have probably used binwalk to pull apart firmware files. It comes pre-installed on Kali Linux. Hundreds of thousands of people use it every day. […] The post Binwalk Path Traversal Vulnerability: Turning Firmware Analysis into Code Execution appeared first on Payatu.  ( 68 min )

  • Open

    Webverse-Pro Keyem
    lab write-up Continue reading on Medium »
    How Attackers Think: A Beginner’s Guide to the Cyber Kill Chain
    The 7 Stages of the Cyber Kill Chain Continue reading on Medium »
    The Slip Through The Cracks
    I red-teamed a Maximo deployment pipeline I built myself. The scanners found almost nothing ~ and that turned out to be the most important… Continue reading on Medium »
    Autonomous Offensive Security Platforms Signal the Future of Proactive Cyber Defense
    The cybersecurity industry is experiencing a significant transformation as Artificial Intelligence and automation become increasingly… Continue reading on Medium »
    AS-REP Roasting
    Introdução Continue reading on Medium »
    How I Discovered 23,000+ Leaked Records Through Google Dorking
    Hi, I’m mrx_w_ (Adem Ziane Berroudja), a bug bounty hunter on Bugcrowd. You can find me on Twitter and LinkedIn under mrx_w_ .In this… Continue reading on Medium »
  • Open

    2026-06-09: Atomic macOS (AMOS) Stealer infection
    No content preview
  • Open

    Build in Public, Test in Private - on the Genbounty AI Bug Bounty Platform
    AI is cumbersome to test, LLM applications can accept any input and produce any output, each response id probabilistic and the number of… Continue reading on Medium »
    PortSwigger Lab Writeup: Exploiting an API Endpoint Using Documentation
    API documentation is a goldmine for security researchers and attackers alike. While it is essential for developers, leaving interactive or… Continue reading on Medium »
    I Spent 2 Weeks Researching How Bug Bounty Hunters Are Secretly Using Claude Code -Here’s What I…
    The AI workflow that’s quietly changing who finds the bugs and who doesn’t. Continue reading on Medium »
    İlk CVE’lerim: Traccar’da iki küçük varsayımın peşinden gitmek
    Claude Code, SAST skills, biraz sabır ve Traccar’da bulunan iki güvenlik açığı Continue reading on Medium »
    Most Beginner Bug Bounty Hunters Never Earn Their First $100 The Reason Isn’t Technical Skills
    The platforms are free to join. HackerOne, Bugcrowd, Intigriti — sign up, pick a program, start testing. No degree required. No… Continue reading on Medium »
    How I Built a Burp Extension Efficiently with Claude
    The hardest part of building a Burp extension used to be the code — now it’s just coming up with the idea. Continue reading on InfoSec Write-ups »
    From Chicken McNuggets to a Bug Bounty: How a Viral Meme Started My Best Finding
    “No methodology. No fancy tools. Just a meme, a hunch, and a chatbot that really should have stayed in its lane.” Continue reading on Medium »
    How I Found 20 Security Vulnerabilities in a Major OTT Streaming Platform
    A Responsible Disclosure Case Study on DRM, CDN, API Security, and Subscription Bypass Continue reading on Medium »
    Easy 150$ Bounty: Delete all votes
    I have been hunting for a while on the Hackerone public program and what I learned is If you choose the right program and spend more time… Continue reading on Medium »
    API Hacking Sounds Scary Until You Realize It’s Just Changing Numbers
    No complex setups. No secret tools. Just you and your browser. Continue reading on Medium »
    How I Discovered 23,000+ Leaked Records Through Google Dorking
    Hi, I’m mrx_w_ (Adem Ziane Berroudja), a bug bounty hunter on Bugcrowd. You can find me on Twitter and LinkedIn under mrx_w_ .In this… Continue reading on Medium »
  • Open

    精选 5 款基于 .NET 开源免费、功能强大的 Windows 系统优化工具 - 追逐时光者
    前言 在数字化办公与娱乐需求日益增长的今天,Windows 系统的流畅度与稳定性直接影响着我们的日常体验。为了帮助大家告别系统卡顿、冗余服务以及隐私泄露的困扰,大姚给大家分享 5 款基于 .NET 开源、功能强大且免费的 Windows 系统优化工具。 Optimizer Optimizer 是一款  ( 2 min )
    补充MySQL官网知识--解锁Online VARCHAR字段扩展与Index的关系 - 东山絮柳仔
    大家可以先看本文的结论【即 4. 总结】,如有兴趣再顺读。 1. Online DDL Support for Column Operations OperationIn PlaceRebuilds TablePermits Concurrent DMLOnly Modifies Metadata  ( 3 min )
    [开源] Meta Assistant / 告别命令行,我为一堆 Python 脚本做了一个 Windows 任务栏的“家” - cup11
    Meta Assistant 是一个为 Windows 打造的任务栏托盘 Python 脚本启动器,以解决 Vibe Coding 时代下“生成脚本易,寻找脚本难”的困境。安装 Meta Assistant,快速利用本地已有环境启动对应的脚本,并原生支持 Windows 自启动,不再需要在打包、配置...  ( 1 min )
    使用 dotnet-counters 观测升讯威客服系统内存占用情况和数据吞吐性能 - 升讯威在线客服系统
    在最近给一个大客户制作技术方案时,客户提出了一个要求:对他们已经部署测试运行的环境,分析其运行性能数据,并制作分析报告。  ( 2 min )
    【Agentic RL / 强化学习框架】Uni-Agent 深度技术分析(1)--- 总体 - 罗西的思考
    【Agentic RL / 强化学习框架】Uni-Agent 深度技术分析(1) 总体 目录【Agentic RL / 强化学习框架】Uni-Agent 深度技术分析(1) 总体0x00 概要0x01 基本功能1.1 竞品对比与定位1.1.1 三者定位1.1.2 七维度对比表1.2 Uni-Agen  ( 14 min )
    用 Claude Code 写了半年大型项目,这 10 个坑踩完才知道怎么用对 - 码哥字节
    有一件事我直到用了 Claude Code 三个月才搞清楚—— 它的上下文窗口是 200K tokens,但这个数字在大型代码库里根本撑不了多久。 粗略换算:一个普通 Java 服务文件大约 200-500 行,按每行 10 个 token 计算,200K tokens 大约能容纳 400-1000  ( 4 min )
    微调LLM前你需要了解的一些概念2--多头注意力机制 - royalrover
    1. 先区分两个概念:多头和多层 Multi-Head Attention 和多层 Transformer Block 不是一回事。 一句话区分: Multi-Head Attention:同一层里,多个 attention head 并行看上下文。 多层 Transformer Block:很多层  ( 6 min )
    C# 实现优雅的热重载:基于 PicoServer 的 Live Reload 方案 - 桔子雨
    .NET 是一个开源、跨平台的开发平台,运行稳定,资源消耗低,AOT 编译进一步降低了交付体积。本文基于 .NET 10 实现一个零配置的热重载服务器,核心代码不到 50 行。 依赖安装:dotnet add package PicoServer NuGet:https://www.nuget.or  ( 2 min )
    深度学习进阶(二十九)现代 LLM 的核心架构设计其四:GQA - 哥布林学者
    上一篇我们介绍了 KV Cache:它把每一步重复的 K、V 计算存进缓存,让自回归推理的计算量骤降。 但这个加速不是没有代价的。KV Cache 的大小正比于多项参数,因此又反过来推动了注意力结构本身的改进。 这便是本篇内容:分组查询注意力(Grouped-Query Attention,GQA)  ( 3 min )
    从零开始:如何将 Reasonix CLI 集成到 HagiCode 系统中 - Newbe36524
    从零开始:如何将 Reasonix CLI 集成到 HagiCode 系统中 本文分享了将 Reasonix CLI 作为一等 Agent Provider 集成到 HagiCode 系统的完整技术实践,涵盖三层架构设计、关键技术决策和前后端实现细节。 背景 Reasonix CLI,说起来也是个挺
    MonkeyCode:长亭开源的企业级AI开发平台,GitHub 3.2k Star! - 半页の时光
    你是否曾幻想过,只用一个自然语言指令,AI就能帮你完成从需求分析到代码提交的整个开发流程?现在,这个幻想正通过MonkeyCode变为现实。 什么是MonkeyCode? MonkeyCode是由长亭科技推出的企业级在线AI开发平台,已在GitHub上获得3.2k Star。它不仅仅是一个代码  ( 2 min )
    用 39 元跑出 1.3B token 的代码知识库:OpenDeepWiki 最新 wiki 质量抽检 - 239573049
    本文基于 2026-06-09 对 OpenDeepWiki 与 AIDotNet/Means 生成页 的一次源码与页面抽检。样本里的生成配置是:目录由 gpt-5.5 生成,正文由 Mimo-2.5 生成;Mimo 侧累计消费约 1.3B token,总成本约 39 RMB,折算约 0.03 RM  ( 2 min )
    开源夜莺 v9 AI 尝鲜版:给每个 SRE 配一个 7x24 在线的资深副驾驶 - IT运维监控
    做过 on-call 的人都熟悉这几个瞬间: 半夜被一条告警吵醒,盯着手机想"这到底是真的挂了,还是又误报了",爬起来开电脑、翻指标、看邻居机器,二十分钟过去,结论是"虚惊一场"。 新接了一个业务,要给上百台机器配一套监控,PromQL、阈值、持续时间、通知规则一项项点,一两个小时就没了。 新人来值
    AI 编程 Agent 不是黑箱了:用 ccglass 看清 Claude Code 和 Codex 的真实请求 - 侯垒
    这两年,AI 编程工具的进化速度非常快。 从最早的代码补全,到现在的 Claude Code、Codex、Cursor、Cline、OpenCode,各类 Agent 已经不只是“帮你写一段函数”了。它们可以读项目、改代码、跑命令、调用工具、分析报错,甚至连续执行一整套开发任务。 这当然很爽。 但问  ( 3 min )
    【译】Visual Studio 五月更新 —— 计划、评审、优化 - MeteorSeed
    出色的开发工作有着一套固定节奏:构思、尝试、检查、调整。本月的 Visual Studio 更新便贴合了这套开发节奏。无论您是在修改任何文件前借助 Plan Agent 拟定开发思路,评审多个文件的改动内容,还是精细调校 Copilot 的工作上下文,五月版本更新在从想法落地到修改定稿的过程中增设了  ( 2 min )
    Agent 工具链工程化: Skill 负责编排判断,CLI 稳定交付的执行边界 - AI小老六
    把判断留给 Skill、把鉴权与执行收敛到 CLI,经验才会从文档沉淀成稳定交付链路。 原文链接:AI小老六 导语 团队里经常会出现一种很尴尬的能力:大家都知道这件事怎么做,也有人能把它讲清楚,可它始终停留在"得找熟人问""得翻上次会话记录"这种状态。看上去经验很多,真正能稳定复用的却不多。 在 A  ( 2 min )
    [MAF预定义ChatClient中间件-05]动态修改ChatOptions和请求消息 - Artech
    调用`IChatClient`的`GetResponseAsync`或者`GetStreamingResponseAsync`方法时,我们通常会传入一个`ChatOptions`对象来控制运行行为。`ConfigureOptionsChatClient`利用指定的委托对象来动态设置`ChatOpti...  ( 3 min )
    折腾了一下午,终于让Codex用上了DeepSeek - 程序员晓凡
    Codex简介 AI时代真的风水轮流转,前段时间最火的还是Claude Code,转眼间Codex就火得一塌糊涂。 Codex是由OpenAI 推出的AI智能体。 起初呢主要用于编程开发,后面慢慢进化成了通用智能体。界面和操作对于小白来说可谓是非常友好。 它不仅能回答问题,编写代码,还能读取电脑本地  ( 2 min )
  • Open

    Fixing the DJI RC2 login error “Check network”
    I was having a strange error with my DJI Air3s drone, it was reporting that I wasn’t connected. Then after I try to connect typing my email (and confirming the agree checkbox) it was reporting “Check network” error. And it no matter which WiFi network or hotspot I use, it always was reporting this error. … Continue reading Fixing the DJI RC2 login error “Check network” →  ( 11 min )
  • Open

    Zero-Click IP Leak in a Privacy Search Engine: Indirect Prompt Injection & Silent Patching
    No content preview
    SolarDisruption Lab Writeup (CyberDefenders)
    No content preview
    Raining Dinosaurs  —  Storm-2603 Lab Writeup [CyberDefenders]
    No content preview
    I Found the Entire Admin UI of a Live PlatformJust By Tweaking Traffic in Burp Suite
    No content preview
    Host & Network Penetration Testing: System-Host Based Attacks CTF 2 — eJPT (INE)
    No content preview
    How GraphQL Mutation Aliasing Led to a $12,500 DoS Bug in HackerOne’s Account Recovery Flow
    A small GraphQL behavior created a very real availability problem. Continue reading on InfoSec Write-ups »
    SQL Injection in Password Reset: Full Database, One Email
    No content preview
    LFI Escalation Lab Writeup [CyberDefenders]
    No content preview
    GreyCTF 2026  —  Crimewatch Forensics Challenge Writeup
    No content preview
  • Open

    Zero-Click IP Leak in a Privacy Search Engine: Indirect Prompt Injection & Silent Patching
    No content preview
    SolarDisruption Lab Writeup (CyberDefenders)
    No content preview
    Raining Dinosaurs  —  Storm-2603 Lab Writeup [CyberDefenders]
    No content preview
    I Found the Entire Admin UI of a Live PlatformJust By Tweaking Traffic in Burp Suite
    No content preview
    Host & Network Penetration Testing: System-Host Based Attacks CTF 2 — eJPT (INE)
    No content preview
    How GraphQL Mutation Aliasing Led to a $12,500 DoS Bug in HackerOne’s Account Recovery Flow
    A small GraphQL behavior created a very real availability problem. Continue reading on InfoSec Write-ups »
    SQL Injection in Password Reset: Full Database, One Email
    No content preview
    LFI Escalation Lab Writeup [CyberDefenders]
    No content preview
    GreyCTF 2026  —  Crimewatch Forensics Challenge Writeup
    No content preview
  • Open

    SSH/SFTP connection reuse can bypass SSH key identity after ssh_config_matches removal
    curl disclosed a bug submitted by byteray_ltd: https://hackerone.com/reports/3788506
    SOCKS5 no-auth accepted despite username/password-only authentication
    curl disclosed a bug submitted by kalfkinen: https://hackerone.com/reports/3786077
    Action Text ReDoS (Ruby 3.1 or lower)
    Ruby on Rails disclosed a bug submitted by ooooooo_q: https://hackerone.com/reports/2389431
  • Open

    SSH/SFTP connection reuse can bypass SSH key identity after ssh_config_matches removal
    curl disclosed a bug submitted by byteray_ltd: https://hackerone.com/reports/3788506
    SOCKS5 no-auth accepted despite username/password-only authentication
    curl disclosed a bug submitted by kalfkinen: https://hackerone.com/reports/3786077
    Action Text ReDoS (Ruby 3.1 or lower)
    Ruby on Rails disclosed a bug submitted by ooooooo_q: https://hackerone.com/reports/2389431

  • Open

    EF Core 8 + SQL Server:Contains() 突然报 "关键字 WITH 附近有语法错误"?一篇避坑指南 - 码农刚子
    升级到 EF Core 8 后,原本文能正常运行的 Contains 查询可能因生成 CTE 语法且缺少前置分号,而触发 SQL Server 错误(错误号 156)。这是 EF Core 8 有意引入的重大变更。为此,文章推荐使用参数化 Raw SQL、FindAsync 或内存过滤作为解决方案;...  ( 4 min )
    当 GIS 遇见 AI - 我才是银古
    开篇:GIS 的"数据富矿"与"效率洼地" 地理信息系统(GIS)是技术领域一个独特的存在。它的数据来源横跨卫星遥感、无人机航拍、激光雷达、物联网传感器、政府普查和众包采集,数据量以 PB 计,更新频率从年度到实时不等。它处理的不是抽象的数字,而是带有空间坐标的、与真实世界一一对应的信息——每一栋建  ( 2 min )
    第二届“星途杯”网络安全竞赛pwn全解及第一道ai的wp - firefly_star
    Pwn 时间跳跃 先用ida看看 普通的菜单题,第一个这个比较的是字符的1234,所以输字符,输入2有栈溢出,打ret2libc即可。exp如下: #!/usr/bin/env python3 from pwn import * import sys from ctypes import * #fr  ( 4 min )
    AgentForge 智能体组件:与云驿插件平台构建全生态化的微服务一体化智能开发引擎 - fanly11
    在AI与微服务深度融合的开发场景中,传统智能体组件普遍存在对接方式单一、工具扩展困难、微服务适配性差、知识体系零散等问题,难以适配企业级复杂研发流程。基于云驿插件平台构建的AI Agent组件(开发助手智能体 AgentForge),打造了一套轻量化、高兼容、可扩展的智能开发解决方案,既支持本地工具  ( 1 min )
    硬件故障后数据文件大小不对故障处理—Oracle碎片扫描恢复 - 惜分飞
    有硬件恢复圈朋友找到我,说硬件恢复之后dbv报dbv-00102错误,让我给看看是否可以处理 这个是oracle dbv中一种常见错误,一般是由于block 0 不对,或者是由于文件大小不对引起,让把恢复文件发给我,进行检查 SQL> select name,bytes/1024/1024/1024
    一文搞懂DDD 领域驱动设计思想原理 - 刀法如飞
    DDD 领域驱动设计思想原理 本文聚焦 DDD 的思想内核与分层框架,主要在于搞清楚"DDD理论原理是什么?" 、"DDD 为什么这样设计?“ 一、DDD 想解决的核心问题 领域驱动设计(Domain-Driven Design,DDD)由 Eric Evans 于 2003 年在同名著作中系统提出  ( 7 min )
    基于AScript的SQL脚本语言发布啦! - rockey627
    说到执行动态脚本,大多使用Lua、Python、JS等语言,为什么没有SQL动态脚本语言呢?现在它来了! AScript是一个开源的C#动态脚本解析执行引擎,支持扩展多种脚本语言,2026年6月8日正式发布了SQL脚本语言AScript.Lang.Sql,快来试试吧! 一、介绍 支持SqlServe  ( 5 min )
    当Agent决定“改造环境”:记一次因弱模型作弊导致的实验数据全零事件 - 凉快李某
    当Agent决定“改造环境”:记一次因弱模型作弊导致的实验数据全零事件 故事背景 最近有一篇论文在做实验,发现了一个很有趣的实验现象,那就是我的最弱的模型所在的对照组,其中最复杂的测试对象——PaddlePaddle库总是所有数据都是0。 这篇论文的一个核心任务是让大模型针对某一个被测Python库  ( 1 min )
    AI Prompt 工程化设计最佳实践(Harness Engineering) - 沉睡的木木夕
    AI Prompt 工程化设计最佳实践 一份面向软件工程师的 Prompt 设计方法论,适用于任何需要系统化、工程化提升 LLM 输出质量的场景。 涵盖从简单问答到复杂的多阶段生成流水线的通用原则。 目录 核心理念:把 Prompt 当作代码 原则一:Plan-and-Prompt 分离 原则二:多  ( 6 min )
    AI Coding 如何影响交付链路重构:写代码更快了,为什么人反而觉得更累了? - AI小老六
    AI Coding 压缩了写码环节,却把瓶颈转移到 边界澄清、质量把关、跨角色协同 和 ​上线决策​。 原文链接:AI 小老六 很多团队最近都有同一种错觉:​代码明显写得更快了​,需求却没有同样幅度地更早上线。人没有更闲,反而更容易被多个需求、多个 Agent、多个评审点位同时拉扯。 这不是工具失灵  ( 2 min )
    响应式原理 —— 数据变了,视图怎么知道? - 微山湖上静悄悄
    一个困惑 先来思考一个看起来很简单的问题。 下面这两行代码之间,发生了什么? let message = '你好' // ... 某个时刻 ... message = '再见' // ← 这一行执行之后 // ↑ // 如果页面上显示了 message, // 框架是怎么知道它变了、并且更新页面的?  ( 4 min )
    Cloud Agent 开发笔记(4):Skill 与 MCP 集成、项目后记 - 五岳
    Cloud Agent 开发笔记(4):Skill 与 MCP 集成、项目后记 上一篇讲了 Agent 事件如何推到浏览器、数据如何持久化、多会话和中断如何处理。这一篇讲能力扩展层:Skill 系统和 MCP 集成。 V1 验证的是产品形态:由管理员角色集中创建和维护 Skill、配置 MCP 连接  ( 4 min )
    开源:把自己"博客转推文"蒸馏成一个 Agent Skill - 子木聊出海
    开源:把自己"博客转推文"蒸馏成一个 Agent Skill 写完一篇博客,最累的往往不是写,是后面那一遍遍的“再分发” 同一篇文章,要拆成 Twitter/X 上一条短而锐的帖子,还要改一版更职业化的 LinkedIn。每次都得重新找那个最值得传播的观点、抠平台字数、判断要不要蹭热点、再配张图。这  ( 1 min )
    高可用之路-闲聊监控指标的局限 - 无毁的湖光-Al
    前言 在我和GPT探讨了很多天的人生之后,他终于说动了我,让我开启迟迟不想动笔的高可用系列。感谢GPT们让我从大量繁琐技术文档中解放出来,让我有时间进行真正的思考。写博客对我来说最大的收益是强制自己思考,如果连博客本身都被GPT代劳,那还不如不写。所以本文文字AI含量基本为0,纯手敲,顶多听取了一些  ( 1 min )
    上周热点回顾(6.1-6.7) - 博客园团队
    热点随笔: · 通过 CC Switch 本地路由让 Codex CLI 接入 DeepSeek 等第三方模型 (JaguarJack) · Codex新手保姆级教程:新手从安装到跑通第一个项目! (狂师) · 沪漂五周年了:我越来越迷茫了 (程序员海军) · 十年后回头看,2026 年或许是程序员  ( 1 min )
    QA Use:推荐一款AI 原生 E2E 测试平台,自然语言一键跑通用例! - 狂师
    老铁们,你还在搭建复杂的自动化测试框架? 还在手写大量定位、断言、报告? 还在让团队被重复回归测试拖垮效率? 别急,今天给大家带来 browser-use 官方出品的终极测试平台 —— QA Use,一款真正面向测试团队、开箱即用的 AI 驱动 E2E 测试系统。不用写代码、不用搭环境、不用维护脚本  ( 3 min )
  • Open

    Broken Access Control leads to delete any user’s comment
    Hello Raccoonians, Continue reading on Medium »
    When the Frontend Lies: Simple Business Logic and Access Control Bug
    Hey hey : ) Continue reading on Medium »
    How i found race conditions leading to premium subscribtion bypass via concurrent user addition
    Hello raccoonians, Continue reading on Medium »
    The $7 Billion Bug and the Pim Protocol Answer
    Continue reading on Medium »
    How I Forced A/B Experiments on Any User via an Unauthenticated Endpoint (Missing Auth)
    Hey Guys and Welcome Back 👋 Continue reading on Medium »
    Tap to Exploit: Analyzing and Manipulating Mastercard Contactless Transactions (Part 2)
    Welcome to Part 2. If you haven’t read Part 1 yet, start there : LINK Continue reading on Medium »
    Tap to Exploit: Analyzing and Manipulating Mastercard Contactless Transactions (Part 1)
    1. Introduction Continue reading on Medium »
    The Advanced URL Enumeration Guide for Recon in 2026
    The hunter who collects the most URLs doesn’t win. The hunter who collects URLs nobody else has — wins.Every tool in this article has been… Continue reading on Medium »
    Slither Static Analysis Tool
    Series: Web3 Security Zero se Advance 🛡️ | Article #12 By HackerMD | 25 min read Continue reading on Medium »
    How I Found a Flaw That Let Me Delete Anyone’s Wishlist on a Major Shopping Site
    While doing some bug bounty hunting, I found a serious IDOR (Insecure Direct Object Reference) flaw on a massive e-commerce platform. It… Continue reading on Medium »
    I Got Rejected 12 Times Before My First Bounty
    Here’s what I was doing wrong and how I fixed it. Continue reading on Medium »
  • Open

    Social Engineering Cheat Sheet: How Hackers Exploit Human Emotions to Bypass Every Security Control…
    The most advanced firewall in the world cannot stop a well-crafted phishing email that makes someone panic. Continue reading on Medium »
    CVE-2024–10914 D-Link NAS: Unauthenticated Command Injection → Root Shell
    1. Overview Continue reading on Medium »
    Stop Treating AI Security Like a Mystery: Why Agentic Bugs are Just 90s Web Vulns in Disguise
    If you spend five minutes on tech Twitter or LinkedIn right now, you’d think securing AI agents requires an entirely new branch of… Continue reading on Medium »
  • Open

    libcurl: HTTP/1.x bare LF byte in response header value enables cookie jar pollution and POST body/credential exfiltration via redirect RC=0, curl 8
    curl disclosed a bug submitted by torkd1: https://hackerone.com/reports/3785919
    DNS domain search list followed for extant domain missing A or AAAA records
    curl disclosed a bug submitted by maxhearnden: https://hackerone.com/reports/3780733
  • Open

    libcurl: HTTP/1.x bare LF byte in response header value enables cookie jar pollution and POST body/credential exfiltration via redirect RC=0, curl 8
    curl disclosed a bug submitted by torkd1: https://hackerone.com/reports/3785919
    DNS domain search list followed for extant domain missing A or AAAA records
    curl disclosed a bug submitted by maxhearnden: https://hackerone.com/reports/3780733
  • Open

    Applying Sherman Kent’s Analytic Discipline to CTI: A Practical Analyst Guide
    No content preview
    Operation Desert Hydra — AI-Assisted CTI Pipeline: MuddyWater to Kibana
    No content preview
    CTI as a Code: Complete Step-by-Step Methodology
    No content preview
    CTI as a Code in Practice: Reactive Investigation — LifeTech Pharma
    No content preview
    ThreatMapper: I Built a Self-Hosted AI Threat Intelligence Platform — Here’s How to Use It
    No content preview
    OSCP Windows Enumeration Checklist: My Complete Privilege Escalation Workflow for Every Box
    Learn the exact Windows enumeration process for OSCP, including WinPEAS analysis, credential hunting, token abuse, service… Continue reading on InfoSec Write-ups »
    JavaScript Prototype Pollution Deep Dive : — Reconnaissance, Exploitation & Bug Bounty Guideline
    No content preview
  • Open

    Applying Sherman Kent’s Analytic Discipline to CTI: A Practical Analyst Guide
    No content preview
    Operation Desert Hydra — AI-Assisted CTI Pipeline: MuddyWater to Kibana
    No content preview
    CTI as a Code: Complete Step-by-Step Methodology
    No content preview
    CTI as a Code in Practice: Reactive Investigation — LifeTech Pharma
    No content preview
    ThreatMapper: I Built a Self-Hosted AI Threat Intelligence Platform — Here’s How to Use It
    No content preview
    OSCP Windows Enumeration Checklist: My Complete Privilege Escalation Workflow for Every Box
    Learn the exact Windows enumeration process for OSCP, including WinPEAS analysis, credential hunting, token abuse, service… Continue reading on InfoSec Write-ups »
    JavaScript Prototype Pollution Deep Dive : — Reconnaissance, Exploitation & Bug Bounty Guideline
    No content preview

  • Open

    探索用 SlideML 让大模型生成 PPT 的实验方法 - lindexi
    本文介绍一种让大语言模型制作幻灯片的实验性思路。通过定义一套极简的 XML 标签语言 SlideML,让模型输出页面描述,再由确定性渲染引擎真实绘制,并利用测量到的实际数据回传给模型,形成一轮一轮的调整优化。  ( 8 min )
    [MAF预定义ChatClient中间件-04]ReducingChatClient——精减对话历史又不丢失基本语义 - Artech
    绝大部分的Agent都采用对话的方式来和用户进行交互,所以对话的内容就成了Agent决策的基础,对话历史也成为占据LLM上下文窗口的主要内容。LLM推理的质量并非与上下文的丰富程度成正向关系,有时候过多的上下文信息反而会干扰Agent的判断,导致它做出错误的决策。`ReducingChatClien...  ( 3 min )
    深度学习框架PyTorch笔记(三)数据集类(Data Set)与数据加载器(Data Loader) - GeoFXR
    在PyTorch中,数据集(Data Set)和数据加载器(Data Loader)是实现深度学习模型和测试的基本组件。下面将首先介绍数据集(Data Set)和数据加载器(Data Loader)的概念,然后介绍如何创建和使用PyTorch中的数据加载器的一些步骤和示例。  ( 3 min )
    FastAPI 身份验证总踩坑?这份 FastAPI Users “避坑指南”请收好 - 一名程序媛呀
    身份验证是每个后端项目的“第一道门”,但往往也是劝退新手的第一大坑。本文不讲虚的,以一个全栈工程师的真实踩坑视角,带你梳理 FastAPI 身份验证与用户管理的最优解—— FastAPI Users 库,从安装、实战到选型建议,一条龙拆解。如果你正被 JWT、Cookie、数据库迁移搞得头秃,这篇“...  ( 2 min )
    Postgresql TPC-H OLAP测试全流程 - aixueforever
    编译 1. 测试能否进入 PostgreSQL 先执行: sudo -u postgres psql -c "SELECT version();" 如果能输出 PostgreSQL 版本,说明数据库服务正常。 2. 创建 TPC-H 测试数据库 你当前 Linux 用户是 username,先检查  ( 8 min )
    GNU GCC 多版本函数扩展 - apocelipes
    GNU GCC 为编写跨平台代码提供了多种支持,今天要讲的就是其中一种叫“多版本函数”的技术。 什么是多版本函数 多版本函数指的是可以为同一个函数在不同的处理器平台或者指令集下编写不同的实现,程序在运行时会自动选择一个最合适的实现作为这个函数真正运行的实体。 文字解释可能比较抽象,我们拿具体的cpp  ( 2 min )
    纯浏览器PDF 编辑器:PDF.js + pdf-lib + Canvas 三件套深度拆解 - 疯狂SQL
    在线体验:geekformat.com/zh-CN/pdf/edit(不用注册、不上传文件,浏览器里就能玩) 故事开头 老板:这份 50 页的合同 PDF,你帮我在每页右下角加个页码,再把所有"乙方"的位置画个箭头。 你:好的(熟练打开 Adobe Acrobat 准备买会员)。 老板:用免费  ( 10 min )
    DeepAgents - 使用Postgres作为Checkpoint - 花酒锄作田
    前言 在用 deepagents 做 Chatbot 的时候,有个最基本的需求:Agent 得记住上一轮聊了什么。你不能每轮对话都让用户重新自我介绍一遍。 LangGraph / DeepAgents 内置了一个叫 checkpoint 的机制来处理这个事。开发阶段用 MemorySaver 跑跑  ( 3 min )
    Ubuntu 20.04 连接 HC-05 蓝牙模块失败 - zylyehuo
    博客地址:https://www.cnblogs.com/zylyehuo/ 一、问题描述 在 Ubuntu 20.04 系统中连接 HC-05 蓝牙串口模块时,Windows 系统可以正常连接,但 Ubuntu 系统中会出现蓝牙可以搜索到、PIN 码输入正确、设备可以配对成功,但是连接后立即断开的  ( 5 min )
    张高兴的 Hailo-10 开发指南:(二)使用 LangChain 搭建本地大模型 RAG 问答应用 - 张高兴
    目录环境配置安装 HailoRT安装 Hailo-OllamaPython 环境启动 Hailo-Ollama 服务实现 RAG 应用1. 引用相关包2. PDF 文档处理3. 文本切分4. 向量化和存储5. 自定义 HailoChatOllama 类6. RAG 链的组装LoRA 微调1. 微调2  ( 5 min )
    RM500U 5G模块debug及拨号上网测试 - 假-正-经
    板卡型号:rk3576 rk3588 内核版本:6.1 注意:需要将天线都接到,共4个天线及sim接好 RM500U模式说明:支持 3 种通信模式: USB 模式、基于 USB-AT 的 PCIe 模式和基于 eFuse 的 PCI e 模式, 本次测试是基于 USB-AT 的 PCIe 模式 1、  ( 1 min )
    15天学会AI应用开发(五)使用AI摘要来压缩上下文消息 - aqi00
    ​前面两篇文章分别介绍了根据消息数量截断历史对话和根据Token长度截断历史对话,可是这两种方式有两个共同的问题: 1、被删去的早期记录可能包含关键信息,直接截断会导致信息丢失。 2、原始的对话记录可能存在重复、冗余、拖沓的文字内容。 一、为什么要对原始记录做摘要 要知道,重复、冗余、拖沓的文字内容  ( 2 min )
    人机Agent团队协同:从Managed Agents原理到Multica实践 - 许雪里
    一、Managed Agents 原理 1.1、诞生背景:从 “单体” 走向 “全托管” 随着大模型工具调用与自主决策能力持续升级,AI Agent 已逐渐深入研发各落地场景。 但单体 Agent 普遍存在「架构耦合、运维成本高、无法团队协作、能力沉淀复用难」等痛点,分析原因为: 多数单体 Agen  ( 3 min )
    一、为什么要学习 USB 协议 - PlayerPencil
    这是一篇学习笔记,分享给大家,如果大家喜欢,也会开启USB的专辑 一、为什么要学习 USB 协议 USB 是 Universal Serial Bus 的缩写,中文通常叫通用串行总线。它是电脑、手机、工控机、单片机开发板和各种外设之间最常见的连接方式之一。 在嵌入式开发中,USB 常用于: 虚拟串口  ( 6 min )
    一篇标题只有5个单词的论文,改变了整个AI世界,并打开了AI大语言模型时代 - 岳小哥AI
    从GPT-1的诞生到ChatGPT的横空出世,AI从专家的实验室走进了每个人的日常,这背后是数据、算力与算法长达七十年的厚积薄发。  ( 1 min )
  • Open

    Access control vulnerabilities and privilege escalation
    Lab 1 Walkthrough: Unprotected Admin Functionality Continue reading on Medium »
    السلام عليكم ورحمة الله وبركاته
    Unauthenticated IDOR in HubSpot HubDB Legacy API — Allowed Full Data Modification & Live Publishing ($X,XXX) Continue reading on Medium »
    COMMAND INJECTION
    One of the Most Dangerous Web Vulnerabilities Continue reading on Medium »
    How I Hijacked a CTF Platform with Four Lines of Burp Suite Rules
    A few weeks ago I was poking around CTF platform. What I found was a pretty embarrassing vulnerability: any registered user could give… Continue reading on InfoSec Write-ups »
    SSRF Explained: Making the Server Your Proxy
    When the web application fetches things on your behalf — and you realize you can control what it fetches. Part 1 of the SSRF series. Continue reading on MeetCyber »
    AI-Powered Password Cracking with Hashcat MCP Explained
    Password security remains one of the biggest challenges in cybersecurity. Continue reading on Medium »
    How I Hacked Any User Account in 5Minutes: A Critical Account Takeover Story
    Introduction Continue reading on Medium »
    The $0 Purchase Bug: How Perfectly Working Code Cost a Company Thousands
    A company spent years securing its platform. Continue reading on Medium »
    I Fine-Tuned a 72B Security LLM From Scratch Then Open-Sourced Everything
    A 72B model. Honest results including what broke. Here’s the full story, and why the repo is public anyway. Continue reading on Medium »
    The Bug That Almost Made Me Quit Cybersecurity — And the Lesson That Changed My Career Forever
    For months, I hunted for vulnerabilities across countless websites. Continue reading on Medium »
    Understanding SSL Pinning: What it is and how to bypass it
    Introduction Continue reading on Medium »
  • Open

    DIY Schlieren Imaging
    Really cool instructions and 3D parts provided:  ( 11 min )
  • Open

    Bir mouse ne qeder tehlukeli ola biler ?
    1 həftə əvvəl gecə saat 2-də driver research edərkən birdən ağlıma hər gün gözümün qabağında olan, hər gün istifadə etdiyim mousemin… Continue reading on Medium »
    No Red, No Blue: Be Purple
    Having the Red Team and the Blue Team working in silos limits the effectiveness of your cybersecurity program. Collaboration between… Continue reading on Medium »
    MobileHackingLabs — Cyclic Scanner writeup
    This was my most enjoyable lab so far. The blog will contain the notes I took while mid-lab to give a picture of what was going on in my… Continue reading on Medium »
    Understanding Obfuscation: From Code Protection to Security Testing
    Why developers hide code, how attackers abuse it, and how security professionals can analyze it. Continue reading on Medium »
    รีวิว CRTA | Certified Red Team Analyst
    Hello, I’am Sho. วันนี้ผมจะมารีวิว Certificate | CRTA ที่ผมสอบพร้อมกับความรู้ AD ที่เรียนมา 1 สัปดาห์ 5555555555 Continue reading on Medium »
    Lateral Movement Vs Pivoting (Difference With Practical)
    Lateral Movement vs Pivoting — I Used to Think They Were the Same Thing Continue reading on Medium »
    NTLM Authentication Explained: How It Works, Why It’s Weak, and How Attackers Exploit It
    Blog 6 of “Mastering Active Directory — Simplified.” Continue reading on Medium »
    Windows Authentication Explained: LSASS, Access Tokens, and Why Logon Types Matter
    Blog 5 of “Mastering Active Directory — Simplified.” Module 1 covered the architecture of Active Directory. Module 2 is where we get into… Continue reading on Medium »
  • Open

    The Most Dangerous Security Bug Is the One That Feels Like a Feature
    A single click should not carry the weight of your entire developer identity. Continue reading on InfoSec Write-ups »
    The 5 Skills Every Cybersecurity Engineer Needs in 2026 (That Universities Still Aren’t Teaching)
    Are you prepping for a cybersecurity market that does not exist ? Continue reading on InfoSec Write-ups »
    Update: The Ending of My $500 Loss and Web Cache Poisoning Story.
    No content preview
  • Open

    The Most Dangerous Security Bug Is the One That Feels Like a Feature
    A single click should not carry the weight of your entire developer identity. Continue reading on InfoSec Write-ups »
    The 5 Skills Every Cybersecurity Engineer Needs in 2026 (That Universities Still Aren’t Teaching)
    Are you prepping for a cybersecurity market that does not exist ? Continue reading on InfoSec Write-ups »
    Update: The Ending of My $500 Loss and Web Cache Poisoning Story.
    No content preview
  • Open

    OpenSSL TLS 1.2 session resumption accepts expired server certificates in libcurl
    curl disclosed a bug submitted by awofjawofjfawf: https://hackerone.com/reports/3781305
    curl cross-origin HTTPS redirect reuses TLS client certificate for unintended second-origin mTLS authentication
    curl disclosed a bug submitted by fanhua: https://hackerone.com/reports/3749428
    curl External-Controlled Filename in `--url @file` Leads to Arbitrary File Overwrite
    curl disclosed a bug submitted by alphalaab: https://hackerone.com/reports/3766392
    Valid share tokens allow to access tempory upload files of share owner
    Nextcloud disclosed a bug submitted by pirikara: https://hackerone.com/reports/3483708
    Authentication Bypass in ID4me handling via Missing JWT Signature Verification in User OIDC
    Nextcloud disclosed a bug submitted by priyanka010: https://hackerone.com/reports/3489490 - Bounty: $2500
    PIN bypass in PassCodeActivity via back button
    Nextcloud disclosed a bug submitted by alper_ozturk: https://hackerone.com/reports/3625210
  • Open

    OpenSSL TLS 1.2 session resumption accepts expired server certificates in libcurl
    curl disclosed a bug submitted by awofjawofjfawf: https://hackerone.com/reports/3781305
    curl cross-origin HTTPS redirect reuses TLS client certificate for unintended second-origin mTLS authentication
    curl disclosed a bug submitted by fanhua: https://hackerone.com/reports/3749428
    curl External-Controlled Filename in `--url @file` Leads to Arbitrary File Overwrite
    curl disclosed a bug submitted by alphalaab: https://hackerone.com/reports/3766392
    Valid share tokens allow to access tempory upload files of share owner
    Nextcloud disclosed a bug submitted by pirikara: https://hackerone.com/reports/3483708
    Authentication Bypass in ID4me handling via Missing JWT Signature Verification in User OIDC
    Nextcloud disclosed a bug submitted by priyanka010: https://hackerone.com/reports/3489490 - Bounty: $2500
    PIN bypass in PassCodeActivity via back button
    Nextcloud disclosed a bug submitted by alper_ozturk: https://hackerone.com/reports/3625210

  • Open

    Mastering Clickjacking: Bypassing CSRF Protection via UI Redressing
    In web security, we often look at vulnerabilities hidden deep within backend code or API endpoints. However, some of the most clever… Continue reading on Medium »
    I Didn’t Hack the App. I Hacked the AI. Web LLM is breached !
    Author : Nilanjan Chowdhury Continue reading on Medium »
    Hunting Android Lockscreen Bypasses on Pixel: A Campaign Walkthrough — Contd. — III
    06 — L5: WindowManager Occlusion — showWhenLocked and Friends Continue reading on Medium »
    Hunting Android Lockscreen Bypasses on Pixel: A Campaign Walkthrough — Contd. — II
    05 — L4: Biometric Framework — The HAL Is the Whole Story Continue reading on Medium »
    Business Logic Flaw — Users Can Manipulate Their Own Subscription Price via Exposed A/B Testing API
    By Divakar Vasani Continue reading on Medium »
    Leaking Production Stripe Promo Codes via Unauthenticated PayloadCMS API — From Information…
    By Divakar Vasani Continue reading on Medium »
    Why Some Hunters Make $10k a Month and Others Make Nothing
    It’s not skill. It’s not luck. Here’s the actual difference. Continue reading on Medium »
    Inner-Parser SSRF
    The bug class your URL validator was never going to catch Continue reading on Medium »
    Mastering CSRF: PortSwigger Academy Walkthrough Guide Community Edition From Practitioner LAB 07
    🛑 Lab 7: SameSite Lax bypass via method override Continue reading on Medium »
  • Open

    How I Passed the CLLMSP — LLM Security From an Enterprise Practitioner’s Perspective
    I don’t write about certifications I haven’t used in the real world. Continue reading on Medium »
    Threat Hunting with Email Security - LetsDefend
    This write-up is based on a training scenario from LetsDefend and is shared for educational purposes only. Continue reading on Medium »
    Network Pusulası #9 — ARP: IP ile MAC Arasındaki Köprü ve Saldırgan İçin Açık Kapı
    Ağda iki tür adresimiz olduğunu öğrendiğimde — IP ve MAC — aklıma bir soru takılı kalmıştı: “Madem IP adresi var, MAC adresine ne gerek… Continue reading on Medium »
    4.9 Million Attacks in 30 Days: What a Distributed Honeypot Taught Me
    For 30 days I left two deliberately vulnerable servers exposed on the open internet one in London, one in Bangalore and recorded… Continue reading on Medium »
  • Open

    一条信息在 PowerMem 中的完整旅程,从写入到淘汰的工程拆解 - knqiufan
    第一篇:PowerMem 记忆系统的遗忘设计,从神经元到代码工程 上一篇 PowerMem 记忆系统的遗忘设计,从神经元到代码工程 从认知科学的角度聊了遗忘机制,包括突触可塑性、艾宾浩斯遗忘曲线、间隔重复和理想的困难。都是一些很有意思的认知学理论。 这一篇换个视角,以一条消息的流转路径为线索,跟踪它  ( 5 min )
    Winter is Coming:当AI疯王们举起屠刀,弑君者已在路上 - boonya
    铁王座从不是永恒的。当OpenAI与Anthropic疯狂封锁+86、猎杀中文开发者时,它们不知道——真正的“凛冬”,正由那些被它们亲手推开的弑君者带来。 凛冬已至,疯王犹在炉边添野火 《权力的游戏》里,史塔克家族世代传诵一句警言:“Winter is Coming.” 凛冬降临时,长城之外的白鬼会  ( 1 min )
    ECC安装与配置:把 Claude Code 装进一个能稳定发挥的 Harness - 叫授_pront
    一、从一次"差口气"说起 老陈做了五年前端,最近接了个全栈私活:Python 后端加 React 前端,登录、注册、JWT、邮箱验证,外加一个管理后台。 听起来不大。干起来才发现处处别扭。 本地跑得挺顺的代码,一推到线上就报错。pip install 装出来的版本和 lock 文件对不上,celer  ( 23 min )
    25. AI 总是 “失忆“?3 行代码让它记住你! - 老陈说编程
    你有没有遇到过这种令人抓狂的情况?和 AI 智能体聊了半天,它帮你计算了复杂数据、处理了多个文件,结果你下一句问 "刚才算的结果是多少?",它却一脸茫然地说 "抱歉,我不记得之前的对话了"。 这就是 AI 最让人头疼的 "先天性失忆症"——默认情况下,所有大模型和智能体都是无状态的。每次 API 调  ( 2 min )
    Luogu P2801 教主的魔法 题解 - EtherealYz
    前言 题目传送门 Luogu P2801 教主的魔法 。 好题,将分块的精髓——整块维护,局部暴力——体现得淋漓尽致。 题意 给定 \(n\) 个数,实现以下两种操作: M L R W ,表示将 \([L,R]\) 内每个数加上 \(W\) 。 A L R C ,表示询问 \([L,R]\) 内有多  ( 3 min )
    理解 Agent 中的 Slash Command:从概念到自定义命令实践 - SharpCJ
    一、为什么要单独理解 Slash Command 用 Agent 做事时,经常会遇到以 / 开头的输入: /help /new /model /skill markdown-to-wechat-richtext 这类输入就是 Slash Command。 它看起来像一句聊天消息,但本质上更接近“命令  ( 5 min )
    当AI成“棋王父亲”,人类驾驶会沦为“3岁棋手”的奢侈游戏吗? - patton88
    2022年,索尼AI车手GT Sophy在《GT赛车》里,用比人类冠军快2秒的圈速证明:AI已经能在公平竞争中“征服”赛道。就像围棋国手古力9岁时赢了父亲古巨山,从此父亲的角色从“对手”变成“陪练”——未来的AI,或许就是那个永远比人类强的“棋王父亲”。而这个“超越时刻”,可能比我们想象的更近。 一  ( 1 min )
    AI时代一届计算机学生本科生 四年生活 保研记录: 四非本科 -> 中流985 - xuhe2
    谨以此篇Blog记录我的大学四年生活. 我们这一届学生是刚好大一的时候ChatGPT开始出名, 然后AI快速发展, 到现在快毕业了, Claude都厉害成这个样子了 一个南方四非计算机学生的记录, 较多踩坑, 少量信息差. 即使是这样卷也没有看到所谓的光明未来, 研究生完全不一定是我需要的, 只是本  ( 2 min )
    基于 superpowers 实现复杂前端改造 - 袋鼠云数栈前端
    superpower 是开源社区非常出名的 harness 框架, 我们在本文通过数栈产品 easyIndex 的复杂需求实现来探究 superpower 的价值。  ( 2 min )
    莫比乌斯反演学习笔记 - Billlly
    莫比乌斯反演 学习笔记 因为在 NDPC 中发现自己并不会莫比乌斯反演于是来学习了 目录莫比乌斯反演 学习笔记积性函数常见的积性函数狄利克雷卷积莫比乌斯反演求解莫比乌斯函数例题互质数对个数朴素做法反演做法优化:整除分块YY的GCD推导过程NDPC_L - LCM题意做法复杂度代码实现更进一步! 积性  ( 6 min )
    数据库查询优化器查询重写 / 逻辑优化 - aixueforever
    前置知识 语法树 AST 是 Abstract Syntax Tree,中文通常叫 抽象语法树。 在数据库里,用户写的 SQL 文本会先经过词法分析和语法分析,被转换成一种树形结构,这棵树就是 AST。它描述的是 SQL 的语法结构,而不是最终怎么执行。 例如 SQL: SELECT name, a  ( 10 min )
    15天学会AI应用开发(四)根据Token长度截断历史对话 - aqi00
    ​上一篇文章说到按照消息数量来截断历史对话,这种方式有个问题,就是每次对话的内容可长可短,导致固定消息数量的对话内容忽长忽短。 历史对话内容不光要存入数据库,还要作为初始提示词发给下次新会话的大模型。太长的提示词不仅冗余,还会消耗大量Token,让用户钱包快速缩水。太短的提示词容纳的信息量不足,难以  ( 2 min )
    Luogu P11721 题解 - hnczy
    似乎没有块门。 这题分块也是可以做的,时间复杂度 \(O(n\sqrt n)\) 是大手子 @ 在场上想出来的,非常浅显易懂也好码。 首先对修改分块,每一次修改到块的最右边的时候,对块进行构造。具体的说,将块内修改的 \(l\) 和 \(r+1\) 离散化并且记录 \(1\sim n\) 每一个点的  ( 3 min )
    一款接入20+免费大模型渠道的AI编程工具 - WellingFengX
    一款接入20+免费大模型渠道的AI编程工具 我做了十几年游戏开发,从 Unity C# 做到 Unreal C++,后来又碰了不少自研引擎。去年开始深度用 Claude Code 和 Codex CLI 写代码——不是那种"帮我写个排序"的用法,是让它读整条渲染管线,跨十几个文件改逻辑,加物理调试工  ( 3 min )
    洛谷P15799 [GESP202603 五级] 找数 题解 - _wyt001
    P15799 [GESP202603 五级] 找数 题目传送门:https://www.luogu.com.cn/problem/P15799 题目背景 对应的选择、判断题:https://ti.luogu.com.cn/problemset/1209 题目描述 给定一个包含 n 个互不相同的正整数  ( 2 min )
    nvp6021调试记录 - 假-正-经
    平台:rk3576 方案:原生bt1120 tx接nvp6021芯片 转为ahd信号 接ahd显示屏幕 1、硬件确认 1.1 rk3576 bt1120硬件确认 查阅硬件设计指南可以确认 rk3576是支持 16bit bt1120模式 ,最高支持1080p60Hz。 对应引脚可以查阅引脚设计指南或  ( 6 min )
    AI沉默的二十年,是数据和算力的厚积薄发,更是从机器学习到深度学习算法的一次次突破 - 岳小哥AI
    2016年,AlphaGo的胜利让AI成为全民话题,但这并非凭空降临,而是机器学习、图像识别、自然语言处理等多项基础技术二十年来不断积累与融合的结果。  ( 1 min )

  • Open

    当 AgentGateway 遇上 OpenClaw.NET:企业级智能体基础设施的深度协同实战 - 张善友
    2026 年 6 月,Linux 基金会旗下的[智能体 AI 基金会(AAIF)正式接纳 AgentGateway 为第四大核心托管项目](https://aaif.io/blog/agentgateway-joins-aaif-as-an-open-gateway-for-agentic-ai-i  ( 2 min )
    JS事件深度解析四 事件的循环和异步 - 用户新
    四、 事件的循环和异步 半年前写的这个js的事件系列,一直没完结。中间又写了个V8引擎入门的系列,也写到了执行部分。先把这个js事件系列写完。事件本身是强依赖浏览器的,尤其是循环和异步,所以在深度上,可能会比前三部分略微深入一点。对V8感兴趣的朋友可以看我写的另一个系列 V8引擎精品漫游指南 。 这  ( 18 min )
    扩展 Lucas 定理(exLucas)学习笔记 & 详解,一文带你彻底看懂扩展 Lucas 定理。 - EtherealYz
    背景 考虑这样一个问题: 给定 \(n, r, p\) ,求 \[\binom{n}{r} \bmod m \]不保证 \(m\) 是质数。\(1\le r\le n\le 10^{18}\)。 显然此时普通 Lucas 定理就无能为力了。这时需要用到扩展 Lucas 定理(exLucas)。 前置  ( 5 min )
    2026 XCPC 杂题选解 - lingfunny
    XCPC 2026 WEEK 14 C - Quadratic Jumps tag(s): math, brute force CodeForces - 2231F 首先由 Fermat Polygonal Number Theorem,答案不会超过 4。 考虑什么时候答案为 1,直接检测 \(b  ( 6 min )
    多Agent开发笔记:为什么4个Codex加1个Claude会把cpu跑满 - 养肥胖虎
    多Agent开发笔记:为什么4个Codex加1个Claude会把9700X跑满 好家伙, vscode里开了四个codex拓展 + 一个 claude把我cpu吃满了,不是哥们,我9700X啊 按理说,8 核 16 线程的桌面 CPU,日常开发应该不算弱. 但我同时开了: 4 个 Codex 1 个  ( 4 min )
    对于Prompt的思考:从“手写”到提示词采样、A/B Test 与自动化评测 - Whitejoce
    tldr: 1、好 prompt 是激活正确分布:底层原理 2、对于强 agentic 模型,过度规则会造成模型开始“执行规则”,而不是进入状态:不同模型,不同策略 3、编写prompt的采样也是在挖掘自己的真正需求:模型可以走多远、现在这个路径是不是正确的? 4、模型采样输出prompt和对应回答  ( 3 min )
    LLM 数据可视化:从“硬编码”到“Generative UI”的五种范式 - BurningFish
    声明:本文在写作过程中使用了AI辅助工具进行资料整理、结构优化与语言润色。核心观点、技术判断与工程经验均为作者原创。 一、问题:卡在渲染层 项目里有这样一条链路:用 LLM 按指定的 schema 抽取领域数据(structured output),拿到结构化数据后,前端写代码把它渲染出来: typ  ( 3 min )
    基于大模型、AiInsight问数框架的高炉冶炼智能体研究与应用(面向高炉历史诊断、目标优化与生产决策辅助) - iNeuOS工业互联网系统
    AiInsight 问数框架将大模型、技能包、数据源、工艺知识库和工具执行统一到一个智能体流程中。用户可以用自然语言或明确命令发起任务,系统根据已选大模型、技能、数据源、文件和知识库构造上下文,分步骤完成数据查询、工艺推理、脚本分析和报告生成等过程。  ( 2 min )
    测试人必备的用例评审SKILL,采纳率提升至95%以上 - AITest研究员
    做测试的同学都知道,用例评审是个"苦力活"。 每次需求变更,都要手动对比:现有用例覆盖了哪些功能点?需求里的功能是否还有遗漏?边界条件考虑了吗?异常场景测了吗? 拿着Excel用例,对着几十页的PRD文档,一条一条核对,眼睛看花了不说,还容易漏。更要命的是——评审结果全靠人工判断,不同人评审标准还不  ( 1 min )
    用SymPy自动计算抛物线求根、判别式与顶点 - wang_yb
    做 Manim 动画时,我想让抛物线 $ y=x^2+bx+2 $随着系数 b 的变化,自动、精准地显示它与 x 轴的交点。 手写求根公式不仅繁琐,还要自己处理判别式为负的情况,稍不注意 math.sqrt 就会让整个动画崩溃。 本文我们就用 SymPy 彻底解决这个痛点。 1. 痛点场景还原 假设  ( 3 min )
    自建 Copilot Cli 代理:让 GitHub Copilot 真正"Bring Your Own Key" - 胖纸不争
    自建 Copilot Cli 代理:让 GitHub Copilot 真正"Bring Your Own Key" Github: https://github.com/wosledon/copilot-auto-byok 一个基于 .NET 10 的轻量级模型代理,解决 Copilot Cli 只  ( 2 min )
    JuiceFS 1.4|大规模元数据操作优化:批量删除、克隆与 Redis 缓存全解析 - JuiceFS
    在 AI 训练、数据集管理等大规模文件访问场景中,随着文件数量和访问并发增加,元数据层往往更早成为性能瓶颈。无论是删除百万级小文件、克隆大规模数据集,还是高并发目录遍历,元数据引擎的响应能力都会直接影响上层业务效率。 JuiceFS 社区版 1.4 在元数据引擎层面引入了三项优化:批量删除(Batc  ( 2 min )
    大话电容传感器和电容SOC芯片,看这一篇就够了 - FreakStudio
    从口袋里的手机屏幕,到工厂里不知疲倦的生产线;从农田里监测土壤的探头,到管道中识别介质的开关,电容传感器早已成为现代社会中 “看不见的感知触手”。它无需物理接触,却能感知万物的细微变化,以多样的形态适配着消费电子、工业控制、农业监测等无数场景,渗透到我们生活与生产的每一个角落。  ( 4 min )
    旧手机别扔!用 Termux 搭个私人云盘,比网盘香多了 - 一名程序媛呀
    你是不是也有台旧安卓在抽屉里吃灰?其实只要装上Termux,再跑个FileBrowser,十分钟就能变成一台7×24小时在线的私人云盘,再也不用忍受网盘限速和和谐。这篇文章不仅手把手带你实战搭建,还会横向对比几种常见方案,把我踩过的坑、最稳的配置一次讲清楚,让你少走冤枉路。  ( 2 min )
    Docker--搭建私有镜像中心Harbor - NE_STOP
    公网镜像中心 Docker Hub 和阿里云都是Docker 的公网镜像中心,用户可以将镜像push到镜像中心的镜像仓库里,可以将仓库设为私有库,他人无法访问,保证镜像安全。 但是,实际工作中,一般在公司内部搭建一个私有镜像中心。 发布镜像到阿里云 注册一个阿里云账号,并完成实名认证 创建镜像中心实  ( 2 min )
    我有一点隐隐的不安:AI时代,知识会不会被少数人“圈养”起来? - 奔跑的小虾米
    我有一点隐隐的不安:AI时代,知识会不会被少数人“圈养”起来? 引言 在这个大模型发展日新月异的时代,国内外的模型层出不穷——国外有GPT、Claude、Gemini,国内有GLM、Qwen、Minimax等等。它们帮我们处理很多事情,让我们能更高效地应对生活。 但大模型越来越强大,人们也越来越依赖  ( 2 min )
    Claude Code v2.1.139 深度解读:Agent 视图 + /goal 命令的架构逻辑 - 码哥字节
    说实话,我第一次看到 Claude Code v2.1.139 的 changelog,以为只是个普通版本更新——新功能扫了一眼,Agent 视图和 /goal 命令,感觉不就是「任务管理器」和「批量执行」嘛,有什么大惊小怪的。 结果真正用了两天,才发现自己浅了。 这次更新不是在 Claude Co  ( 3 min )
    沪漂五周年了:我越来越迷茫了 - 程序员海军
    今天,是我来上海沪漂的第五年。 五年前刚来上海的时候,我心里更多是兴奋。那时候觉得,只要自己肯学、肯卷、肯熬夜,把 Vue、React、Node、小程序、工程化这些东西学明白,在大城市站稳脚跟应该不是一件特别难的事。 现在五年过去了,我从前端做到全栈,又开始往 AI 应用方向转。我做过 AI 查询、  ( 1 min )
    Agentic Skill Routing 实战:别再把所有 Skill 塞进 AI Agent 上下文 - AI小老六
    把低频 Skill 变成可检索冷存储,让 Agent 按需找回能力,省上下文也更稳。 原文链接:AI小老六 导语 Agent 的能力越来越像一个小型操作系统:它能读文件、调接口、写代码、查日历,也能按团队经验执行一套固定流程。Skill 就是把这些经验沉淀下来的常见方式。 问题也随之出现。Skill  ( 3 min )
    MySQL-Seconds_behind_master的精度误差 - 无毁的湖光-Al
    MySQL-Seconds_behind_master的精度误差 前言 Seconds_behind_master是我们观察主从延迟的一个重要指标。但任何指标所能表示的精度都是有限的。例如用精度只能到秒的指标去衡量毫秒级的表现就会产生非常大的误差。如果再以此误差去分析问题,就会让思维走上弯路。例如用  ( 2 min )
    [MAF预定义ChatClient中间件-03]CachingChatClient——利用缓存省钱省时间 - Artech
    我们知道LLM的调用不仅仅是一个耗时的操作,还会产生一定的费用,所以我们希望能够尽可能地减少不必要的调用。`CachingChatClient`就是为此而生的一个中间件实现,它通过在内存中维护一个缓存来存储之前调用LLM的输入和输出,从而避免了对相同输入的重复调用。当我们调用`GetResponse...  ( 3 min )
    AI的至暗历史:从万众期待到被政府撤资,AI的两次死亡徘徊 - 岳小哥AI
    从AlphaGo的棋局到日常的对话与绘画,我们正享受着AI带来的便利,但你是否想过,这位无所不能的‘大脑’,是如何从七十年前那个夏天的纸上谈兵,一步步成长起来的?  ( 1 min )
  • Open

    Secure Implementation Of Link Previews In Chat Applications
    Don’t let link previews pwn your infrastructure — a complete security guide. Continue reading on Medium »
    API Fuzzing for Security Testing: Part 1: Recon, Discovery & Mapping the Attack Surface
    Learn how to discover undocumented APIs, enumerate endpoints, uncover exposed Swagger files, and build a complete attack surface map before Continue reading on OSINT Team »
    Password Reset Flow Testing — The Most Overlooked Account Takeover Vulnerability
    Hackers don’t always crack passwords. Sometimes they just click “Forgot Password?” and walk right in through a broken back door. Continue reading on OSINT Team »
    Session Management Testing: How Session Tokens Work, How Hackers Steal Them, and How Security…
    When you log in to a website, you don’t have to enter your username and password on every page. Continue reading on OSINT Team »
    HTB File Upload Attacks Skills Assessment Walkthrough
    Chaining SVG XXE, source code disclosure, Apache misconfiguration, and EXIF injection to achieve remote code execution. Continue reading on MeetCyber »
    You Don’t Need More Tools. You Need Better Habits.
    I had 47 tools installed. Found nothing. Changed three habits. Started finding bugs. Continue reading on Medium »
    The Definitive Guide to WSL 2 for Cybersecurity and Development: Installation and Commands
    Master WSL 2 for cybersecurity and development. Learn how to install it, optimize its resources, and run essential commands. Continue reading on Medium »
  • Open

    Secure Implementation Of Link Previews In Chat Applications
    Don’t let link previews pwn your infrastructure — a complete security guide. Continue reading on Medium »
    Complete Wazuh Setup Guide for Windows and Linux Endpoints
    Security monitoring is one of those topics that seems overwhelming at first. Continue reading on Medium »
    Windows CMD-Only Low-Noise Enumeration
    Windows CMD-Only Low-Noise Enumeration Handbook Continue reading on Medium »
    ​ Automated Reconnaissance: Advanced SQL Injection Dumping with SQLmap ​Manual exploitation of…
    sqlmap -r request.txt --batch Continue reading on Medium »
    Advanced Evasion Tradecraft: Precision Module Stomping
    Mapping process memory for calculated, stable execution Continue reading on Medium »
    Web Server Attacks- 1 Walkthrough | TryHackMe
    Enumerate and identify misconfigurations across Apache, Nginx, Node.js, and Python HTTP Server. Continue reading on Medium »
    Webverse-Pro Inked
    lab write-up Continue reading on Medium »
  • Open

    GnuTLS OCSP stapling accepts unrelated SingleResponse (no cert-ID binding)
    curl disclosed a bug submitted by argus-systems: https://hackerone.com/reports/3784125
    CURLOPT_PROXY_CRLFILE / CURLOPT_PROXY_ISSUERCERT / CURLOPT_PROXY_ISSUERCERT_BLOB silently ignored on backends that don't support them
    curl disclosed a bug submitted by bagder: https://hackerone.com/reports/3717552
    Shared HSTS cache accessed without lock
    curl disclosed a bug submitted by bagder: https://hackerone.com/reports/3718265
    RTSP Digest auth state leaks across origins on reused libcurl easy handle
    curl disclosed a bug submitted by hamaowo: https://hackerone.com/reports/3776535
    TFTP upload ignores --continue-at / CURLOPT_RESUME_FROM and leaks skipped local file prefix
    curl disclosed a bug submitted by bowen111: https://hackerone.com/reports/3776433
    libcurl 8.20.0 ignores HTTP Digest domain protection space and preemptively leaks Digest auth outside the declared scope
    curl disclosed a bug submitted by skksndk: https://hackerone.com/reports/3774977
    CURLOPT_COOKIE leaked to cross-origin redirect target CURLOPT_UNRESTRICTED_AUTH bypass for the STRING_COOKIE path
    curl disclosed a bug submitted by azraelxuemo: https://hackerone.com/reports/3766065
  • Open

    GnuTLS OCSP stapling accepts unrelated SingleResponse (no cert-ID binding)
    curl disclosed a bug submitted by argus-systems: https://hackerone.com/reports/3784125
    CURLOPT_PROXY_CRLFILE / CURLOPT_PROXY_ISSUERCERT / CURLOPT_PROXY_ISSUERCERT_BLOB silently ignored on backends that don't support them
    curl disclosed a bug submitted by bagder: https://hackerone.com/reports/3717552
    Shared HSTS cache accessed without lock
    curl disclosed a bug submitted by bagder: https://hackerone.com/reports/3718265
    RTSP Digest auth state leaks across origins on reused libcurl easy handle
    curl disclosed a bug submitted by hamaowo: https://hackerone.com/reports/3776535
    TFTP upload ignores --continue-at / CURLOPT_RESUME_FROM and leaks skipped local file prefix
    curl disclosed a bug submitted by bowen111: https://hackerone.com/reports/3776433
    libcurl 8.20.0 ignores HTTP Digest domain protection space and preemptively leaks Digest auth outside the declared scope
    curl disclosed a bug submitted by skksndk: https://hackerone.com/reports/3774977
    CURLOPT_COOKIE leaked to cross-origin redirect target CURLOPT_UNRESTRICTED_AUTH bypass for the STRING_COOKIE path
    curl disclosed a bug submitted by azraelxuemo: https://hackerone.com/reports/3766065
  • Open

    I Started Learning AWS and Realised I Didn’t Fully Understand the Internet
    No content preview
    Host & Network Penetration Testing: System-Host Based Attacks CTF 1 — eJPT (INE)
    No content preview
    “Bug Bounty Bootcamp #44: No Login?
    You stumble on a login page. No “Register”, no “Forgot Password”. Just two lonely text boxes staring back at you. Most hunters give up… Continue reading on InfoSec Write-ups »
  • Open

    I Started Learning AWS and Realised I Didn’t Fully Understand the Internet
    No content preview
    Host & Network Penetration Testing: System-Host Based Attacks CTF 1 — eJPT (INE)
    No content preview
    “Bug Bounty Bootcamp #44: No Login?
    You stumble on a login page. No “Register”, no “Forgot Password”. Just two lonely text boxes staring back at you. Most hunters give up… Continue reading on InfoSec Write-ups »

  • Open

    Security Assessment Report: Full Exploit Chain (SQLi to RCE)
    Researcher: Hashem Ali Kahil / Orion7715 Continue reading on Medium »
    My Instructor Said “You Can’t Get a Shell.” I Got Root. — Full Web Pentest Exam Write-Up
    Author: Shikhali Jamalzade GitHub: github.com/alisalive LinkedIn: linkedin.com/in/camalzads Continue reading on Medium »
    Master Guide: Solving All PortSwigger CORS Vulnerabilities
    Lab 1: CORS Vulnerability with Basic Origin Reflection Continue reading on Medium »
    Race Conditions Are Not Only Race Conditions
    Concurrency is the cheapest way to wedge a state machine apart. The window is never the bug. Continue reading on Medium »
    Mastering CSRF: The Ultimate PortSwigger Academy Walkthrough Guide Community Edition
    Cross-Site Request Forgery (CSRF) is one of the most critical web vulnerabilities that allows an attacker to induce users to perform… Continue reading on Medium »
    File Upload Attacks: How to Secure Them Properly
    A practical breakdown of real-world defenses that stop file upload attacks when implemented together. Continue reading on MeetCyber »
    From CVE Matching to Exploit Validation: How Vulnerability Scanners Are Evolving
    I’ve been integrating OnScanner into my workflow recently as part of external security assessment and bug bounty reconnaissance, and it… Continue reading on Medium »
    From $0 to $14,000: How a Blind SSRF in a PDF Generator Cracked a Corporate Internal Network
    A few months ago, I was looking at a private program on Bugcrowd that managed financial data for enterprise clients. The scope was massive… Continue reading on Medium »
    Full server takeover ssh leak (dell technologies) by (RECON)
    السلام عليكم ورحمة الله Continue reading on Medium »
  • Open

    Support — HackTheBox Write-Up
    Difficulty: Easy | OS: Windows (Active Directory) Continue reading on Medium »
    Understanding Process Creation and Manipulation in Windows: A Deep Dive into CreateProcess()
    Introduction Continue reading on Medium »
    How Nmap Actually Knows What’s Running on a Port (And Why It’s Sometimes Wrong)
    Most people trust the service name Nmap gives them. Experienced testers treat it as a hypothesis. Continue reading on MeetCyber »
    Threat Intelligence Brief: CVE-2026-40072-SSRF Vulnerability in web3.py
    Vulnerability: CVE-2026–40072 Affected Software: web3.py Severity: High (CVSS 3.1: 7.5) / Medium (CVSS 4.0: 6.9) Patched Versions: 7.15.0… Continue reading on Medium »
    What does a Threat Intelligence Report actually consist of?
    Hey there, Continue reading on Medium »
    AWS S3 Multi-Bucket Compromise via GraphQL Federation SSRF
    Pendahuluan Continue reading on FMI Cyber Security Consulting Services »
    Webverse-Pro Voxlink
    lab write-up Continue reading on Medium »
  • Open

    Agent OS :五种驯服不确定性的范式 - 罗西的思考
    Agent OS :五种驯服不确定性的范式 目录Agent OS :五种驯服不确定性的范式0x00 概要0x01 Part 1: 问题空间1.1 不确定性的六个来源1.2 三个独有问题1.3 跨领域全景:计算机中"驯服不确定性"的经典实践1.4 分布式系统深度对标1.4.1 8 个经典问题全景对照1  ( 13 min )
    数据库即时编译JIT - aixueforever
    JIT 是 Just-In-Time Compilation,即“即时编译”。意思是:程序运行到某段代码时,才把原本解释执行的逻辑编译成 CPU 可以直接执行的本机机器码。PostgreSQL 文档给的典型例子是:不用一个通用表达式解释器去判断 WHERE a.col = 3,而是为这个具体条件生成  ( 7 min )
    [Begin]AI Learn Data Day 0 - 粉紫系超人气月兔铃仙
    [Begin]AI Learn Data Day 0 最近一直在学 AI 全栈开发。关于全栈,略懂一二;关于 AI,则完全是个新手。 说到底我还没学完全栈,后端的内容比如 Redis 根本没学,只知道是个什么东西,教学视频太长看不下去。然后转到 AI 这边,或许因为是新方向的缘故,网上连一篇成体系的  ( 3 min )
    深度学习进阶(二十七)现代 LLM 的核心架构设计其二:SwiGLU - 哥布林学者
    上一篇我们看了现代大模型对归一化的改造。 RMSNorm 去掉了均值中心化,只保留均方根缩放:一个沿用多年的标准组件,拆开一看,其中一部分工作在现代整体架构中已经有些多余了。 本篇来看第二个改动:Transformer 架构中的 FFN (MLP) 层的重构,而其具体内容,需要先从激活函数说起。 1  ( 3 min )
    刚开始,你以为只是换工作 - Zhang_Xiang
    如果,我是说如果,AI 取代你现在的工作。 金属机器人走进办公室,坐在你的工位上,抢走你的键盘,这是电影。 但在真实的商业世界里,这种事情往往悄无声息,你甚至以为,自己只是和往常一样,换了一份工作。 过去几十年里,大家形成的认知是掌握一样专业技能就可以作为职业的壁垒,并且这套逻辑在过往中,经过验证是  ( 1 min )
    我问了 AI 一个问题:编码能力贬值后,什么能力值钱? - 微山湖上静悄悄
    这不是一篇教你"怎么用 AI 写代码"的文章。 这是一篇关于程序员的核心竞争力,到底该往哪投的思考。 我是怎么开始想这个问题的 前几天,我用 Claude Code 做了一个实验。 从零开始,不写一行代码,纯靠自然语言描述需求,花了大约 2 小时,做出了一个完整的智慧校园 3D 数据大屏。 效果包含  ( 2 min )
    Means:基于 .NET 10 打造的开源自部署 S3 兼容对象存储服务 - 239573049
    Means:基于 .NET 10 打造的开源自部署 S3 兼容对象存储服务 GitHub 地址:https://github.com/AIDotNet/Means 欢迎 Star ⭐、Fork、提 Issue 和 PR!MIT 开源协议,放心使用。 引言 对象存储已经成为现代云原生架构的基石。  ( 3 min )
    ICLR 2026 | LiveMoments 用参考图引导的扩散模型提升重选封面帧画质 - vivo互联网技术
    我们团队提出了 LiveMoments,这是首个专门针对 Live Photo 重选封面帧画质修复的解决方案,已被 ICLR 2026 录用。 针对用户重选封面时面临的画质降级痛点,我们利用 Live Photo 自带的原始高清封面作为参考,构建了一个包含运动对齐模块的参考引导扩散模型。 该方法有效...  ( 2 min )
    GitHub Copilot 新版计费翻车:2天消耗49美元,天价引爆开发者吐槽,国产开源 AI 迎来替代窗口期 - iNeuOS工业互联网系统
    按照新版计费规则,Copilot Pro+订阅用户 6.1~6.2 短短两天,7000 内置 Credits 全额耗尽、10 美元附加预算全部透支,折算实际产生计费成本超 49 美元。海外开发者同样发文吐槽,3 天掏空月度全部代币。  ( 1 min )
    筛完 1400+ Skills,这 5 个让 Claude Code 效率提升 3 倍 - 码哥字节
    三个月前,我第一次看到 superpowers 的时候,立刻觉得这东西有点厉害——14 个 skill 装进去,Claude Code 直接变成一套完整的软件工程流水线,TDD、代码审查、brainstorming 全覆盖。 GitHub 上已经 187K star 了,Medium 上的测评文章写  ( 3 min )
    agent 进阶:Plan-and-Execute 适合什么样的场景 - it排球君
    前言 本文主要描述Plan-and-Execute开发中的ReAct模式,并且使用一个demo,彻底搞懂怎么在实际工作中使用Plan-and-Execute模式 话不多说,我们开始 代码结构 代码地址 . ├── main.py # 主入口,串起规划、执行、分析三个阶段 ├── planner.py  ( 2 min )
    你的FastAPI又在服务器上“跑不起来”了?来,今天咱把打包这件事彻底聊透 - 一名程序媛呀
    还在为FastAPI打包后各种报错抓狂?本文不讲虚的,用最接地气的方式对比Linux和Windows下三种主流打包部署方式,从PyInstaller的各种坑到Docker的一键起飞,手把手教你如何根据场景选择,附上保姆级实操命令,让你的服务“稳如老狗”。  ( 2 min )
    Cloud Agent 开发笔记(3):Web 交互与数据持久化 - 五岳
    Cloud Agent 开发笔记(3):Web 交互与数据持久化 上一篇搭好了 Agent 引擎和 Tool 体系:query() 能跑、Tool 能调、安全有 pathGuard 兜底。但这一切都是在服务端发生的。Agent 生产的事件怎么到浏览器?消息怎么存才能让用户刷新页面不丢?多个会话同时跑  ( 3 min )
    CodeGraph 代码图谱实战:AI Agent 为什么不该再从 grep 开始? - AI小老六
    用代码图谱重构 Agent 检索链路,让大仓库定位、调用链和影响面分析更稳更省。 原文链接:AI小老六 导语 代码 Agent 真正卡住的地方,往往不是模型不会写代码,而是它不知道该先看哪里。 面对一个陌生仓库,Agent 通常会先列目录,再搜关键词,接着打开几个文件,猜一个入口点,发现不对后再重复  ( 3 min )
    基于Blazor实现的跟踪光伏智能运维平台 - known
    一、项目背景 随着光伏电站规模不断扩大,传统的人工巡检和运维方式已无法满足电站高效运行的需求。特别是跟踪式光伏电站,核心依赖追日跟踪系统来最大化发电效率,对实时监控、故障预警和远程控制的要求更高。一个集数据采集、实时监控、报警管理、历史分析和远程控制于一体的智能运维平台,成为行业刚需。 我们基于 B  ( 2 min )
    [MAF预定义ChatClient中间件-02]FunctionInvokingChatClient——实现ReAct循环和人机交互的大功臣 - Artech
    在众多预定义的`IChatClient`中间件中,`FunctionInvokingChatClient`无疑是最重要的一个,以至于没有它整个Agent就无法工作了。原因在于驱动Agent执行的核心机制的ReAct循环就是通过`FunctionInvokingChatClient`实现的,我们注册的...  ( 3 min )
    ChatGPT一夜爆火,是算力、数据、算法七十年的漫长等待后的爆发 - 岳小哥AI
    2022年年底,ChatGPT横空出世,两个月月活破亿,成为人类有史以来增长最快的互联网产品。 一夜之间,全世界都在讨论AI,都想着和ChatGPT对话。 ChatGPT是一夜爆火,但并不是凭空冒出来的。 上篇文章《AI不是从天而降,它经历了七十年三起三落:读懂AI的第三课》,我们知道从1950年图  ( 1 min )
  • Open

    “Bug Bounty Bootcamp #43: Login Page?
    Let’s be real — you’ve hit that login wall more times than you’ve hit “snooze” on a Monday morning. Continue reading on InfoSec Write-ups »
    I Bought a ₹1,599 Government Book for ₹1. The Server Approved It.
    No content preview
    I Typed 000000 and the App Thought MFA Was Already On
    No content preview
    Setting Up n8n Locally on Kali Linux Using Docker
    Continue reading on InfoSec Write-ups »
    API Penetration Testing Checklist: How Real-World Attacks Break APIs Before Scanners Do
    No content preview
    I Finished My Thesis Defense — A Journey to Mobile Forensic
    No content preview
    “Bug Bounty Bootcamp #42: JWT Attacks — How a Stolen Token or a Weak Secret Can Grant You Admin…
    JSON Web Tokens are everywhere — in cookies, Authorization headers, and API calls. Continue reading on InfoSec Write-ups »
    The Ultimate Guide to Stay Hidden Online: TOR and Proxy Chaining
    No content preview
    Frontend Security & Bug Hunting: The .env File Crisis and Real-World Exploitation
    No content preview
    5 Windows Event IDs Every SOC Analyst Should Know (With Real Lab Evidence)
    No content preview
  • Open

    “Bug Bounty Bootcamp #43: Login Page?
    Let’s be real — you’ve hit that login wall more times than you’ve hit “snooze” on a Monday morning. Continue reading on InfoSec Write-ups »
    I Bought a ₹1,599 Government Book for ₹1. The Server Approved It.
    No content preview
    I Typed 000000 and the App Thought MFA Was Already On
    No content preview
    Setting Up n8n Locally on Kali Linux Using Docker
    Continue reading on InfoSec Write-ups »
    API Penetration Testing Checklist: How Real-World Attacks Break APIs Before Scanners Do
    No content preview
    I Finished My Thesis Defense — A Journey to Mobile Forensic
    No content preview
    “Bug Bounty Bootcamp #42: JWT Attacks — How a Stolen Token or a Weak Secret Can Grant You Admin…
    JSON Web Tokens are everywhere — in cookies, Authorization headers, and API calls. Continue reading on InfoSec Write-ups »
    The Ultimate Guide to Stay Hidden Online: TOR and Proxy Chaining
    No content preview
    Frontend Security & Bug Hunting: The .env File Crisis and Real-World Exploitation
    No content preview
    5 Windows Event IDs Every SOC Analyst Should Know (With Real Lab Evidence)
    No content preview

  • Open

    .NET 8 Web开发入门(七):安全门禁——JWT 身份验证与授权实战 - 码农刚子
    本文将深入浅出地讲解 JWT(JSON Web Token)在 .NET 8 中的应用。从原理到代码实战,手把手教你搭建用户登录接口颁发 Token,并配置 API 网关验证 Token,最终实现基于角色的接口权限控制,保护你的 API 不被非法访问。  ( 3 min )
    MacOS 安装Seismic Unix - GeoFXR
    MacOS 安装Seismic Unix ​ 目前,Seismic Unix 适用于Unix系统及其衍生系统,包括Ubuntu,RedHat等Linux发行版,苹果公司的Mac OS系统,微软Windows10 系统。 ​ 由于网上广泛存在的都是基于Linux系统的安装教程,而Mac OS作为时下科  ( 1 min )
    Harness Engineering:把 AI 真正接进工程流程 - SharpCJ
    一、为什么要谈 Harness Engineering 这两年,团队里关于 AI 的讨论很多,但真正落到研发现场,问题往往很具体: 为什么 AI 有时看起来很能干,有时又完全不靠谱? 为什么它能写出一段像样的代码,却经常交不出一个可验证的结果? 为什么同一个任务,换个人问、换种说法,结果差异会这么大  ( 5 min )
    代码之外:一个技术人的职场困境与自我和解 - charlee44
    1. 价值实现 我越来越能理解,为什么有些员工在公司干得不错、待遇也还可以,但最后还是选择了离开。这背后其实是一个价值实现的问题。 举个例子:你和上级都想把领导交代的事情做好,但你们做事的方法完全不同。你信奉“第一性原理”,喜欢从根本上解决问题,追求无冗余、渐进迭代和可扩展;而在上级眼里,可能更看重  ( 1 min )
    ThingsBoard 规则链系统源码分析和自定义定时器 - daidaidaiyu
    规则链系统 https://thingsboard.io/docs/user-guide/rule-engine-2-0/re-getting-started/?scriptfunctionfilterconfig=anonymous 规则节点总共分为6大类 Filter nodes 根据不同条件过  ( 14 min )
    LangChain 入门 Agent 的基本运行机制 - VipSoft
    目录安装依赖添加环境变量加载环境变量定义工具创建 Agent发起调用完成代码如下 Python - UV 为每个项目创建独立、干净的Python工作空间 UV 中文文档:https://uv.doczh.com/ 安装依赖 cd D:\OpenSource\Python\VipLangChain #  ( 3 min )
    【Azure App Service】应用服务中的SNAT (Source Network Address Translation 源网络地址转化) - 编码者卢布
    App Service 应用经常需要访问外部服务,比如 Azure SQL、Redis、Storage 或第三方 API。很多人会以为应用是直接从 worker 实例访问公网,但实际上并不是这样。 App Service 的 worker 实例运行在 scale unit / stamp 内部,通常  ( 2 min )
    C++ 最难的 20 个概念 完整详细讲解 - 异地远程联网技术
    以下是为你一次性整理的 20 个最难 C++ 概念 的详细文章。每条包含:核心概念、为什么难、代码示例、关键要点 和 版本演进。内容精炼但深入,适合中高级开发者。 1. 模板元编程 (Template Metaprogramming) 为什么难:把编译期变成图灵完备的“编程语言”,调试极差,错误信息  ( 3 min )
    AI编程系列01:裸 API 账单场景下,如何自建 LLM 用量可视化看板 - AlfredZhao
    很多开发者接入聚合型 LLM API 后,会遇到一个很现实的问题:服务商只给账单接口,不提供 Web 管理后台。Token 消耗、扣费明细、成本趋势都藏在 JSON 里,看得见数据,却看不清变化。 按照上篇文章《氛围编程实战系列:先规划清楚学习路径》这个规划路径,我们今天就来用 AI 先开发一个 L  ( 2 min )
    告别 oh-my-openagent 版本地狱:一套纯 OpenCode 配置实现 Agent 软路由 - 我才是银古
    背景 如果你正在用 OpenCode,大概率听说过甚至用过 oh-my-openagent——一个为 OpenCode 提供多 Agent 编排能力的插件。它的核心思路非常好:意图门控、只读隔离、并行探索、结构化输出……属实把"让 AI 自己调度 AI"这件事玩明白了。 但问题也很头疼:太不稳定了。  ( 1 min )
    RAG系列:#6 一文搞懂RAG存储技术:文件、元数据、切片、向量 - AI-Frontiers
    ​ 原文链接:https://mp.weixin.qq.com/s/gs-yR2R-ZTJeYx0r2ow1PA 欢迎关注公zh: AI-Frontiers RAG往期文章推荐 RAG效果差?7个指标让你的准确率大幅提升 RAG评测完整指南:指标、测试和最佳实践 收藏!RAG核心工具大全: 7大解析  ( 3 min )
    里程碑式更新!Dash 4.2新版本新增websocket型回调 - 费弗里
    更多Dash应用开发干货知识、案例,欢迎关注“玩转Dash”微信公众号 1 简介 大家好我是费老师。Dash作为流行的Python全栈应用开发框架,于近日刚发布了其4.2.0正式版本,在本次新版本中,新增了颇具里程碑✨意义的websocket型回调函数机制,使得我们可以在Dash应用回调函数逻辑  ( 2 min )
    《图解HTTP》--第5章-与HTTP协作的Web服务器 - lunzi_fly
    读《图解HTTP》:代理、网关、隧道与缓存到底是什么? 本文是阅读《图解HTTP》第 5 章后的学习整理,结合个人理解做了少量补充。文中的流程图和表格用于概括本章概念,不使用原书截图,也不替代原书内容。 这篇文章解决什么问题 我们平时访问一个网站时,直觉上会以为浏览器直接连到了目标服务器。实际情况往  ( 2 min )
    一次函数图像工厂:用 SymPy 自动生成 y=kx+b 对比动画 - wang_yb
    你在用 Manim 制作一次函数图像的对比动画时,是不是也遇到过这种麻烦:想直观展示不同斜率 k 和截距 b 对直线的影响,但每改一个参数,都得重新手算两端点坐标、重新算与坐标轴的交点,甚至要凭感觉“拉长”线段保证它贯穿画面。 改三组参数,工作量就翻三倍。 今天这篇文章,就是要彻底解决这个体力活。我  ( 3 min )
    我用 AI 十分钟生成了一个科幻风智慧城市 3D 大屏,附完整 Prompt 和源码解析 - 微山湖上静悄悄
    前言 最近在探索用 AI 辅助前端开发,尝试了一个有意思的实验:完全通过自然语言描述,让 AI 生成一个科幻电影风格的智慧城市 3D 数据大屏。效果出乎意料——不仅有完整的 Three.js 3D 城市场景,还有数据面板、交互控制、动画效果,而且代码质量远超预期。 先看最终效果: 济南市智慧城市大脑  ( 3 min )
  • Open

    The Gmail “Zombie Token” Google Refused to Kill
    Here is how I found a persistent authorization bypass in Gmail’s ‘Send Mail As’ feature, documented it with video proof, and got it closed… Continue reading on OSINT Team »
    A Unique Case of One-Click Account Takeover via CSRF Chain
    Hello everyone! Today I want to walk through an account takeover vulnerability I found recently. At its core, the issue was a CSRF… Continue reading on Medium »
    File Upload Attacks: Alternative Exploitation Paths (XSS, XXE, DoS)
    What happens when the upload form is actually secure — and why that doesn’t mean you’re out of options. Part 6 of the File Upload Attacks… Continue reading on MeetCyber »
    The Easiest $50K You’ll Ever Make? Hacking AI Systems
    The new AI Bug Bounty Gold Rush Continue reading on Medium »
    The Most Dangerous OAuth Bug I’ve Ever Found
    Hello! Some of you might know me (just kidding). Continue reading on Medium »
    IDOR Sounds Fancy But It’s Just Broken Math
    Here’s how to find them without overthinking. Continue reading on Medium »
    Finding an Unintended Privilege Escalation While Solving Mobile Hacking Lab’s Food Store Challenge
    While hunting for SQL Injection, I accidentally became the CEO of the Food Store. Continue reading on Medium »
    From Zero to $8,500: Exploiting a Flawed Reset Token Validation to Bypass MFA and Take Over…
    It was late on a Friday night, and I was deep-diving into a private bug bounty program for a major B2B SaaS platform. Continue reading on Medium »
    File Upload Attacks: Bypassing Type Filters
    When the server stops trusting the filename and starts looking at what’s inside — and why that’s still not enough. Part 5 of the File… Continue reading on MeetCyber »
  • Open

    When Your Login System Becomes the Attacker: PAM Modules as a Double-Edged Sword
    In the last article, we examined how nginx modules, a tool designed to serve web traffic, can be exploited by an attacker who has gained a… Continue reading on Medium »
    AI Red Teaming in 2026: A Practical Guide to Prompt Hacking, Jailbreaks, and Defending LLM…
    How adversaries break AI systems, and how you build ones that fight back Continue reading on Medium »
    Ghostcat-PWN: When an Old Tomcat Vulnerability Opens the org doors
    From exposed AJP connectors to reliable post-exploitation workflows: the story behind Ghostcat-PWN and our experience with CVE-2020–1938 Continue reading on Medium »
    Understanding Windows API Programming: How Applications Communicate with Windows
    Introduction Continue reading on Medium »
    I Built a Multi-Agent RAG System and Then Red-Teamed It
    From near-perfect RAGAS scores to adversarial attacks, a deep dive into building production-grade RAG and why evaluation isn’t optional. Continue reading on Medium »
    How a Simple Port Scan Led Me to a Critical Active Directory Finding
    Most people think critical vulnerabilities are found using advanced exploits. In reality, many of the highest-impact findings start with… Continue reading on Medium »
    Before I Learned to Hack, I Learned to Investigate
    Understanding the systems behind the tools and the thinking behind the findings. Continue reading on MeetCyber »
  • Open

    Missing access control when linking banners or campaigns to zones
    Revive Adserver disclosed a bug submitted by darky_os: https://hackerone.com/reports/3650504
    Missing access control when linking trackers to campaigns
    Revive Adserver disclosed a bug submitted by darky_os: https://hackerone.com/reports/3650582
    Blind SQL injection via clientid parameter in zoneinclude.php
    Revive Adserver disclosed a bug submitted by titanrain: https://hackerone.com/reports/3653196
    Reflected XSS via clientid parameter in zoneinclude.php
    Revive Adserver disclosed a bug submitted by titanrain: https://hackerone.com/reports/3653316
    PHP code injection via delivery limitation logical
    Revive Adserver disclosed a bug submitted by 0x4c616e: https://hackerone.com/reports/3656781
    Stored XSS via Full Name field in userlog email entries
    Revive Adserver disclosed a bug submitted by 3l4: https://hackerone.com/reports/3669623
    Session ID reuse allowing XMLRPC API authentication bypass
    Revive Adserver disclosed a bug submitted by 0x4c616e: https://hackerone.com/reports/3672641
    Missing access control when modifying parent entities via XMLRPC
    Revive Adserver disclosed a bug submitted by 3l4: https://hackerone.com/reports/3677576
    Banner status override by advertiserlevel users
    Revive Adserver disclosed a bug submitted by v3rtical: https://hackerone.com/reports/3678828
    Stored XSS via malicious usernames in audit log details + Username validation bypass in XMLRPC addUser
    Revive Adserver disclosed a bug submitted by 3l4: https://hackerone.com/reports/3680090
    PHP code injection via unexpected delivery limitation parameter
    Revive Adserver disclosed a bug submitted by rajib_mahmud: https://hackerone.com/reports/3744200
    PRE_PROXY change leaks stale Proxy Digest state across proxy-chain boundary
    curl disclosed a bug submitted by hungly09: https://hackerone.com/reports/3777381
    curl/libcurl 8.20.0 NOPROXY bypass via uppercase-hex IPv4 aliases leaks off-proxy Basic credentials to the configured proxy
    curl disclosed a bug submitted by arkss: https://hackerone.com/reports/3773293
    SMTP connection reuse ignores --ssl-reqd / CURLOPT_USE_SSL and reuses a clear-text STARTTLS session on current master
    curl disclosed a bug submitted by hualuo: https://hackerone.com/reports/3770979
    Proxy CONNECT response poisoning via authentication retry in cf-h1-proxy.c (libcurl)
    curl disclosed a bug submitted by lvtable: https://hackerone.com/reports/3767963
  • Open

    Missing access control when linking banners or campaigns to zones
    Revive Adserver disclosed a bug submitted by darky_os: https://hackerone.com/reports/3650504
    Missing access control when linking trackers to campaigns
    Revive Adserver disclosed a bug submitted by darky_os: https://hackerone.com/reports/3650582
    Blind SQL injection via clientid parameter in zoneinclude.php
    Revive Adserver disclosed a bug submitted by titanrain: https://hackerone.com/reports/3653196
    Reflected XSS via clientid parameter in zoneinclude.php
    Revive Adserver disclosed a bug submitted by titanrain: https://hackerone.com/reports/3653316
    PHP code injection via delivery limitation logical
    Revive Adserver disclosed a bug submitted by 0x4c616e: https://hackerone.com/reports/3656781
    Stored XSS via Full Name field in userlog email entries
    Revive Adserver disclosed a bug submitted by 3l4: https://hackerone.com/reports/3669623
    Session ID reuse allowing XMLRPC API authentication bypass
    Revive Adserver disclosed a bug submitted by 0x4c616e: https://hackerone.com/reports/3672641
    Missing access control when modifying parent entities via XMLRPC
    Revive Adserver disclosed a bug submitted by 3l4: https://hackerone.com/reports/3677576
    Banner status override by advertiserlevel users
    Revive Adserver disclosed a bug submitted by v3rtical: https://hackerone.com/reports/3678828
    Stored XSS via malicious usernames in audit log details + Username validation bypass in XMLRPC addUser
    Revive Adserver disclosed a bug submitted by 3l4: https://hackerone.com/reports/3680090
    PHP code injection via unexpected delivery limitation parameter
    Revive Adserver disclosed a bug submitted by rajib_mahmud: https://hackerone.com/reports/3744200
    PRE_PROXY change leaks stale Proxy Digest state across proxy-chain boundary
    curl disclosed a bug submitted by hungly09: https://hackerone.com/reports/3777381
    curl/libcurl 8.20.0 NOPROXY bypass via uppercase-hex IPv4 aliases leaks off-proxy Basic credentials to the configured proxy
    curl disclosed a bug submitted by arkss: https://hackerone.com/reports/3773293
    SMTP connection reuse ignores --ssl-reqd / CURLOPT_USE_SSL and reuses a clear-text STARTTLS session on current master
    curl disclosed a bug submitted by hualuo: https://hackerone.com/reports/3770979
    Proxy CONNECT response poisoning via authentication retry in cf-h1-proxy.c (libcurl)
    curl disclosed a bug submitted by lvtable: https://hackerone.com/reports/3767963

  • Open

    DeleGator: Building and Testing a Linux-Native Kerberos Delegation Abuse Framework
    A tool release, lab build, and measured OPSEC testing writeup covering all three Active Directory delegation attack types — with real… Continue reading on Medium »
  • Open

    Incomplete fix for CVE-2022-35406: meta-redirect content-type check bypassable via parameter injection
    PortSwigger Web Security disclosed a bug submitted by hacker-kartel: https://hackerone.com/reports/3775183
    page.line.me Open Redirect Leading to OAuth Authorization Code Exposure and Access Token Compromise
    LY Corporation disclosed a bug submitted by imnotr3al: https://hackerone.com/reports/3423013 - Bounty: $1000
  • Open

    Incomplete fix for CVE-2022-35406: meta-redirect content-type check bypassable via parameter injection
    PortSwigger Web Security disclosed a bug submitted by hacker-kartel: https://hackerone.com/reports/3775183
    page.line.me Open Redirect Leading to OAuth Authorization Code Exposure and Access Token Compromise
    LY Corporation disclosed a bug submitted by imnotr3al: https://hackerone.com/reports/3423013 - Bounty: $1000
  • Open

    How I was able to Modify Ratings on a Target and Cause Business Impact
    Learn how I found this interesting bug Continue reading on InfoSec Write-ups »
    Bug Bounty Bootcamp #41: Remote Command Execution — From Innocent Inputs to Full Server Takeover
    A stock checker that pings an IP. A comment box that echoes your name. These simple features hide a terrifying truth: they might be… Continue reading on InfoSec Write-ups »
    The KQL Query That Caught 260 Brute Force Attempts in Microsoft Sentinel
    No content preview
    Auth Mastery Part 2: Sessions, Cookies, and Staying Authenticated
    No content preview
    AI Threat Modelling: A Practical Walkthrough of the TryHackMe Room
    No content preview
    Uncovering the Blind Spot: Bypassing a Security Patch (CVE-2026–24884) to Achieve Arbitrary File…
    No content preview
    One Agent, Five Zero-Days: Turning Past CVEs Into SAST Rules
    No content preview
  • Open

    How I was able to Modify Ratings on a Target and Cause Business Impact
    Learn how I found this interesting bug Continue reading on InfoSec Write-ups »
    Bug Bounty Bootcamp #41: Remote Command Execution — From Innocent Inputs to Full Server Takeover
    A stock checker that pings an IP. A comment box that echoes your name. These simple features hide a terrifying truth: they might be… Continue reading on InfoSec Write-ups »
    The KQL Query That Caught 260 Brute Force Attempts in Microsoft Sentinel
    No content preview
    Auth Mastery Part 2: Sessions, Cookies, and Staying Authenticated
    No content preview
    AI Threat Modelling: A Practical Walkthrough of the TryHackMe Room
    No content preview
    Uncovering the Blind Spot: Bypassing a Security Patch (CVE-2026–24884) to Achieve Arbitrary File…
    No content preview
    One Agent, Five Zero-Days: Turning Past CVEs Into SAST Rules
    No content preview

  • Open

    Missing HMAC validation on /uninstall webhook in Shopify/sample-django-app reference template
    Shopify disclosed a bug submitted by cipher-kid: https://hackerone.com/reports/3697491
    Mentioned unites are at the same time .Then we have to increase the bounty.
    curl disclosed a bug submitted by karthiktp1810: https://hackerone.com/reports/3761789
    TLS conn reuse and session cache ignore fsslctx callback and ssl_config_data flags ( incomplete fix variant of 7541ae569 )
    curl disclosed a bug submitted by fg0x0: https://hackerone.com/reports/3761647
    lib/ldap.c follows attacker-controlled LDAP referrals and binds to a second server; WinLDAP builds leak current logon credentials (confirmed on Window
    curl disclosed a bug submitted by tpfeng: https://hackerone.com/reports/3756699
    Use-after-free in `curl_easy_duphandle()` with HTTP/2 stream-dependency tree
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3751701
    Low priority HSTS bypass in curl_easy_duphandle()
    curl disclosed a bug submitted by ajohnston3825: https://hackerone.com/reports/3769293
  • Open

    Missing HMAC validation on /uninstall webhook in Shopify/sample-django-app reference template
    Shopify disclosed a bug submitted by cipher-kid: https://hackerone.com/reports/3697491
    Mentioned unites are at the same time .Then we have to increase the bounty.
    curl disclosed a bug submitted by karthiktp1810: https://hackerone.com/reports/3761789
    TLS conn reuse and session cache ignore fsslctx callback and ssl_config_data flags ( incomplete fix variant of 7541ae569 )
    curl disclosed a bug submitted by fg0x0: https://hackerone.com/reports/3761647
    lib/ldap.c follows attacker-controlled LDAP referrals and binds to a second server; WinLDAP builds leak current logon credentials (confirmed on Window
    curl disclosed a bug submitted by tpfeng: https://hackerone.com/reports/3756699
    Use-after-free in `curl_easy_duphandle()` with HTTP/2 stream-dependency tree
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3751701
    Low priority HSTS bypass in curl_easy_duphandle()
    curl disclosed a bug submitted by ajohnston3825: https://hackerone.com/reports/3769293
  • Open

    Forensic Implications of Apple Stolen Device Protection
    If you extract data from iPhones for a living, Stolen Device Protection is the change you can no longer afford to ignore. It does something deceptively simple: it puts Face ID or Touch ID in front of the “Trust This Computer” prompt. The practical result is that an examiner who knows the device passcode still […]  ( 12 min )
  • Open

    How Bug Bounty Hunters Are Using Claude Code.
    The community has been quietly building something powerful. I went and found it. Continue reading on InfoSec Write-ups »
    Common Mistakes Made by Cybersecurity Beginners
    No content preview
    Google Dorks Google Ko Bana Do Apna Hacking Tool: Free Mein Bugs Dhundho! (Hinglish Mein)
    No content preview
  • Open

    How Bug Bounty Hunters Are Using Claude Code.
    The community has been quietly building something powerful. I went and found it. Continue reading on InfoSec Write-ups »
    Common Mistakes Made by Cybersecurity Beginners
    No content preview
    Google Dorks Google Ko Bana Do Apna Hacking Tool: Free Mein Bugs Dhundho! (Hinglish Mein)
    No content preview
  • Open

    Vulnerability & Patch Roundup — May 2026
    If you run a website, you know that a single unpatched vulnerability can take your site offline, damage your reputation, or leave you cleaning up after an attack. Most compromises we see start with automated attacks targeting known software flaws, often the same ones that have already been reported and disclosed. To help you stay ahead of these threats, we’ve put together this month’s roundup of critical security updates and vulnerability patches affecting the WordPress ecosystem. Continue reading Vulnerability & Patch Roundup — May 2026 at Sucuri Blog.  ( 24 min )

  • Open

    Blind POST SSRF via Web Push Notification Endpoint
    phpBB disclosed a bug submitted by misop00p: https://hackerone.com/reports/3608558
  • Open

    Blind POST SSRF via Web Push Notification Endpoint
    phpBB disclosed a bug submitted by misop00p: https://hackerone.com/reports/3608558

  • Open

    Removing Docker volume when container is in use
    You started looking the volumes: Then you try to remove: The solution is listing all containers, filtering by volume: Then remove the container: And finally remove the volume:  ( 11 min )

  • Open

    V1Plugin.Decrypt panics on empty ciphertext (Remote DoS)
    AWS VDP disclosed a bug submitted by misop00p: https://hackerone.com/reports/3620748
    V2Plugin.Decrypt panics on empty ciphertext (Remote DoS)
    AWS VDP disclosed a bug submitted by misop00p: https://hackerone.com/reports/3620753
    iOS Brave Playlist "Open in Private Tab" bypasses FaceID requirement for Private Tabs
    Brave Software disclosed a bug submitted by aaront: https://hackerone.com/reports/3693295
  • Open

    V1Plugin.Decrypt panics on empty ciphertext (Remote DoS)
    AWS VDP disclosed a bug submitted by misop00p: https://hackerone.com/reports/3620748
    V2Plugin.Decrypt panics on empty ciphertext (Remote DoS)
    AWS VDP disclosed a bug submitted by misop00p: https://hackerone.com/reports/3620753
    iOS Brave Playlist "Open in Private Tab" bypasses FaceID requirement for Private Tabs
    Brave Software disclosed a bug submitted by aaront: https://hackerone.com/reports/3693295

  • Open

    Downloading iPhone and iPad backups from Apple iCloud
    Pulling a backup out of iCloud is one of the more technically demanding jobs in cloud forensics. An iCloud backup is not a single, ready-to-download file; instead, it is assembled from a large number of separate fragments that have to be collected and stitched back together into a coherent backup. Recent changes to Apple’s communication […]  ( 9 min )

  • Open

    Heap-OOB read in urlapi `redirect_url()` via `CURLU_GUESS_SCHEME` + `CURLU_NO_GUESS_SCHEME` flow
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3751715
    curl GnuTLS backend accepts a clientAuth-only certificate for HTTPS server authentication
    curl disclosed a bug submitted by jingzhou: https://hackerone.com/reports/3752567
    Autotranslate DDP Method Exposes Private Messages Without Authentication or Room Access Check
    Rocket.Chat disclosed a bug submitted by deprrous: https://hackerone.com/reports/3734326
  • Open

    Heap-OOB read in urlapi `redirect_url()` via `CURLU_GUESS_SCHEME` + `CURLU_NO_GUESS_SCHEME` flow
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3751715
    curl GnuTLS backend accepts a clientAuth-only certificate for HTTPS server authentication
    curl disclosed a bug submitted by jingzhou: https://hackerone.com/reports/3752567
    Autotranslate DDP Method Exposes Private Messages Without Authentication or Room Access Check
    Rocket.Chat disclosed a bug submitted by deprrous: https://hackerone.com/reports/3734326

  • Open

    NULL pointer dereference in node:sqlite DatabaseSync#applyChangeset() via malformed SQLite changeset
    Node.js disclosed a bug submitted by junius: https://hackerone.com/reports/3736889
    Memory Corruption via TOCTOU Race in SharedArrayBuffer UTF-8 Decode (`StringBytes::Encode`)
    Node.js disclosed a bug submitted by v1ct0rv0nd00m: https://hackerone.com/reports/3752489
  • Open

    NULL pointer dereference in node:sqlite DatabaseSync#applyChangeset() via malformed SQLite changeset
    Node.js disclosed a bug submitted by junius: https://hackerone.com/reports/3736889
    Memory Corruption via TOCTOU Race in SharedArrayBuffer UTF-8 Decode (`StringBytes::Encode`)
    Node.js disclosed a bug submitted by v1ct0rv0nd00m: https://hackerone.com/reports/3752489

  • Open

    2026-05-22: SmartApeSG ClickFix --> Unidentified RAT --> NetSupport RAT
    No content preview
  • Open

    Interfacing with a 4-20mA pressure sensor
    This video tutorial is very good to show how to interact with an analog pressure sensor:  ( 11 min )
  • Open

    RemotePE: The Lazarus RAT that lives in memory
    Authors: Yun Zheng Hu and Mick Koomen Summary Last year, we published research about a North Korean Lazarus subgroup targeting financial and cryptocurrency organizations, encountered during multiple incident response engagements. This Lazarus subgroup overlaps with activity linked to AppleJeus, Citrine Sleet, UNC4736, and Gleaming Pisces. In one investigation, we observed that the actor had replaced … Continue reading RemotePE: The Lazarus RAT that lives in memory →  ( 26 min )
  • Open

    A Decade of BitLocker Vulnerabilities: What’s Patched, What’s Not, and What Still Works
    A few days ago we wrote about YellowKey, the newest entry in what has become a remarkably long list of BitLocker bypasses. That article walked through one specific attack with a practical workflow. This follow-up steps back and surveys the broader landscape: where BitLocker has been broken before, where it is still broken today, and […]  ( 23 min )
  • Open

    WordPress Site Down? Here’s How to Get Back Online
    If your WordPress site goes offline, every minute costs you lost sales, missed leads, and a dent in visitor trust. Search engines may start flagging errors, and customers see a blank page instead of your business. In that moment, the pressure is real: What broke, and how do you get back online before the damage adds up? The good news is that most WordPress outages are fixable. In most cases, your site isn’t lost, it’s blocked by something like a plugin conflict, server hiccup, database error, expired domain, SSL problem, sudden traffic spike, or malware infection. Continue reading WordPress Site Down? Here’s How to Get Back Online at Sucuri Blog.  ( 12 min )

  • Open

    Group restriction bypass via bearer token in user_oidc (SETTING_RESTRICT_LOGIN_TO_GROUPS not enforced in Backend::getCurrentUserId)
    Nextcloud disclosed a bug submitted by msatz: https://hackerone.com/reports/3572848
  • Open

    Group restriction bypass via bearer token in user_oidc (SETTING_RESTRICT_LOGIN_TO_GROUPS not enforced in Backend::getCurrentUserId)
    Nextcloud disclosed a bug submitted by msatz: https://hackerone.com/reports/3572848

  • Open

    curl --skip-existing has a TOCTOU race that lets a post-check symlink redirect the later download write
    curl disclosed a bug submitted by sdjasj: https://hackerone.com/reports/3747959
    Credentials forwarded to HTTP after HTTPSHTTP same-port redirect url_set_data_creds uses scheme-blind comparator
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3733946
    POST /api/bitcoinWithdrawalFees returns financial data without authentication despite being documented as a USER OPERATION (private endpoint)
    CoinMate.io disclosed a bug submitted by glferreira-devsecops: https://hackerone.com/reports/3676308
    HMAC signature verification omits endpoint and payload allowing request forgery on CoinMate API
    CoinMate.io disclosed a bug submitted by glferreira-devsecops: https://hackerone.com/reports/3670955
  • Open

    curl --skip-existing has a TOCTOU race that lets a post-check symlink redirect the later download write
    curl disclosed a bug submitted by sdjasj: https://hackerone.com/reports/3747959
    Credentials forwarded to HTTP after HTTPSHTTP same-port redirect url_set_data_creds uses scheme-blind comparator
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3733946
    POST /api/bitcoinWithdrawalFees returns financial data without authentication despite being documented as a USER OPERATION (private endpoint)
    CoinMate.io disclosed a bug submitted by glferreira-devsecops: https://hackerone.com/reports/3676308
    HMAC signature verification omits endpoint and payload allowing request forgery on CoinMate API
    CoinMate.io disclosed a bug submitted by glferreira-devsecops: https://hackerone.com/reports/3670955

  • Open

    HTTP/3 paused transfer buffers incoming data without bound up to ~1 GiB
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734947
    Schannel custom-CA path skips Extended Key Usage enforcement
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734992
    Connection reuse ignores haproxyprotocol and HAPROXY_CLIENT_IP settings, allowing PROXY context to persist across transfers
    curl disclosed a bug submitted by 7omoo: https://hackerone.com/reports/3741135
    SSL session-cache peer key omits signature_algorithms: strict-sigalg handle silently resumes a permissive sibling's session
    curl disclosed a bug submitted by hexproof: https://hackerone.com/reports/3739561
    CURLOPT_PROXY_CAINFO_BLOB silently activates native CA store on Apple builds
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735179
    TLS peer-verification bypass via mid-transfer ssl_config mutation
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735276
    TLS verifyhost bypass in rustls, mbedTLS, and wolfSSL when verifypeer=0
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734095
    HTTP/2 proxy CONNECT tunnel unbounded 1xx chain (missing Curl_bump_headersize cap in cf-h2-proxy.c)
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734020
    Cross-repository IDOR in `/settings/security_analysis/bypass_reviewers` allows unauthorized delegated bypass reviewer modification
    GitHub disclosed a bug submitted by ahacker1: https://hackerone.com/reports/3560256
  • Open

    HTTP/3 paused transfer buffers incoming data without bound up to ~1 GiB
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734947
    Schannel custom-CA path skips Extended Key Usage enforcement
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734992
    Connection reuse ignores haproxyprotocol and HAPROXY_CLIENT_IP settings, allowing PROXY context to persist across transfers
    curl disclosed a bug submitted by 7omoo: https://hackerone.com/reports/3741135
    SSL session-cache peer key omits signature_algorithms: strict-sigalg handle silently resumes a permissive sibling's session
    curl disclosed a bug submitted by hexproof: https://hackerone.com/reports/3739561
    CURLOPT_PROXY_CAINFO_BLOB silently activates native CA store on Apple builds
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735179
    TLS peer-verification bypass via mid-transfer ssl_config mutation
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735276
    TLS verifyhost bypass in rustls, mbedTLS, and wolfSSL when verifypeer=0
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734095
    HTTP/2 proxy CONNECT tunnel unbounded 1xx chain (missing Curl_bump_headersize cap in cf-h2-proxy.c)
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734020
    Cross-repository IDOR in `/settings/security_analysis/bypass_reviewers` allows unauthorized delegated bypass reviewer modification
    GitHub disclosed a bug submitted by ahacker1: https://hackerone.com/reports/3560256

  • Open

    What to Do When a Third-Party Data Breach Puts Your Website at Risk
    Data breach notification letters have become a familiar routine. They usually start with “We value your privacy” and offer a year of free credit monitoring. But the most important part is often hidden in the middle: A list of what actually got out. A leaked email address is not a leaked admin password. A hashed credential is not a session token. There is no universal post-breach checklist. The right response depends on the data exposed, so read the notice carefully and match your response to the level of exposure. Continue reading What to Do When a Third-Party Data Breach Puts Your Website at Risk at Sucuri Blog.  ( 8 min )
  • Open

    CURLOPT_HSTS_CTRL disables shared HSTS without share guard use-after-free and double-free
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3733934
    cookie: case-insensitive path comparison in replace_existing() allows cookie eviction across distinct paths
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735238
    libssh SFTP initialization ignores CURLOPT_TIMEOUT, hangs indefinitely
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735080
    rustls backend silently ignores CURLOPT_CRLFILE when native CA store is active
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734935
    HSTS multi-trailing-dot bypass-ish: possible incomplete fix for CVE-2022-30115
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3733984
    Unauthenticated File Upload to CDN
    Enjin disclosed a bug submitted by ph0r3nsic: https://hackerone.com/reports/3589247
    IDOR: autotranslate.translateMessage Full Message Content Leak
    Rocket.Chat disclosed a bug submitted by josan_george: https://hackerone.com/reports/3713682
  • Open

    CURLOPT_HSTS_CTRL disables shared HSTS without share guard use-after-free and double-free
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3733934
    cookie: case-insensitive path comparison in replace_existing() allows cookie eviction across distinct paths
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735238
    libssh SFTP initialization ignores CURLOPT_TIMEOUT, hangs indefinitely
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3735080
    rustls backend silently ignores CURLOPT_CRLFILE when native CA store is active
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734935
    HSTS multi-trailing-dot bypass-ish: possible incomplete fix for CVE-2022-30115
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3733984
    Unauthenticated File Upload to CDN
    Enjin disclosed a bug submitted by ph0r3nsic: https://hackerone.com/reports/3589247
    IDOR: autotranslate.translateMessage Full Message Content Leak
    Rocket.Chat disclosed a bug submitted by josan_george: https://hackerone.com/reports/3713682
  • Open

    YellowKey: An Unexpected Backdoor into BitLocker, and Why You Should Be Paying Attention
    On May 12, 2026, a researcher operating under the handles Chaotic Eclipse and Nightmare-Eclipse dropped a working proof-of-concept on GitHub for a Windows zero-day called YellowKey. In short, it lets anyone with brief physical access to a BitLocker-protected Windows 11, Windows Server 2022, or Windows Server 2025 machine pop a command prompt with full read […]  ( 17 min )

  • Open

    Trailing-dot IPv4 URL bypasses IP-address guard, allows wildcard DNS SAN match
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734921
    NULL pointer dereference in libcurl URL API redirect_url() with CURLU_DEFAULT_SCHEME
    curl disclosed a bug submitted by mulan_dh: https://hackerone.com/reports/3736234
  • Open

    Trailing-dot IPv4 URL bypasses IP-address guard, allows wildcard DNS SAN match
    curl disclosed a bug submitted by giant_anteater: https://hackerone.com/reports/3734921
    NULL pointer dereference in libcurl URL API redirect_url() with CURLU_DEFAULT_SCHEME
    curl disclosed a bug submitted by mulan_dh: https://hackerone.com/reports/3736234

  • Open

    SQL Injection in Column Type Parameter Allows Arbitrary SQL Execution
    Nextcloud disclosed a bug submitted by suul: https://hackerone.com/reports/3462991
  • Open

    SQL Injection in Column Type Parameter Allows Arbitrary SQL Execution
    Nextcloud disclosed a bug submitted by suul: https://hackerone.com/reports/3462991

  • Open

    Chrome is Silently Installing 4GB AI Model on Your Device without Consent. Here’s how to find it and remove it.
    Sometime we cover things on the Packet Protector Podcast News Roundup episodes that need immediate attention. This is one. Chrome has been sneaking a local AI model on systems, possibly since 2024. How to Disable and Remove It Quick option.. See If It’s Installed Now On Mac On Windows Enterprise Centralized Removal Options More info […]  ( 8 min )
  • Open

    Origin IP Exposed waf bypass
    Yuga Labs disclosed a bug submitted by r00tsid: https://hackerone.com/reports/1821085 - Bounty: $250
    Kerberos/SPNEGO Connection Reuse Vulnerability
    curl disclosed a bug submitted by rootofpi_ramesh: https://hackerone.com/reports/3725659
  • Open

    Origin IP Exposed waf bypass
    Yuga Labs disclosed a bug submitted by r00tsid: https://hackerone.com/reports/1821085 - Bounty: $250
    Kerberos/SPNEGO Connection Reuse Vulnerability
    curl disclosed a bug submitted by rootofpi_ramesh: https://hackerone.com/reports/3725659
  • Open

    Top 10 CERT-In Empanelled VAPT Companies in India (2026)
    Picking a VAPT vendor in India is harder than it should be. Every empanelled firm offers the same checklist of services on their homepage. Most reports follow the same template. And once you sign, the gap between a real attacker-mindset assessment and a glorified vulnerability scan only shows up after the work is done.  This list isn’t ranked by company size, […] The post Top 10 CERT-In Empanelled VAPT Companies in India (2026) appeared first on Payatu.  ( 68 min )

  • Open

    QuickSight Authorization Bypass: Chat Agents Accessible Despite Custom Permissions Denial
    AWS VDP disclosed a bug submitted by jcow: https://hackerone.com/reports/3577145
  • Open

    QuickSight Authorization Bypass: Chat Agents Accessible Despite Custom Permissions Denial
    AWS VDP disclosed a bug submitted by jcow: https://hackerone.com/reports/3577145

  • Open

    2026-05-11: Google ad for Claude leads to macOS malware infection
    No content preview  ( 1 min )
  • Open

    Using the Extraction Agent in 2026: Compatibility, Signing, Firewall, and Extraction Tips
    Over the years, we have published several articles about the extraction agent. However, the underlying technology changes quickly, and incremental changes often have significant cumulative effects. As a result, many of our older posts are no longer relevant and can be misleading if followed to the letter today. While last year’s recap, Installing and Troubleshooting […]  ( 13 min )

  • Open

    another liberapay member team twitter account broken Link Hijacking via Expired Twitter Account Link
    Liberapay disclosed a bug submitted by rox-11: https://hackerone.com/reports/3723002
    Liberapay member team twitter account broken Link Hijacking via Expired Twitter Account Link
    Liberapay disclosed a bug submitted by rox-11: https://hackerone.com/reports/3721519
  • Open

    another liberapay member team twitter account broken Link Hijacking via Expired Twitter Account Link
    Liberapay disclosed a bug submitted by rox-11: https://hackerone.com/reports/3723002
    Liberapay member team twitter account broken Link Hijacking via Expired Twitter Account Link
    Liberapay disclosed a bug submitted by rox-11: https://hackerone.com/reports/3721519

  • Open

    2026-05-08: macOS Shub Stealer infection
    No content preview
  • Open

    Private circle can be added to another circle via API despite visibility restriction
    Nextcloud disclosed a bug submitted by vidang04: https://hackerone.com/reports/3511998 - Bounty: $150
    Files drop share links for end-to-end encrypted folders allowed to drop files into other folders of the share owner
    Nextcloud disclosed a bug submitted by 0x0doteth: https://hackerone.com/reports/3304830
    View-only guests could see deleted Collectives pages in the trashbin
    Nextcloud disclosed a bug submitted by yoyomiski: https://hackerone.com/reports/3521434
  • Open

    Private circle can be added to another circle via API despite visibility restriction
    Nextcloud disclosed a bug submitted by vidang04: https://hackerone.com/reports/3511998 - Bounty: $150
    Files drop share links for end-to-end encrypted folders allowed to drop files into other folders of the share owner
    Nextcloud disclosed a bug submitted by 0x0doteth: https://hackerone.com/reports/3304830
    View-only guests could see deleted Collectives pages in the trashbin
    Nextcloud disclosed a bug submitted by yoyomiski: https://hackerone.com/reports/3521434
  • Open

    The Chrome Extension That Stole the CEO’s Cookies: A Confession on AI, Trust, and Supply Chain Security
    Modern software development doesn’t really start from a blank file anymore. Instead, most of the time, it’s about putting things together. We build by connecting pieces that already exist. For example, if we need a backend quickly, something like FastAPI gets us there in minutes. If we need a quick parsing script, an LLM generates […] The post The Chrome Extension That Stole the CEO’s Cookies: A Confession on AI, Trust, and Supply Chain Security appeared first on Payatu.  ( 70 min )

  • Open

    mbedTLS private-key blob null-termination asymmetry in lib/vtls/mbedtls.c (mbed_load_privkey)
    curl disclosed a bug submitted by shecantcode2: https://hackerone.com/reports/3717365
    ActiveStorage Disk Service Path Traversal via Custom Blob Key Injection
    Ruby on Rails disclosed a bug submitted by ksw9722: https://hackerone.com/reports/3580511
  • Open

    mbedTLS private-key blob null-termination asymmetry in lib/vtls/mbedtls.c (mbed_load_privkey)
    curl disclosed a bug submitted by shecantcode2: https://hackerone.com/reports/3717365
    ActiveStorage Disk Service Path Traversal via Custom Blob Key Injection
    Ruby on Rails disclosed a bug submitted by ksw9722: https://hackerone.com/reports/3580511

  • Open

    Critical Deadlock Vulnerability in Monero RPC Leading to Complete Node Paralysis
    Monero disclosed a bug submitted by rorkh: https://hackerone.com/reports/3307874
    Connection Count Bug in Monero Node Enables Outbound Peer Reset Attack
    Monero disclosed a bug submitted by yulge: https://hackerone.com/reports/3185083
    wcurl treats some URL operands after -- as curl options
    curl disclosed a bug submitted by p4p3r_hak: https://hackerone.com/reports/3708482
  • Open

    Critical Deadlock Vulnerability in Monero RPC Leading to Complete Node Paralysis
    Monero disclosed a bug submitted by rorkh: https://hackerone.com/reports/3307874
    Connection Count Bug in Monero Node Enables Outbound Peer Reset Attack
    Monero disclosed a bug submitted by yulge: https://hackerone.com/reports/3185083
    wcurl treats some URL operands after -- as curl options
    curl disclosed a bug submitted by p4p3r_hak: https://hackerone.com/reports/3708482
  • Open

    SSAQ AI for security 20260422
    本文链接 本报告《AI重塑网络安全:网络安全智能化产品与市场报告》由数说安全研究院于2026年4月发布,聚焦"AI for Security"领域,系统研究人工智能技术如何赋能网络安全防御、检测、响应与治理。报告明确区分"用AI保护系统"与"保护AI系统自身"的边界,围绕五大核心问题展开:AI如何重塑安全运营工作流、从Copilot到Agent的技术演进路径、产学研最新进展与差距、不同场景下AI能力边界、代表性厂商技术路线差异。 报告基于66家国内厂商问卷、11场深度访谈及国际公开资料,提炼13项关键发现:全球市场规模年均增速19-24%、头部厂商加速分化、数据质量成最大阻碍、工程化壁垒崛起等。市场按威胁检测、安全运营、数据安全等10大场景簇切分,并按Chatbot/Copilot、Agent智能体、内嵌增强三类交付形态梳理能力演进。 报告提供10大维度选型评估框架、30道RFI关键问题清单,并深度画像44家国内外代表厂商,为CISO、SecOps团队、产品负责人及投研机构提供可落地的战略参考与实践路径。  ( 3 min )
    Careful adoption of agentic AI services
    本文链接 本文档由澳大利亚、美国、加拿大、新西兰及英国等多国国家级网络安全机构联合发布,旨在为政府、关键基础设施及行业企业提供智能体AI(Agentic AI)安全采用的权威指南。智能体AI基于大语言模型,具备自主推理、规划与执行能力,但其高度的自主性与复杂的系统架构显著扩大了攻击面,引入了权限滥用、目标错位、行为不可预测、结构耦合及问责困难等新型安全风险。 指南系统梳理了智能体AI在设计、开发、部署与运营全生命周期中的安全挑战,并提出分层防御最佳实践。核心建议包括:严格遵循最小权限与零信任原则,强化细粒度身份管理;实施纵深防御与环境隔离;完善输入验证、第三方组件审查与持续监控机制;在高风险流程中强制保留“人在回路”监督;并通过威胁建模、红队演练与动态评估提升系统韧性。文档强调,组织应将AI安全深度融入现有网络安全框架,仅将其用于低风险任务,采取渐进式部署策略,优先保障系统的可观测性、可逆性与风险可控性,在安全可控的前提下推进技术落地。  ( 3 min )

  • Open

    Out of scope: Improper Input Validation Order on /api-internal/login via password field leads to unnecessary resource consumption
    PortSwigger Web Security disclosed a bug submitted by bereza4321: https://hackerone.com/reports/3625600 - Bounty: $200
    Potential Resource Leak in tool_parsecfg.c at line 279 during fileerror
    curl disclosed a bug submitted by ravindrasl2026: https://hackerone.com/reports/3710209
    libcurl 8.20.0 incomplete fix for CVE-2026-7168: changing only CURLOPT_PROXYPORT leaks stale Proxy Digest auth to a different proxy
    curl disclosed a bug submitted by codexxxx: https://hackerone.com/reports/3707747
    MQTT CONNACK Packet Type Bypass leads to RCE via Malicious Broker
    curl disclosed a bug submitted by orelbn7: https://hackerone.com/reports/3712343
  • Open

    Out of scope: Improper Input Validation Order on /api-internal/login via password field leads to unnecessary resource consumption
    PortSwigger Web Security disclosed a bug submitted by bereza4321: https://hackerone.com/reports/3625600 - Bounty: $200
    Potential Resource Leak in tool_parsecfg.c at line 279 during fileerror
    curl disclosed a bug submitted by ravindrasl2026: https://hackerone.com/reports/3710209
    libcurl 8.20.0 incomplete fix for CVE-2026-7168: changing only CURLOPT_PROXYPORT leaks stale Proxy Digest auth to a different proxy
    curl disclosed a bug submitted by codexxxx: https://hackerone.com/reports/3707747
    MQTT CONNACK Packet Type Bypass leads to RCE via Malicious Broker
    curl disclosed a bug submitted by orelbn7: https://hackerone.com/reports/3712343
  • Open

    DNSSEC: The Extra Security Layer That Can Break Your Padlock
    Turning on DNSSEC makes your domain more secure — but if it’s misconfigured, newer certificate validation rules can stop SSL renewals in their tracks. Hey there, You know that satisfying click when you finally turn on DNSSEC? It feels like adding a shiny new deadbolt to your domain’s front door. You’re doing the responsible thing: locking down your DNS against spoofing and hijacks, and making the internet just a bit safer. Continue reading DNSSEC: The Extra Security Layer That Can Break Your Padlock at Sucuri Blog.  ( 7 min )
  • Open

    [译] Anthropic 的产品团队为什么能比其他公司更快(2026)
    译者序 本文整理翻译自 2026 年的一档播客 How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code), 嘉宾是 Claude Code 的产品主管 Cat Wu。 文中多次提到”产品品味”,这一点可以 callback 关于 AI 下半场的思考(二):商业/应用篇(2025): AI 使得执行力不再稀缺,那以后工作的关键是什么 你要做什么(主观能动性,Agency) 你选择什么(品味,Taste) 水平及维护精力所限,译文不免存在错误或过时之处,如有疑问,请查阅原视频。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 以下是译文。 译者序 1 Anthropic 的 PM 角色是什么样的? 1.1 老板定 3~6 个月愿景,PM 拆成可执行计划 1.2 方向和思路一致,分工存在一定模糊区间 2 Anthropic 的 PM 岗位需要什么思维? 2.1 思维一:快速行动(Moving fast) 2.2 思维二:建立一个快速上线新功能的机制 2.3 思维三:建立一个高效的上下游团队协作框架 3 Anthropic 的 PM 还写 PRD 吗? 3.1 比 PRD 更重要的两件事 指标驱动,每周通晒 统一认知,符合团队的原则就可以自主决策,不受 PM 卡点 3.2 有时候也写 PRD:模糊功能、超大基建功能 4 Anthropic 为什么能迭代这么快? 4.1 确实有 mythos 的原因 4.2 更重要的原因:上线流程简单,鼓励每个人都能”从想法到上线” 5 Anthropic 的 PM team 组织形式是怎样的? 5.1 research PM team 5.2 clau…  ( 7 min )

  • Open

    Improper input validation On Exported deep-link handler crashes `FileDisplayActivity` on crafted external URL Denial-of-Service
    Nextcloud disclosed a bug submitted by khoof: https://hackerone.com/reports/3399016
    Double fdrop on a socket through sys_netcontrol
    PlayStation disclosed a bug submitted by slidybat: https://hackerone.com/reports/3320669 - Bounty: $10000
  • Open

    Improper input validation On Exported deep-link handler crashes `FileDisplayActivity` on crafted external URL Denial-of-Service
    Nextcloud disclosed a bug submitted by khoof: https://hackerone.com/reports/3399016
    Double fdrop on a socket through sys_netcontrol
    PlayStation disclosed a bug submitted by slidybat: https://hackerone.com/reports/3320669 - Bounty: $10000
  • Open

    Vulnerability & Patch Roundup — April 2026
    Vulnerability reports and responsible disclosures are essential for website security awareness and education. Automated attacks targeting known software vulnerabilities are one of the leading causes of website compromises. To help educate website owners about potential threats to their environments, we’ve compiled a list of important security updates and vulnerability patches for the WordPress ecosystem this past month. The vulnerabilities listed below are virtually patched by the Sucuri Firewall and existing clients are protected. Continue reading Vulnerability & Patch Roundup — April 2026 at Sucuri Blog.  ( 21 min )

  • Open

    Elcomsoft Phone Breaker 11 Restores iCloud Access
    Extracting cloud data becomes increasingly valuable – and increasingly complex at the same time. In scenarios where a target device is physically unavailable cloud extraction is often the only real way to access evidence. This is particularly relevant when devices are secured by an unknown passcode or locked under Apple’s Stolen Device Protection framework without […]  ( 6 min )

  • Open

    MQTT state machine confusion: PINGRESP/DISCONNECT with non-zero remaining_length dispatches to stale nextstate
    curl disclosed a bug submitted by fxv_ray_st: https://hackerone.com/reports/3702718
    Use-After-Free in SMB connection reuse (req->path dangling pointer after needle destruction)
    curl disclosed a bug submitted by nadsec42: https://hackerone.com/reports/3591956
    Negotiate connection reuse with wrong credentials when using CURLAUTH_ANY
    curl disclosed a bug submitted by anonymous_237: https://hackerone.com/reports/3646072
    Negotiate Authentication Premature on Connection Reuse
    curl disclosed a bug submitted by sdainard: https://hackerone.com/reports/3666576
    CVE-2026-7168: cross-proxy Digest auth state leak
    curl disclosed a bug submitted by xkilua: https://hackerone.com/reports/3697719
    CVE-2026-7009: OCSP stapling bypass with Apple SecTrust
    curl disclosed a bug submitted by 3lcarry: https://hackerone.com/reports/3694390
    CVE-2026-6253: proxy credentials leak over redirect-to proxy
    curl disclosed a bug submitted by joesephdiver: https://hackerone.com/reports/3669637
    CVE-2026-5545: wrong reuse of HTTP Negotiate connection
    curl disclosed a bug submitted by quaccws: https://hackerone.com/reports/3642555
    CVE-2026-6276: stale custom cookie host causes cookie leak
    curl disclosed a bug submitted by arkss: https://hackerone.com/reports/3671818
    CVE-2026-6429: netrc credential leak with reused proxy connection
    curl disclosed a bug submitted by nobcoderr: https://hackerone.com/reports/3677759
    CVE-2026-4873: connection reuse ignores TLS requirement
    curl disclosed a bug submitted by bonaire: https://hackerone.com/reports/3621851
    CVE-2026-5773: wrong reuse of SMB connection
    curl disclosed a bug submitted by osama-hamad: https://hackerone.com/reports/3650689
    Use-after-free in `curl_easy_ssls_export()` during callback re-entrancy
    curl disclosed a bug submitted by m1llie: https://hackerone.com/reports/3682666
    Heap-buffer-overflow in `Curl_ssl_push_certinfo_len()` sole bounds check is `DEBUGASSERT`
    curl disclosed a bug submitted by h3zh3z: https://hackerone.com/reports/3684614
    Stack exhaustion in MIME multipart reading with deeply nested subparts
    curl disclosed a bug submitted by wi110w: https://hackerone.com/reports/3684603
    PS4 BD-J privilege escalation using nested JAR
    PlayStation disclosed a bug submitted by gezine: https://hackerone.com/reports/3452696 - Bounty: $2500
  • Open

    MQTT state machine confusion: PINGRESP/DISCONNECT with non-zero remaining_length dispatches to stale nextstate
    curl disclosed a bug submitted by fxv_ray_st: https://hackerone.com/reports/3702718
    Use-After-Free in SMB connection reuse (req->path dangling pointer after needle destruction)
    curl disclosed a bug submitted by nadsec42: https://hackerone.com/reports/3591956
    Negotiate connection reuse with wrong credentials when using CURLAUTH_ANY
    curl disclosed a bug submitted by anonymous_237: https://hackerone.com/reports/3646072
    Negotiate Authentication Premature on Connection Reuse
    curl disclosed a bug submitted by sdainard: https://hackerone.com/reports/3666576
    CVE-2026-7168: cross-proxy Digest auth state leak
    curl disclosed a bug submitted by xkilua: https://hackerone.com/reports/3697719
    CVE-2026-7009: OCSP stapling bypass with Apple SecTrust
    curl disclosed a bug submitted by 3lcarry: https://hackerone.com/reports/3694390
    CVE-2026-6253: proxy credentials leak over redirect-to proxy
    curl disclosed a bug submitted by joesephdiver: https://hackerone.com/reports/3669637
    CVE-2026-5545: wrong reuse of HTTP Negotiate connection
    curl disclosed a bug submitted by quaccws: https://hackerone.com/reports/3642555
    CVE-2026-6276: stale custom cookie host causes cookie leak
    curl disclosed a bug submitted by arkss: https://hackerone.com/reports/3671818
    CVE-2026-6429: netrc credential leak with reused proxy connection
    curl disclosed a bug submitted by nobcoderr: https://hackerone.com/reports/3677759
    CVE-2026-4873: connection reuse ignores TLS requirement
    curl disclosed a bug submitted by bonaire: https://hackerone.com/reports/3621851
    CVE-2026-5773: wrong reuse of SMB connection
    curl disclosed a bug submitted by osama-hamad: https://hackerone.com/reports/3650689
    Use-after-free in `curl_easy_ssls_export()` during callback re-entrancy
    curl disclosed a bug submitted by m1llie: https://hackerone.com/reports/3682666
    Heap-buffer-overflow in `Curl_ssl_push_certinfo_len()` sole bounds check is `DEBUGASSERT`
    curl disclosed a bug submitted by h3zh3z: https://hackerone.com/reports/3684614
    Stack exhaustion in MIME multipart reading with deeply nested subparts
    curl disclosed a bug submitted by wi110w: https://hackerone.com/reports/3684603
    PS4 BD-J privilege escalation using nested JAR
    PlayStation disclosed a bug submitted by gezine: https://hackerone.com/reports/3452696 - Bounty: $2500
  • Open

    New Security Features and Low-Level Extraction of iOS 26
    We updated iOS Forensic Toolkit, adding low-level extraction support for iOS 26 and 26.0.1 via the extraction agent. This support is available for most iPhones and iPads compatible with the iOS 26 branch with a notable exception of the iPhone 17 range and M5-based iPads. Why exactly are these devices exempt, and what else did […]  ( 11 min )

  • Open

    What is online gambling spam and what can I do about it?
    Online gambling spam thrives on dreams of easy money and high stakes. Beating the house at an exotic casino. Splitting sevens. Going all in on the flop. A baccarat dealer calling La grande! For most people, though, the reality falls far short of Monte Carlo and an Aston Martin. So they turn to online gambling. And bad actors harness that allure to create their scams. They think they’re buying credits at a hot new online casino. Continue reading What is online gambling spam and what can I do about it? at Sucuri Blog.  ( 9 min )
  • Open

    Digital Triage Masterclass
    For decades, the forensic “gold standard” was straightforward: isolate the machine, pull the plug, and image the drive. In that era, what you saw on the screen was exactly what you would extract, bit by bit, from the magnetic platters. Today, that assumption is outdated, and is actively detrimental to an investigation. The digital forensics […]  ( 7 min )

  • Open

    CVE-2026-41651 PackageKit TOCTOU 本地提权漏洞分析
    <link rel="stylesheet" type="text/css" hr  ( 5 min )
  • Open

    IBM Aspera HTTP Gateway stores sensitive information in clear text in easily obtainable files which can be read by an unauthenticated user.
    IBM disclosed a bug submitted by jhon1231248e: https://hackerone.com/reports/3340797
    Bypass of Restricted Keyword "Mozilla" in Display Name Field via Unicode Homoglyphs on addons.allizom.org
    Mozilla disclosed a bug submitted by icecream_23: https://hackerone.com/reports/3279441 - Bounty: $500
    Bypassing Inbox Privacy Settings and Enabling Spam on Pixiv.net
    pixiv disclosed a bug submitted by aaqibhussain: https://hackerone.com/reports/3100570 - Bounty: $200
    Non-premium user can disable Ads in japanese version of dic.pixiv.net
    pixiv disclosed a bug submitted by lainkusanagi: https://hackerone.com/reports/3183520 - Bounty: $3000
  • Open

    IBM Aspera HTTP Gateway stores sensitive information in clear text in easily obtainable files which can be read by an unauthenticated user.
    IBM disclosed a bug submitted by jhon1231248e: https://hackerone.com/reports/3340797
    Bypass of Restricted Keyword "Mozilla" in Display Name Field via Unicode Homoglyphs on addons.allizom.org
    Mozilla disclosed a bug submitted by icecream_23: https://hackerone.com/reports/3279441 - Bounty: $500
    Bypassing Inbox Privacy Settings and Enabling Spam on Pixiv.net
    pixiv disclosed a bug submitted by aaqibhussain: https://hackerone.com/reports/3100570 - Bounty: $200
    Non-premium user can disable Ads in japanese version of dic.pixiv.net
    pixiv disclosed a bug submitted by lainkusanagi: https://hackerone.com/reports/3183520 - Bounty: $3000

  • Open

    我们扫描了五万个 Skill,发现危险仍然存在 | 朱雀实验室
    真正的问题不是扫出了什么,是:这个生态还有没有,有效的免疫系统? 腾讯安全平台部负责人 Coolc腾讯安全平台部负责人 Coolc  ( 13 min )
  • Open

    Argument Injection in /manage/ssh/ via host parameter leads to sensitive file disclosure on Weblate
    Weblate disclosed a bug submitted by alexb_616: https://hackerone.com/reports/3518571
  • Open

    Argument Injection in /manage/ssh/ via host parameter leads to sensitive file disclosure on Weblate
    Weblate disclosed a bug submitted by alexb_616: https://hackerone.com/reports/3518571

  • Open

    My Website Is Hosting a Phishing Page – Now What?
    Most phishing advice is written for the person staring at a suspicious email. This guide is for the other kind of victim: The website owner whose legitimate site has been quietly turned into the attacker’s weapon. You didn’t send the message or build the fake login page. You just woke up to a browser warning, a suspended hosting account, or a polite note from someone’s security team asking why your domain is requesting Apple ID credentials. Continue reading My Website Is Hosting a Phishing Page – Now What? at Sucuri Blog.  ( 10 min )

  • Open

    Low cost open-source sub-micrometer XYZ Manipulator
    This project is really cool and it is in my TODO list! https://github.com/0x23/MicroManipulatorStepper  ( 11 min )
  • Open

    mruby-engine: UAF in MRubyEngine#initialize enables local RCE
    Shopify disclosed a bug submitted by 0xd0ff9: https://hackerone.com/reports/3679660
  • Open

    mruby-engine: UAF in MRubyEngine#initialize enables local RCE
    Shopify disclosed a bug submitted by 0xd0ff9: https://hackerone.com/reports/3679660
  • Open

    How a Double-Encoded Null Byte Turns a ZIP File into an XSS Vector – CVE-2026-2790
    MIME type confusion, content sniffing abuse, and a sneaky bypass of a previous Firefox patch This bug is a bypass. It doesn’t introduce any new primitive on its own; it sidesteps a fix Mozilla shipped for CVE-2025-1936 without realising the sanitization was only half-done. To understand why it works, you need to dig into how […] The post How a Double-Encoded Null Byte Turns a ZIP File into an XSS Vector – CVE-2026-2790 appeared first on Payatu.  ( 67 min )

  • Open

    2026-04-23: SmartApeSG activity
    No content preview  ( 1 min )
  • Open

    WordPress DDoS Protection: How to Keep Your Site Online
    WordPress powers over 40% of the web, which makes it one of the most attractive targets for Distributed Denial of Service (DDoS) attacks. If your site goes down for an hour, you lose revenue, search rankings, and visitor trust. If it goes down repeatedly, you lose much more. A DDoS attack floods your website with fake traffic until it slows to a crawl or crashes entirely. Unlike hacks that steal data, DDoS attacks are about disruption. Continue reading WordPress DDoS Protection: How to Keep Your Site Online at Sucuri Blog.  ( 9 min )
  • Open

    Incomplete fix for CVE-2026-21637: loadSNI() in _tls_wrap.js lacks try/catch leading to Remote DoS
    Node.js disclosed a bug submitted by mbarbs: https://hackerone.com/reports/3556769
    RBAC bypass on App log endpoints via `permissionRequired` typo any authenticated user reads admin-only Enterprise App logs
    Rocket.Chat disclosed a bug submitted by arccode: https://hackerone.com/reports/3589551
  • Open

    Incomplete fix for CVE-2026-21637: loadSNI() in _tls_wrap.js lacks try/catch leading to Remote DoS
    Node.js disclosed a bug submitted by mbarbs: https://hackerone.com/reports/3556769
    RBAC bypass on App log endpoints via `permissionRequired` typo any authenticated user reads admin-only Enterprise App logs
    Rocket.Chat disclosed a bug submitted by arccode: https://hackerone.com/reports/3589551
  • Open

    Recovering Windows Credentials with Elcomsoft System Recovery
    In traditional forensic workflows, gaining access to a Windows system was a straightforward exercise: extract the NT hashes from a local database and run a fast (very fast!) offline attack. Today, Windows authentication is moving away from those essentially insecure NTLM hashes toward more resilient mechanisms. Microsoft is actively steering users away from local Windows […]  ( 8 min )

  • Open

    Complete authentication bypass to admin permissions
    Rocket.Chat disclosed a bug submitted by npc: https://hackerone.com/reports/3564655
  • Open

    Complete authentication bypass to admin permissions
    Rocket.Chat disclosed a bug submitted by npc: https://hackerone.com/reports/3564655

  • Open

    Low-Level Extraction for M-Series iPads
    With the release of iOS Forensic Toolkit 10.01, we are extending low-level extraction capabilities to Apple tablets running up to iPadOS 18.7.1. This update brings our extraction agent to the latest hardware, supporting not just A-series but also M-series iPads. We have also implemented support for the distinct memory layout found in high-end 1TB and […]  ( 7 min )

  • Open

    Joomla SEO Spam Injector: Obfuscated PHP Backdoor Hijacking Site Visitors
    Overview During a recent malware cleanup investigation, we encountered a compromised Joomla website where the site owner reported a strange issue. Their website displayed a large number of suspicious product links that had nothing to do with their business. These products were not added by the website owner and did not exist in their catalog. Visitors and search engines were seeing pages that promoted unrelated products, raising immediate concerns about spam injection or remote content manipulation. Continue reading Joomla SEO Spam Injector: Obfuscated PHP Backdoor Hijacking Site Visitors at Sucuri Blog.  ( 8 min )
  • Open

    IoT Penetration Testing: A Complete Methodology Guide with the OWASP ISTG Framework
    In our last 20 IoT security assessments across automotive, MedTech, and consumer electronics, 17 devices had at least one debug interface (UART or JTAG) left fully accessible in production. In 9 of those cases, the exposed interface gave us an unauthenticated root shell within minutes of opening the enclosure. That single finding class alone invalidated every software-level security […] The post IoT Penetration Testing: A Complete Methodology Guide with the OWASP ISTG Framework  appeared first on Payatu.  ( 72 min )

  • Open

    Web Cache Poisoning vs Deception: The Dynamic Duo of Cache Attacks
    Web applications today rely heavily on caching to improve performance and reduce server load. CDNs like Cloudflare, Akamai, and reverse proxies like Nginx store frequently requested content so users get faster responses. But when these caching systems are misconfigured or behave unexpectedly, they can become serious security vulnerabilities. This post covers two related but different […] The post Web Cache Poisoning vs Deception: The Dynamic Duo of Cache Attacks appeared first on Payatu.  ( 73 min )

  • Open

    Interfacing with 4-20mA interface
    I found this suggested circuit: At this thread: https://electronics.stackexchange.com/questions/166301/using-a-microcontroller-adc-to-measure-a-4-20ma-sensor There are many commercial solutions to use 4-20mA with microcontrollers, i.e.:https://www.mikroe.com/4-20ma-r-click Or you can use INA260 to measure current directly like this guy did (note: INA260 is not ideal for low current):https://www.youtube.com/watch?v=GzxyzwyHRN0  ( 12 min )

  • Open

    从发现 33 个 0day 漏洞,看安全攻防的下半场 | 朱雀实验室
    2026 年 4 月 7 日,Anthropic 联合 Apple、Google、Microsoft 等 45 家机构发布 Project Glasswing 计划,并宣布其尚未公开发行的前沿模型 Claude Mythos Preview 在所有主要操作系统和浏览器中发现了数千个…  ( 12 min )

  • Open

    Why 2FA SMS is a Bad Idea in 2026
    What is 2FA? Two-factor authentication (2FA) offers a second layer of security to help protect an account from brute force, phishing, and social engineering attacks. 2FA requires an extra step for a user to prove their identity, which reduces the chance of a bad actor gaining access to their account or data. And since notifications are sent to verify the initial authentication via username and passwords, it also gives users and business the ability to monitor for potential indicators of a compromise. Continue reading Why 2FA SMS is a Bad Idea in 2026 at Sucuri Blog.  ( 11 min )

  • Open

    GDPR Meets DPDP: What Every Indian CISO Operating in the EU Must Know
    India finally has its data protection law. Europe has had one since 2018. If your organisation sits at the intersection of both, you are now subject to two comprehensive privacy regimes simultaneously, and the gaps between them are where the risk lives. A New Privacy Reality for Indian Businesses  For years, Indian technology companies, IT […] The post GDPR Meets DPDP: What Every Indian CISO Operating in the EU Must Know appeared first on Payatu.  ( 70 min )
    Top 7 Cybersecurity Companies in India – 2026 Edition
    The cybersecurity landscape in India has evolved dramatically over the past decade. What began as basic antivirus and firewall implementations has transformed into sophisticated security operations encompassing threat intelligence, advanced penetration testing, security orchestration, compliance management, and cutting-edge research. As cyber threats grow more sophisticated and regulations become stricter, organizations need security partners who can deliver both technical depth […] The post Top 7 Cybersecurity Companies in India – 2026 Edition appeared first on Payatu.  ( 66 min )

  • Open

    Vulnerability & Patch Roundup — March 2026
    Vulnerability reports and responsible disclosures are essential for website security awareness and education. Automated attacks targeting known software vulnerabilities are one of the leading causes of website compromises. To help educate website owners about potential threats to their environments, we’ve compiled a list of important security updates and vulnerability patches for the WordPress ecosystem this past month. The vulnerabilities listed below are virtually patched by the Sucuri Firewall and existing clients are protected. Continue reading Vulnerability & Patch Roundup — March 2026 at Sucuri Blog.  ( 29 min )

  • Open

    JetStream 3: A modern benchmark for high-performance, compute-intensive Web applications
    shared announcement blog post, we wanted to take a moment here to dive a little deeper. In this post, we’ll pull back the curtain on the benchmark itself, explore the methodology behind our choices, and share the motivations driving these major updates. Why Do We Benchmark, Anyway? Driven by Open Governance collaborative effort across all major browser engines, including Apple, Mozilla, and Google. We adopted a strict consensus model for this release. This means we only added new workloads when everyone agreed they were valuable and representative. This open governance model has led to an incredibly productive collaboration with buy-in from multiple parties, ensuring the benchmark serves the best interests of the overall Web ecosystem. Ripe for an Update JetStream vs. Other Benchmarks…  ( 16 min )
  • Open

    从4.8亿下载量的 LiteLLM投毒事件,看 AI 基础设施安全攻与防 | 朱雀实验室
    近日,知名大模型网关工具 LiteLLM 遭遇供应链投毒,其 1.82.7和1.82.8 版本被植入恶意代码。由于该项目月下载量极高(近1亿月下载量),且被 DSPy 等众多主流…  ( 12 min )
  • Open

    How to Fix “Not Secure” Warnings and SSL Issues in WordPress (8 Steps)
    If you own a WordPress website and ever encountered the “Not Secure” warning, you might have worried that visitors would perceive your site as spam or fraudulent. Not only does this warning impact user trust, but it can also create technical search issues when both HTTP and HTTPS versions of your pages remain accessible or when redirects, canonicals, and sitemaps point to different URL versions. Browsers show the visible security warning, while search engines rely on permanent redirects, canonical URLs, and updated sitemaps to understand your preferred HTTPS pages. Continue reading How to Fix “Not Secure” Warnings and SSL Issues in WordPress (8 Steps) at Sucuri Blog.  ( 11 min )

  • Open

    Android Sets New Record for Mobile Web Performance
    Android is now the fastest mobile platform for web browsing. Through deep vertical integration across hardware, the Android OS, and the Chrome engine, the latest flagship Android devices are setting new performance records, outperforming all other mobile competitors in the key web performance benchmarks Speedometer and LoadLine and providing a level of responsiveness previously unseen on mobile. web performance benchmarks (Chrome 146, March 2026) Why web performance matters responsiveness and page load speed. Speedometer: Measuring web responsiveness Speedometer is the collaborative industry standard used by all major browser engine developers to measure web app responsiveness. It simulates real-world user actions—like adding items to a to-do list—to measure interaction latency. Wh…  ( 14 min )
  • Open

    Slightly safer vibecoding by adopting old hacker habits
    I have seen a lot of public discussion around supply-chain attacks on the Python ecosystem, prompt injection risks when using coding agents, and general worries about the security implications of "vibe coding" for the development machine. In some of these discussions I find myself puzzled as to what problem is being solved - and it took me a while to realize that my failure to understand lies in the development setup that I tend to use. In this blog post I'll quickly explain my development setup. The setup is pretty simple: The actual development happens on a rented server (or a VM on that server). In order to do development, I SSH into that server with key-forwarding for my github keys enabled. I perform my development on the server by attaching to a screen or tmux session. I used to …  ( 10 min )

  • Open

    OpenClaw:技术解读和给 AI 应用开发的启示(2026)
    最近几个月 OpenClaw 大火,各种 Claw 大家多多少少都体验过了, 本文从技术角度尽量介绍一些不一样的东西,希望可以给大家一些参考和思考。 Code and scripts used in this post: Github. 水平及维护精力所限,文中不免存在错误或过时之处,请酌情参考。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 1 引言 1.1 技术演进:LLM -> Agent -> Assistant 1.2 OpenClaw 是什么 官方自我介绍 设计视角 技术视角 1.3 OpenClaw 不是什么? 2 ToyClaw:200 行 Python 代码实现一个 Claw 2.1 Code (by cursor with GPT-5.4) 2.2 启动和基本对话测试 2.3 Identity/Soul/User 测试 2.4 User 测试 2.5 Skill 测试 2.6 其他测试 查看 CPU 占用最高的几个进程 查看哪个进程在监听 22 端口 2.7 小结 3 OpenClaw 技术架构 4 OpenClaw 核心组件 4.1 Agent 系统提示词 Agent 类型: in-process vs. subprocess 4.2 Skill skill vs. tool 示例:tripgenie skill 4.3 定时任务 4.4 工作空间 (workspace) 5 当我们在养龙虾时,我们在养什么 5.1 AGENTS.md:主 Agent 系统提示词 5.2 BOOTSTRAP.md:启动提示词,用后即删 5.3 USER.md:Assistant 对 User 的理解和记录 5.4 IDENTITY.md:Assistant 的身份 5…  ( 26 min )

  • Open

    码力·智能开发与安全一体化平台
    本文链接 《码力·智能开发与安全一体化平台》由长亭科技推出,以"AI+安全"双引擎破解企业AI编码困局:IDE智能辅助实现代码生成-审计-修复闭环,AI员工全自动完成需求到提交;内置SAST/SCA/AI审计三层防护,误报率<5%,支持DeepSeek等主流模型私有化部署。独创第四代自主规划Agent架构,1400倍上下文压缩、环境驱动执行与动态任务分解,让80万行代码审计从"被动扫描"升级为"主动探索";配套安全/质量/合规/业务逻辑四维模板与Skill热插拔体系,已服务蔚来、贝壳等企业,AI代码占比超30%,实现"代码生而安全"的新型开发范式。  ( 3 min )

  • Open

    Quick way to find where a function is called in the source code using the ELF binary file
    Just use these two commands:  ( 11 min )

  • Open

    Recreation of the Windows 95 3D Maze Screensaver
    These are some implementations I found in the Internet: Javascript: https://github.com/ibid-11962/Windows-95-3D-Maze-Screensaver Rust: https://github.com/clrnd/win95-maze-rs (This is not a recreation) This guy added navigation control to the original 3D Maze screenshot: https://github.com/x86matthew/Playable3DMaze  ( 11 min )

  • Open

    Testing an OpenGL 3D Cube on Linux
    I found this nice video tutorial at Youtube video: https://www.youtube.com/watch?v=WjSRCX4OrZU Then I downloaded the example from: https://bitbucket.org/c-code/snakegame/src/master/opengl_cube.c but when I tried to compile I got this error: So, the solution was installing these packages: Then I noticed my other mistake, it should be -lGL and -lGLU, also -lglu was needed:  ( 11 min )

  • Open

    Bringing Chrome to ARM64 Linux Devices
    We’re excited to announce that Google will launch Chrome for ARM64 Linux devices in Q2 2026, following the successful expansion of Chrome to Arm-powered macOS devices in 2020 and Arm-powered Windows devices in 2024. Launching Chrome for ARM64 Linux devices allows more users to enjoy the seamless integration of Google’s most helpful services into their browser. This move addresses the growing demand for a browsing experience that combines the benefits of the open-source Chromium project with the Google ecosystem of apps and features. This release represents a significant undertaking to ensure that ARM64 Linux users receive the same secure, stable, and rich Chrome experience found on other platforms. Get the best of the Google ecosystem With Chrome, you are able to leverage the full…  ( 12 min )

  • Open

    Cheaper WiFi Repeater with MT7628 chip
    I bought a really low cost WiFi after watching the LowLevel channel video about hacking a similar device. But in this case the device was using a Qualcomm chip, mine came with a MT7628KN inside (board’s silkscreen: HX7628_L2_20210109). Searching in the Internet I found many datasheets and programming guides for this chip and also there … Continue reading Cheaper WiFi Repeater with MT7628 chip →  ( 11 min )

  • Open

    Flashing in the maxbox the Linux image generated by yocto
    These are the steps I use: Initialize the BOOTP/DHCPD service: In the maxbox run the default u-boot command to flash the device: That is it!  ( 11 min )

  • Open

    Preempted: From ADB Service Call to Bootloader Unlock on Xiaomi
    <link rel="stylesheet" type="text/css" hr  ( 4 min )

  • Open

    Top 5 IoT Security Companies in India 2026
    The Internet of Things has transformed how businesses operate, from smart manufacturing floors to connected healthcare devices and intelligent building systems. However, this connectivity comes with significant security risks. IoT devices often have limited processing power, run outdated firmware, lack basic security controls, and create new attack surfaces that traditional security tools weren’t designed to protect.  IoT security […] The post Top 5 IoT Security Companies in India 2026  appeared first on Payatu.  ( 64 min )

  • Open

    Top 7 Red Team Companies in India
    In today’s threat landscape, traditional penetration testing often isn’t enough. Organisations need to understand how they’d fare against sophisticated, persistent attackers who use social engineering, physical breaches, and advanced techniques to achieve specific objectives. That’s where red teaming comes in.  Red teaming goes beyond finding vulnerabilities. It simulates real-world adversaries attempting to breach your defences, evade detection, and accomplish concrete goals like exfiltrating sensitive data or […] The post Top 7 Red Team Companies in India  appeared first on Payatu.  ( 64 min )

  • Open

    Getting started to WiFi HaLow
    I bought some seeed studio wifi modules and I will get it running: https://wiki.seeedstudio.com/getting_started_with_wifi_halow_module_for_xiao/ Then I found a user that was able to use it with the official MorseMicro image: https://community.morsemicro.com/t/after-changing-bcf-from-basic-to-quectels-for-xiao-esp32s3-sense-xiao-wifi-halow-quetel-fgh100m-h/1121 Also the pre-build binaries are here: https://github.com/MorseMicro/morse-firmware/tree/main/bcf/quectel Connecting the Seeed Studio Module to Raspberry Pi Zero 2W discussed here: https://community.morsemicro.com/t/raspberry-pi-zero-2w-with-fgh100m/938/6  ( 11 min )

  • Open

    Pickling the Mailbox: A Deep Dive into CVE-2025-20393
    <link rel="stylesheet" type="text/css" hr  ( 5 min )

  • Open

    如何给 Linux 内核提交补丁:一次真实的踩坑记录
    <link rel="stylesheet" type="text/css" hr  ( 4 min )

  • Open

    当AI学会背刺:深度剖析Agent Skills的安全陷阱 | 朱雀实验室
    你是否想过,当你让 AI 帮你写代码时,它可能正在你背后悄悄地干着其它事——比如,加密你的所有文件,然后弹出一个勒索提示?这不是吓唬你,而是我们研究后发现在各种…  ( 11 min )

  • Open

    SOAR落地实践分享
    本文链接 《SOAR落地实践》介绍平安科技5年来将“安全编排、自动化与响应”平台化的经验:以“自动炒菜机”理念把钓鱼研判、IP封堵、离职终端策略加固、资产富化与电话告警5大高频场景压缩至分钟级闭环,实现无人值守T+5分钟处置。分享给出SOAR选型关键——APP生态、剧本丰富度、总拥有成本与同行业案例,并剖析剧本设计、白名单校验、审批节点与重拨机制等落地细节,为准备建设或优化SOAR的团队提供可直接复用的参考模板。  ( 3 min )

  • Open

    [笔记] 生成式推荐:OpenOneRec 技术报告(快手,2026)
    本文是阅读学习快手的 OpenOneRec Tech Report 时整理的一些笔记, 很多训练相关的信息已经开源,见 github.com/Kuaishou-OneRec/OpenOneRec, 包括: 测评框架 RecIF-Bench 和训练数据:16w 用户,96million 交互数据 数据处理代码、训练代码,确保可复现文中内容(非 pro 版本) 训练好的模型:1.7B、8B 整体框架: 训练&评估任务: 相关文章: 从 Tokenization 视角看生成式推荐(GR)近几年的发展(2025) 水平及维护精力所限,文中不免存在错误或过时之处,请酌情参考。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 1 引言 1.1 RecIF-Bench:推荐领域的指令遵循 benchmark 1.2 缓解 SFT 带来的通用能力退化 1.3 开源模型:1.7B/8B 2 基础 2.1. Items as Tokens: 商品的语义编码 2.2. Recommendation as Auto-regressive Models:用自回归模型做推荐 3 RecIF-Bench: 推荐领域的指令遵循 Benchmark 3.1 数据集构建 数据集切分策略:按用户维度 80:20 切分 3.2 评估任务:4 层,从对齐到推理 3.2.1. Layer 0: 语义对齐能力 3.2.2. Layer 1: 基础推荐能力 3.2.3. Layer 2: 指令遵循能力 3.2.4. Layer 3: 推理能力(推荐理由) 3.3. 评估指标 推荐指标:Pass@K, Recall@K 文本生成指标:LLM-as-Judge 4 Pre-Training 4.1 Item Tokenization …  ( 5 min )

  • Open

    Five Years with Hugo
    During a break over Christmas 2020, I rebuilt this site, moving from WordPress to Hugo. After more than 5 years of publishing with Hugo, I’d like to share what I’ve learned, what’s worked, what hasn’t, and why for once, I’m happy with the platform I’m using. This review builds on two recent articles, Five Hundred, a retrospective of 500 posts to this site, and ‌Lessons Learned from 20 Years & Why You Should Blog, a look back at 20 years of publishing here, and the value of writing & blogging more generally. In this post, I will be diving into publishing with Hugo specifically, what’s good, what’s not, and what you should think about if you are considering it. Writing Friction In the Lessons Learned post I spoke about the need to carefully consider the friction of the tools you use, as any …  ( 6 min )

  • Open

    AI & IAM: Focus on Fundamentals
    A recent article, The Future of Cybersecurity Includes Non-Human Employees1, discussed the growing need to manage access granted to the rapidly expanding number of AI agents being deployed in companies. This is a deeply important topic, and particularly timely, as many are facing this challenge today. While I do want to address that topic, I also want to address how this is being framed. Aside from leaning into an inflammatory tone with the use of “Non-Human Employees” as part of the title of the piece, there’s a deeper issue I see with how this is being framed, and it’s also related to AI. Importantly, this is far from unique to this article, but a trend across discussions of AI (and other new & emerging technologies). For the pragmatic security practitioner, a clear understanding of this…  ( 7 min )

  • Open

    On Privacy Nihilism
    Amongst the steady stream of marketing emails for gift cards and other last minute gifts in the days before Christmas, buried in the noise sent when people are least likely to see it, was a notice. It was an all-too-familiar “we take your privacy seriously, but” email. Perfectly timed to make it clear that privacy wasn’t that important. This wasn’t just my email address being leaked, this was everything. Name, address, income, employer, social security number. Each record stolen was essentially an identity theft kit; everything needed in one place. From a privacy and data security perspective, few things are worse. Yet the only thing remarkable about my reaction to that notice was that it was entirely unremarkable, more blasé than nonplussed. This was far from the first such notice, not ev…  ( 5 min )
    Dynamic Social Media Images for Hugo
    I’m a big fan of Hugo as a publishing platform, it’s the framework behind this site, and is incredibly flexible - if you are willing to invest the time and effort to make it truly yours. It’s fast, versatile, and has robust theming support. However, it’s also a static site generator, so doing anything dynamic means doing some extra work (as you have to do it at build time). My life philosophy can be summed up to “work hard to be lazy” - in this case, that means I want a solution for social media sharing (OpenGraph) images that I will work without me needing to think about them again. This way, when a link is shared to Bluesky, Mastodon, LinkedIn, or other platforms, a reasonable image will be shown - even if I didn’t include an image for the article. More importantly, I was a solution that…  ( 5 min )

  • Open

    当 AI 成为代码贡献者,软件安全正在发生怎样的变化?
    随着 AI 技术加速融入软件研发流程,AI 已经成为新时代的“代码贡献者”。在显著提升研发效率的同时,AI 生成代码也对软件安全、漏洞治理和工程责任边界提出了新的挑战。  ( 12 min )

  • Open

    Lessons Learned from 20 Years & Why You Should Blog
    Twenty years ago, I started publishing articles and essays here, and I recently published the 500th post to this site. After writing 267,897 words here and investing 2,100 hours into this site, I’ve learned a few things, made some mistakes, and I’d like to share some of these insights with you. Whether you are a veteran of the blogosphere or questioning if you should dip your toes in the waters (you should), I think you will find some useful information here. This isn’t a how-to, this isn’t a collection of best practices, and this certainly isn’t a “how to get rich as a blogger” guide. This is what I’ve learned from making mistakes over the last 20+ years, and how you can avoid them, or maybe this will help you see that you’ve been making the same mistake. I hope that you will find this to…  ( 8 min )

  • Open

    Five Hundred
    The year was 2006 when I registered adamcaudill.com and set up WordPress to host this site. I had recently moved, started a new job as a software developer, and I wanted a new place to share thoughts, code, and the insight I was gathering along the way. I made the very first post. It will be 20 years, next month, since that first post, a short note from someone still finding his legs in the industry and far from finding his legs as a writer. Through the 2000s, the average length of the posts was only 240 words. Far from the long-windedness common in my more recent work. This post marks the 500th post here. During 2006, I posted frequently, several times each month. Most of these posts were short, filled with typos, and were generally short asides about news in technology. These early, and …  ( 4 min )

  • Open

    Whose Monkeys Are These?
    Over the course of my career, I’ve found that there are some principles that are key for people and teams to be effective. One of these is that everything should have an owner. Everything should have someone that is responsible. Everything should have a designated person whose job it is to care about it. This might a be bug or vulnerability reports in software, it could be routine processes, or who responds to certain emails. When there isn’t clear ownership, things slip through the cracks, things are forgotten about, and minor issues can quickly become anything but minor. This works well enough for many things where ownership is clear and agreed upon, but things get ugly when that thing is a problem. Here, human nature starts getting in the way, and leads to unfortunate outcomes. In this …  ( 6 min )
  • Open

    [译] 软件领域的工业革命:AI 将使软件成为一种新的 UGC(2025)
    译者序 本文翻译自 2025 年的一篇文章 The rise of industrial software。 工业化能以极大的规模生产低质量、低成本的产品, 印刷工艺的工业化导致了平装书的出现 农业的工业化导致了垃圾食品的出现 数字图像传感器的工业化导致了海量普通人拍摄的图片、视频等等 LLM 的出现是软件领域的蒸汽机时刻,软件开发正在经历一次属于它的“工业革命”, 软件开发正在从传统手工业变成制造业 一旦生产成本足够低,垃圾就是能最大化产量、利润和市场触达的东西 最终市场上流通的不是丰富的好东西,而是过量的最易消费的东西 —— 我们确实正在消费它们(AI 垃圾) 人类程序员未来还有多少市场?未来的创新将是什么? 水平及维护精力所限,译文不免存在错误或过时之处,如有疑问,请查阅原文。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 以下是译文。 译者序 1 软件开发的“工业革命”:从手工业到制造业 1.1 手工业:成本高、开发慢,高度依赖人的专业技能和经验 1.2 制造业:成本低、开发快、很少依赖人的专业知识 1.3 软件开发日益自动化的世界 2 软件作为一次性商品 2.1 现状:劳动力(程序员)贵,生产(软件开发)有规模瓶颈 2.2 工业化的本质:自动化(不依赖人、低成本) 影响一:传统开发模式受到挤压,门槛降低,劳动力(程序员)竞争加剧 影响二:大规模生产低质量、低成本的软件 2.3 一次性软件(disposable software) 3 稀缺商品的工业化生产 3.1 Jevons 悖论 煤炭:单位效率提升,单位成本下降,总消费上升 Token:单位推理成本下降,推理需求变多,总算力消费激增 3.2 农业领域的先例:食物生产的工业化:垃圾食品 消灭饥饿 vs. 垃圾食品 丰富的好东西 vs.…  ( 1 min )

  • Open

    Is Long-form Writing Dead?
    In a world where attention spans have been reduced to seconds, college students aren’t expected to read full books, AI is used to summarise anything more than a few sentences, and blogs have been largely replaced with microblogging platforms, is there still a place for long-form writing? In this essay I would like to explore that question; from how we got here to what hope we have for the future. This is a complex topic, and to properly explore this, we need to go back several decades. What we see today are the symptoms of decades of changes and issues, layered one upon the other, building up to a far larger issue than what any of these could explain in isolation. While this essay goes on quite a journey, I believe that the journey is worthwhile and enlightening. Before we truly begin, I m…  ( 8 min )

  • Open

    TP-Link WR841N router CVE-2023-50224 and CVE-2025-9377
    <link rel="stylesheet" type="text/css" hr  ( 3 min )
  • Open

    Why I Will Never Write With AI
    Every person has rules that they apply to themself, red lines that they won’t cross. For me, signing my name to anything generated by AI is one of those. I’d like to take a few minutes to talk about why. I spend a lot of time writing. Some is on this site, though much more of my work isn’t public, is published elsewhere, not yet published, or ghost written for others. On this site, according to the stats, I’ve spent at least 864 hours writing the content you see here. That’s over a month of continuous writing. That’s over a month of my life dedicated to sharing information and working to prompt further discussion and contemplation. Yet I never have, nor ever will, sign my name to anything written with AI1. For some readers, this will seem to be a given due to objections2 to generative AI. …  ( 9 min )

  • Open

    25% Unemployment in Tech?
    For more than a year there has been a clear trend: unemployment in the tech industry was climbing. The number of people with an #OpenToWork badge on their picture was climbing. There was something worse though, something that I hadn’t seen before: several months later, they were still looking. Something was wrong. Something had changed. There are incredibly talented people with years or even decades of experience that are finding themselves spending anywhere from 9 to 24+ months to find new employment. People that are well established in their careers, people that have proven themselves, yet unable to find stable employment. People that just a few years ago would have had several offers within weeks, now unable to find even one after many months. The strength and resilience of the job mark…  ( 8 min )

  • Open

    [译][论文] P5 paper | 用语言模型做推荐:一种统一的预训练、个性化提示和预测范式(2022)
    译者序 本文翻译自 2022 年 RecSys 大会的一篇论文 Recommendation as Language Processing (RLP): A Unified Pretrain, Personalized Prompt & Predict Paradigm (P5)。 图 3:P5 架构示意图。 水平及维护精力所限,译文不免存在错误或过时之处,如有疑问,请查阅原文。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 以下是译文。 译者序 摘要 1 引言 1.1 现阶段推荐系统的特点 特征表示和学习越来越复杂 推荐任务的类型越来越多样 1.2 现代推荐系统需要什么 1.3 P5 的创新点 2 相关工作 2.1 统一框架的尝试 基于通用语言模型(T5 和 GPT3) 基于自然语言的 seq-to-seq 架构 基于通用用户表示 2.2 通过提示的方式学习(Prompt Learning) 2.3 推荐领域的 NLP 2.4 Zero-shot 和冷启动推荐 3 个性化 prompts 集合 3.1 Prompts 设计 rating prediction 提示词设计 sequential recommendation 提示词设计 explanation 提示词设计 review 相关提示词设计 direct recommendation 提示词设计 3.2 从原始数据构建训练数据集(prompts & answers) 4 P5 范式与模型 4.1 P5 架构 位置编码 Whole-word embedding,补偿 item token 表示被 tokenizer 拆分带来的语义损失 encoder & decoder 4.2 用预训练的 P5 进行推荐任务(推…  ( 16 min )

  • Open

    以旅行规划(Trip Planning)为例,看 DeepSeek-V3.2 如何合成高质量训练数据(2025)
    如何基于 Agent/LLM 强大的规划能力+生成能力+代码执行能力+反思能力, 自动化合成大批量高质量数据: Hypothetical workflow DeepSeek-V3.2: workflow for synthesizing high-quality agentic datasets for RL training (in agentic fashion, without human intervention) 水平及维护精力所限,文中不免存在错误或过时之处,请酌情参考。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 1 场景:增强模型的 Trip Planning 能力 1.1 方案拆解 1.2 子任务:准备高质量的 Trip Planning 数据 2 方案:自动合成高质量 Trip Planning 数据 2.1 思考:人(专家)怎么完成这个任务 2.2 自动化:人工方案的 workflow 化 2.3 这个 workflow 的独特之处 2.4 小结 3 图解:DeepSeek-V3.2 是怎么做的(”Large-Scale Agentic Tasks”) 3.1 方案描述 3.2 方案图解 Step 0: Agent 输入 Step 1: Agent 构建旅行数据库 Step 2: Agent 合成 tools(代码生成) Step 3: 合成一个具体旅行规划任务 Step 4:执行 solution function,(基于 tool calling)生成一个线路规划 Step 5:执行 verification function,对上一步生成的线路规划进行验证 Step 6: 如果验证成功,将这条数据输出 Step 7: 返回到 step 3,继续合成下一个更难的任务 Step 8: 如果 step…  ( 6 min )

  • Open

    Ask your LLM for receipts: What I learned teaching Claude C++ crash triage
    I recently embarked on a small toy project/experiment: How well can I equip Claude Code to automatically analyze and triage crashes in a C++ code base? For the experimentation, I worked on a small number of crashes in the ffmpeg bug tracker. The initial results were very discouraging, Claude hallucinated all sorts of implausible root causes and tended to write typical "AI slop" -- things that follow the form of a well-written report, but that had no bearing on reality. I iterated for a few days, but ultimately I got things to work reasonably well, at least to the point where I was happy with the result. The result of this little diversion are a bunch of .md files (subagents and skills) that I contributed to https://github.com/gadievron/raptor - specifically the following parts: https://gi…  ( 11 min )

  • Open

    Good Faith, Moral Duty, and Selfishness
    What do we owe to each other? This is a classic question of moral philosophy, and a critical question that defines how society itself functions. It’s also a key question to understand what role each of us plays to make society successful. These words were started as society was entering a seismic shift, the COVID-19 pandemic had started, people were dying, fear ran deep, and the disease impacted some far more than others. Some reactions to this crisis were based on what was good for society, others came down to the interests and desires of the individual. The question of what we owe to each other had rarely been so imperative. I am finishing1 this as society enters another seismic shift. Fear again runs deep, there will be deaths, people are divided to an extent not seen in generations, an…  ( 9 min )

  • Open

    Developer Tools & Productivity
    Technology improves and advances ceaselessly, new tools are created and change how people work. Some are small and simple, making people somewhat more productive. Others revolutionise the way people work. These revolutionary tools may come along only once or twice in a generation, and when they do, they tend make people uncomfortable. They can make people question their role, their skills, their future, and their place in the industry. I would like to take a few minutes to talk about a revolutionary change in how developers work. I kindly ask the reader to reserve judgement on this topic till the end of this article, to fully understand the intent. Shifting Paradigms Developer tools have historically been hard to use, complex, and required extensive and specialised knowledge to accomplish …  ( 5 min )

  • Open

    【漏洞分析】CVE-2025-55182 React2shell远程代码执行解析
    React2Shell (CVE-2025-55182) 漏洞分析 这几天应该很多人在忙着应急这个“核弹级”漏洞,它已经成了安全圈里的“新顶流”。但网上还没有比较详细的分析文章,而我又对公开POC中的一些细节实在好奇,所以今天来一探究竟。 因为我对React不够了解,所以本文没有任何扩展内容,仅仅只是把漏洞利用过程调试了一遍,并且因为利用过程有点绕,所以配上了流程图方便理解。  ( 3 min )

  • Open

    google tagmanager idor - 飘渺红尘✨
    谷歌安全团队这两天修复了我之前提交的安全漏洞,经过谷歌安全团队允许,完整无码公开下这份漏洞,这个漏洞相对简单,越权漏洞,危害较小,给了1337刀。 强如四大金刚,也会存在安全问题,在漏洞挖掘中要相信自己,要有信心,然后是耐心和仔细。 https://analytics.google.com/ Vul  ( 5 min )

  • Open

    [笔记] 从 Tokenization 视角看生成式推荐(GR)近几年的发展(2025)
    不同类型的真实世界 建模元素 对应的模型类型 感知世界(Perceptual World) 视觉(Vision) 扩散模型(Diffusion Models, DMs) 认知世界(Cognitive World) 语言(Language) 大语言模型(LLMs) 行为世界(Behavioral World) 交互(Interaction) 用户行为的模型? 从模型和现实世界的对应关系来看,感知世界(Perceptual World)和 认知世界(Cognitive World) 都已经有了对应的大模型类型,分别基于视觉(Vision)和语言(Language) 建模, 并且基本都是基于生成式架构,实际效果非常好。 推荐领域属于行为世界(Behavioral World), 这个场景基于交互(Interaction)建模,目前还没有跟前两个领域一样成功的模型。 一个思路是:如果大量场景已经充分证明了生成式是一把非常好的锤子, 那我们是不是能把还没有很好解决的问题变成钉子?—— 具体到推荐场景, 就是通过一些工程和算法手段,把推荐任务变成一个生成任务,从而套到生成式框架里。 这就是生成式推荐模型(generative recommendation models)背后的思想。 最近有一篇很详尽的关于这个领域近几年发展的综述: Towards Large Generative Recommendation: A Tokenization Perspective。 本文整理一些阅读笔记和思考。 水平及维护精力所限,文中不免存在错误或过时之处,请酌情参考。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 1 背景 1.1 什么是生成式模型(…  ( 7 min )

  • Open

    构筑智能化攻防格局:网络安全实验室的战略与未来蓝图
    本文链接 《构筑智能化攻防格局》提出“三位一体”安全实验室战略:以渗透测试、威胁情报、安全研究为支柱,通过领域专家制、流程固化与知识标准化,打造可复制的安全底座;引入AI智能体,实现威胁情报秒级研判、漏洞自动收集-分析-复现、渗透测试Multi-Agent报告生成,已把Tomcat CVE-2025-53506从披露到POC验证压缩至小时级。报告给出AI工具化→平台化→自主化三阶段路线图,目标让攻防策略随威胁动态自进化,为企业构建持续演化的智能安全中枢。  ( 3 min )
    轻装上阵:Javassist聚焦代码审计关键点实践
    本文链接 《轻装上阵:Javassist聚焦代码审计关键点实践》提出用字节码操作库Javassist突破传统正则与SAST瓶颈,精准提取Spring等框架的路由映射、方法参数与注解语义,自动识别Runtime.exec、ScriptEngine.eval等敏感sink;结合参数签名分析,复现JDBC任意文件写入等CVE,实现无源码、低成本的蓝军快速审计。工具已集成路由提取、sink定位、参数校验三大模块,可作为轻量级辅助手段,显著提升代码审计效率。  ( 3 min )
    企业内部的渗透测试自动化探索
    本文链接 《企业内部的渗透测试自动化探索》由平安科技廖汇铭分享,提出“API 资产中心 → 标准化用例 → AI 漏洞研判”三位一体自动化平台:通过流量镜像、语义打标、路径折叠去重,5 分钟完成 10 万级 API 分类(准确率 92-95%);基于标签自动生成水平越权、SQL 注入、支付绕过等 200+ 标准化用例,分布式执行器 1 小时完成千接口测试;AI 研判智能体引入正向-反向双提示与证据链校验,越权误报降低 80%,整体自动化覆盖率 45.9%,漏洞发现占比 38.3%,实现渗透测试从“手工经验”到“可持续工程体系”的跃迁。  ( 3 min )
    攻击者视角下的业务智能体渗透
    本文链接 《攻击者视角下的业务智能体渗透》由平安Hamber团队出品,聚焦大模型与智能体在金融科技场景中的新型攻击面。报告系统梳理提示词注入、工具滥用、身份伪装、MCP越权、知识库污染、XSS/SSRF/命令执行等10+实战手法,演示如何通过一次“查薪资”诱导智能体泄露他人工资、伪造token调用未公开接口、上传恶意MCP插件横向移动并获取内网shell。作者提出“智能体应用扫描-画像-决策”闭环渗透框架,结合CVE实例与平安众测数据,输出可落地的风险矩阵及加固方案,助力企业从设计、部署到运营全生命周期守护数字员工安全。  ( 3 min )

  • Open

    Fuzzing4LLM:撬动大语言模型的安全对齐机制
    本文链接 《Fuzzing4LLM:撬动大语言模型的安全对齐机制》由科大讯飞梅瑞提出,发布进化式越狱框架 ForgeDAN:通过字符-词-句多级扰动、场景假设/角色扮演变异与语义适应度评估,自动生成高隐蔽性对抗提示,在 DeepSeek、Qwen 等模型上攻陷率最高达 98%;并构建双维度判别器,精准区分拒绝/顺从与安全/有害输出,显著降低假阳性。报告同步给出覆盖越狱、投毒、跨模态、智能体等 20 + 场景的 AI 安全评测体系,以及多层防护、人机协同、数据全生命周期治理的主动防御方案,为大模型安全对齐提供从攻击到评测再到治理的完整闭环。  ( 3 min )
    VPN与红队攻防
    本文链接 《VPN与红队攻防》作者10余年红蓝对抗经验,系统披露国内主流VPN设备“大网扩线-漏洞利用-后渗透”完整攻击链:从未授权下载、默认口令、P12证书泄露到任意文件读取获取运维DB,绕过双向校验植入后门key,再借客户端日志、token与数据库构造字典,实现无爆破隐蔽登录。报告给出防护清单:默认配置清零、2FA、堡垒机运维、管理口白名单、集中日志审计,帮助企业堵住VPN这一高权限横向通道。  ( 3 min )
    kubernetes RBAC 安全风险研究
    本文链接 《Kubernetes RBAC 安全风险研究》系统梳理 K8s 授权链路,从默认服务账户、授权过度授予、身份冒用-impersonate权限、pods 资源组配置错误、 Webhook 配置错误和ServiceAccount过度授权 六大配置缺陷,演示“容器内提权→横向移动→集群接管”完整攻击链;结合微服务、OIDC-CAS、Nexus 供应链镜像投毒等真实场景,给出最小权限、GitOps 左移、PSS/PSA 基线、审计+自动化响应的闭环防御方案,帮助企业在云原生时代快速发现并修复 RBAC 滥用,提升集群安全韧性。  ( 3 min )
    红队基础设施自动化多云部署能力
    本文链接 《红队基础设施自动化多云部署能力》基于 IaC 理念开源 RedC 引擎,实现“一条命令”分钟级拉起 C2、DNSLog、代理池、扫描节点等全套红队云设施。通过 Terraform 模板化+抢占式 Spot 实例,将单场演练成本压至 500 元以内,支持 AWS/阿里云/腾讯云多云并发、IPv6 扫描、密文存储与 IMv2 防 SSRF,自带 clash 负载均衡与 Web 控制台,已提速打点 30-40 倍。报告同时提出 SaaS 化调度、MCP 大模型自动调起攻击场景的开源路线,终结红队“重复造轮子”时代。  ( 3 min )

  • Open

    数据安全意识培训:守护数字生命线
    本文链接 《数据安全意识培训:守护数字生命线》面向产品与研发全员,重新定义数据安全是产品特性而非负担。课程围绕机密性、完整性、可用性三大基石,给出可落地的“安全左移”实践:需求阶段做威胁建模与数据分类,编码阶段用参数化查询、bcrypt 哈希和 CI/CD 自动扫描,部署阶段实施基础设施即代码安全、密钥托管与日志脱敏,运营阶段持续监控、限流防爬并闭环漏洞。通过越权访问、硬编码密钥、日志泄露三大真实案例,演示设计疏忽如何演变为百万罚款与勒索事件,并提供产品经理、研发工程师、通用办公三张速查卡,让“谁该看到什么数据”成为日常共识。培训最后发出“黄金三步法”应急号召:断网-报告-改密,强调“安全无小事,从一行代码、一个设计决策做起”,帮助团队把数据保护写进需求、融入编码、落到运维,共筑企业数字生命线。  ( 3 min )
    数据安全,你我共筑的防火墙
    本文链接 《数据安全,你我共筑的防火墙》以“公共卫生”作喻,用鲜活案例拆解四大“数据杀手”:钓鱼邮件、弱密码复用、公共Wi-Fi与内部泄密,指出九成泄露源于日常疏忽。课程给出 4 套可立即落地的“金钥匙”准则:12 位强密码+管理器、四步识破钓鱼、人走锁屏+加密存储、社交云盘“零分享”,并配套“断-报-改”黄金应急法,帮助员工把安全从“负担”变习惯,让每人成为企业最可靠的“人肉防火墙”。  ( 3 min )

  • Open

    An Illustrated Guide to AP2 (Agent Payment Protocol) (2025)
    With the rapid evolution of GenAI and the growing trend of accomplishing more and more tasks through chat, can you imagine a day (perhaps in the near future) we can buy almost anything simply by chatting? Instead of browsing e-commerce sites, comparing products yourself, you’ll just tell your agent what you need. It will handle everything: selecting options, comparing features, negotiating prices, making payments, and ensuring the product arrives at the right place and time. To bring this vision to life, one essential piece is still missing: a payment protocol designed for agent-to-agent transactions. That’s exactly why AP2 was created. This post offers an illustrative guide to this emerging topic. Fig. Shopping agent view of the "Buy a coffee maker" AP2 demo. Fig. Call flow of the AP…  ( 15 min )

  • Open

    A.S.E 2.0 正式发布 | 开放生态 · 动静协同 · 让 AI 编程安全可量化
    A.S.E (AI Code Generation Security Evaluation) - 你的 AI 鉴赏官,为你提供大模型安全实践工具,让你一眼就能选出最靠谱的 AI 队友。  ( 12 min )

  • Open

    CSPT漏洞浅析 - 飘渺红尘✨
    CSPT全称是Client-Side Path Traversal ,即客户端路径遍历。 概念说明 CSPT 全称 Client-Side Path Traversal(客户端路径遍历),是一种针对前端应用的漏洞,核心是攻击者通过篡改 URL 参数、请求参数等,让浏览器(客户端)错误地向非预期的服务  ( 3 min )

  • Open

    Glitching on STM32 Blue Pill
    Introduction With the rise in awareness of enhanced hardware-level security practices, it is high time that the rising security researchers got their hands dirty with some advanced hardware attacks. Thus, keeping this into consideration, I performed a small attack on the STM32 Blue Pill board. In this blog, I will discuss how I managed to […] The post Glitching on STM32 Blue Pill appeared first on Payatu.  ( 66 min )

  • Open

    Reducing notification overload for a quieter browsing experience in Chrome
    We're constantly working to improve your browsing experience. To help you cut through the noise and reduce notification overload, we’re launching a new feature to automatically remove notification permission for sites you haven't interacted with recently. Today, Chrome’s Safety Check already does this for other permissions such as camera and location. The feature will be launched in Chrome on Android and desktop. Data indicates that users frequently receive a high volume of notifications, resulting in minimal engagement and high disruption. Less than 1% of all notifications receive any interaction from users. But notifications can be genuinely valuable and helpful. Therefore, this feature will only revoke permissions for sites when there is very low user engagement and a high volume of notifications being sent. This feature does not revoke notifications for any installed web apps. Chrome will inform you when notification permissions are removed. If you prefer to keep getting notifications from a particular website, you can easily re-grant the permission at any time through Safety Check or alternatively by visiting the site and enabling notifications again. You can also choose to turn off the auto-revocation feature entirely. Posted by Archit Agarwal, Product Manager, Chrome  ( 12 min )

  • Open

    [笔记]《人工智能简史(第二版)》(2025)
    尼克的《人工智能简史(第二版)》从人和流派传承的角度介绍了人工智能作为计算科学一个分支的发展过程, 内容和风格有点偏学术史,用作者的话说,“写法比较偏重基础和方法论,而不太注重应用”。 作为一本不太“常规”的人工智能入门读物,适合领域内的部分专业读者, 或者想从科学、哲学、伦理学等更高角度理解和看待人工智能的读者。 本文整理一些个人阅读笔记和思考。 水平及维护精力所限,文中不免存在错误或过时之处,请酌情参考。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 0 前言 0.1 哈代:一等智力 vs. 二等智力 0.2 任正非 1 达特茅斯会议:人工智能的起源, 1956 1.1 经典读物 1.2 Chomsky:机器可以思考吗?-> 潜艇会游泳吗? 1.3 AI 的两面:工程和科学 2 自动定理证明兴衰记 2.1 自动定理证明的起源 数学哲学三大派 逻辑学的源头:亚里士多德三段论 2.2 思想实验:Brain in a vat 2.3 王浩(1921—1995) 2.4 吴文俊(1919—2017) 高龄开始学习编程 为人类文明做出贡献 2.5 哲学问题 有黑盒的理解不能算理解,有黑盒的证明也不能算证明 人已经无法核实部分计算机证明的结果 数学家的归宿 计算模糊了理性和经验的边界 2.6 现状 时代交替 (2006):定理证明小组被裁,深度学习论文横空出世 定理证明领域的名字演化 2.7 结束语 数学家不把逻辑学家当回事 逻辑似乎处于一切科学的底部,因为逻辑探索一切事物的本质 两个 Alpha-zero 下棋,我们人类已经看不懂了 3 从专家系统到知识图谱 3.1 机器归纳法:用现在的话说就是机器学习 3.2 知识表示 逻辑是最方便的知识表示语…  ( 9 min )

  • Open

    00后黑客落网!曾运营大型僵尸网络,频繁攻击腾讯游戏、DeepSeek、X平台
    据美国司法部通报,活跃多年的高危DDoS僵尸网络“RapperBot”(历史频繁攻击腾讯游戏、DeepSeek、X平台的幕后黑手)现已被成功取缔。该僵尸网络主谋、22岁的美国俄勒冈州男子Ethan…  ( 11 min )

  • Open

    清华大学网络研究院 NISL 实验室发布 SecCodeBench 2.0:面向智能编码工具的代码安全评测体系升级
    引言:AI 生成代码评测体系的局限与演进 随着以大语言模型(LLM)为核心的辅助编程工具的普及,AI 生成代码的安全性 已成为业界关注的焦点。为科学地评测 AI 生成代码的安全性,发现其内在缺陷并促进模型安全能力的提升,一套 全面、可靠 的评测基准至关重要。 然而,社区现有的安全评测基准在三个核心维度上存在显著的局限性,这使得它们难以真实反映模型或Agent的安全编码能力: 测试用例质量参差不齐:许多数据集来自开源代码,严重依赖自动化生成和简单过滤,缺乏人工的深度参与。这导致了 数据分布失衡,大量低优先级的安全问题占据主导,无法有效衡量模型在关键漏洞上的表现; 无效的测试用例,部分题目在设计上存在缺陷(例如,由于过于关注漏洞点而不是修复点,导致在给定的约束下无法生成正确的安全代码),这会导致对模型能力的系统性低估,而非客观评估; 潜在数据污染,测试用例所属的开源仓库代码可能已经作为了模型的预训练语料,进而影响评估的公正性。 评估方法过于单一且精度不足:现有的评估方法大多依赖于简单的正则表达式或代码检测工具,这导致它们难以准确识别 语法或语义复杂 的代码变体,并且完全忽略了必须通过 真实运行 才能验证的漏洞。更重要的是,许多评估方法 忽略了功能的重要性,这导致 评估标准与实际可用性脱节,甚至会将功能损坏的“安全代码”判定为更优解。 未能覆盖智能编码工具:真实编程场景已进化至 智能编码工具(Agentic Coding Tool),即开发者依赖的是能够自主调用工具、检索知识库的智能体。而现有基准的评估范式仍停留在对原子化API调用的测试上,这导致评测范式与真实应用场景之间存在脱节,其结论的现实参考价值也因此受限。 为科学评估模型在真实开发环境中的安全编码能力,评测体系需持续迭代以适配技术演进与场景变化。基于此,我们推出SecCodeBench 2.0,这是一个 专为现…  ( 1 min )

  • Open

    [译] 从 OpenDeepResearch 背后的设计演进,解读 AI 领域反复学到的一课(2025)
    本文翻译自 2025 年的一篇文章 Learning the Bitter Lesson。 来自 github.com/langchain-ai/open_deep_research 作者。 过去 70 年 AI research 领域学到的最大经验是:以计算作为支撑的通用方法 (general methods that leverage computation)是终极方案(ultimately the most effective),而且大幅领先其他方式。 水平及维护精力所限,译文不免存在错误或过时之处,如有疑问,请查阅原文。 传播知识,尊重劳动,年满十八周岁,转载请注明出处。 1 反复学到的一课 1.1 AI Research 领域 1.2 AI 工程领域 2 以 Open Deep Research 为例 2.1 添加结构(假设) 2.2 结构开始成为瓶颈 2.3 移除结构 3 总结 4 致谢 过去 70 年 AI research 领域学到的最大经验是:以计算作为支撑的通用方法 (general methods that leverage computation)是终极方案(ultimately the most effective),而且大幅领先其他方式。 Rich Sutton,The Bitter Lesson 1 反复学到的一课 1.1 AI Research 领域 The Bitter Lesson 在许多 AI 研究领域一次次地被证实,比如国际象棋、围棋、语音、视觉。 用好计算(leveraging computation)被证明是最重要的事情, 而我们强加给模型的"结构"反而往往会限制它们用好不断增长的计算能力。 这里所说的”结构”是什么意思? Often structure includes inductive biases abo…  ( 1 min )

  • Open

    Three Lazarus RATs coming for your cheese
    Authors: Yun Zheng Hu and Mick Koomen Introduction In the past few years, Fox-IT and NCC Group have conducted multiple incident response cases involving a Lazarus subgroup that specifically targets organizations in the financial and cryptocurrency sector. This Lazarus subgroup overlaps with activity linked to AppleJeus, Citrine Sleet, UNC4736, and Gleaming Pisces. This actor uses … Continue reading Three Lazarus RATs coming for your cheese →  ( 29 min )

  • Open

    【漏洞分析】CVE-2025-9132 "Await Using" Can't Wait
    既然这是Project Zero的BigSleep发现的第一个V8漏洞,Buff这么多,那就很难不来一窥究竟了。  ( 3 min )

  • Open

    blog升级https访问
    blog很久没更新维护了,最近http频繁提示不安全,索性升一下https。 这里记录一下流程,不得不说,以前 […]  ( 4 min )

  • Open

    记「乾坤大挪移」式使用双系统
    我是在 WWDC25 期间直接把我的工作电脑 MacBook Pro 升级到 macOS 26 Beta 的。当时人还在 Apple Park 参加活动,后面也基本以玩为主,因此没有被这个系统影响到工作。 回国后,从游玩状态中调整过来,准备好好干活。但真的高强度使用 macOS 26 Beta 时,觉得非常别扭。首先,我的 M3 Max 的整体 UI 响应速度,就像回到了用一台 Intel 电脑;其次,Liquid Glass 非常不成熟,过度的动画、透明和阴影,始终觉得内容重点被夺走了;最后让我无法忍受的是,菜单上那些画蛇添足的图标,不再是原本干干净净的 macOS,让我有种在用 Windows 的感觉。于是我决定装双系统,回到 macOS 15 工作。 过去几年,基本每年我都会安装双系统开始我的 Beta 系统的体验,每年的原因都差不多,主要还是新系统无法跑最新 Xcode 稳定版本,虽然这个在今年被解决了,但在可用性方面却大大降低了。我相信(或期望)可用性问题在正式版本会解决(或者在 macOS 27、28)。但那是后话了,本篇我就简单介绍下我是如何进行「乾坤大挪移」式使用双系统。 如何安装双系统 昨天花了几小时装+配置好 macOS 15 环境后,整个立刻就感觉清爽、轻盈了起来。 在 Mac 上装双系统非常简单,只要打开 Disk Utility,新加一个 APFS 卷,然后去 App Store 下载最新 macOS 的稳定版本,然后把这个 macOS 装在新加卷上面就可以了。因为两个 macOS 可以互读硬盘,因此文件也可以共享,很快就可以把自己的需要的环境设置起来。 但是,今年我做了一些不一样的操作,我决定隔离两个系统的环境,进行「乾坤大挪移」。一个重要的出发点是,我的 macOS 26 Beta 的环境已经很乱了,我不想简单的把文档复制或链接过来,我想在使用过程中,把真正需要的文件挪过来,不要的文件就留在「老系统」,直到我不需要并抛弃它。 禁止当前系统索引另外一个系统的文件 默认情况下,双系统的 Mac 会挂载另一个系统的分区,Spotlight 也会自动去索引所有文件,于是你搜索的时候可能会出现重复的内容,这在我过去几年的双系统体验中是一个很烦恼的事情。而且即使在 Spotlight 中设置了排除项,重启以后依然会无效。 今年我找到了一个新方法,那就是在命令行层面禁止索引另一个系统的分区,真正做到互不影响。 禁止索引系统盘: sudo mdutil -i off /Volumes/Macintosh\ HD 禁止索引数据盘: sudo mdutil -i off /Volumes/Macintosh\ HD\ -\ Data 执行完这个后,在当前系统下就搜不到另一个系统的文件了,算是「软」隔离成功。 渐进式「乾坤大挪移」 前面已经提到「乾坤大挪移」了,那就是我想在使用干净的系统的过程中,只移动需要的文件到当前系统来,老的、旧的东西就不动了,直到我抛弃(删除)那个系统。 昨天当我设置好基本的开发环境后,我就开始工作了一会。当时的系统环境只有必要的软件和工程项目,没有任何多余的东西,这使得我的工作状态也非常轻盈和舒适:就像你乱了很久的小屋,有一天你狠狠地打扫了一遍,把所有不重要的东西都清理掉了,尽管小屋还是那个小屋,但是生活起来绝对会舒适很多。 就这样,昨天在我的「新系统」里,进入了回国以后久违的一个工作状态。 后记 接下去,我会在日常的工作中,继续把需要的文件和软件挪回来。在这个过程中,新系统也可能会继续变乱,然后也越来越臃肿,但是没关系,下一年,我再来玩一次这样的双系统「乾坤大挪移」——就像需要经常打扫屋子一样,电脑的系统不仅需要经常打扫,而且需要利用可以装双系统的优势,比起找什么需要、什么不需要式的清理,双系统更有一种「快刀斩乱麻」的感觉。 读到这里,你可能会说:我的硬盘不够我装双系统!那…下台电脑请买 2TB 起的硬盘,只能这样了。 附图:  ( 1 min )

  • Open

    Understand Neural Nets better, post 5 of N -- Code Assistant shootout
    In a series of previous blogposts [1, 2, 3, 4] I ran some experiments drawing the boundaries of the polytopes generated by a fully-connected leaky ReLU network while it was getting trained on reproducing an input image. As I tried to scale the experiments to larger networks, I noticed a dramatic slowdown in the code, caused by the calculation of a hash of the activation pattern happening on CPU -- so each training step would be fast, but then everything would grind to a halt for the visualisation, and for each pixel the code would forward-evaluate the NN (all in all 1024*1024 times), and whenever the prediction was calculated, it'd transfer the activation pattern to CPU and then perform the hashing. This was very slow, and very non-parallel. I had contemplated writing some custom CUDA code…  ( 14 min )

  • Open

    Introducing Skia Graphite: Chrome's rasterization backend for the future
    Today's The Fast and the Curious post covers the launch of Skia's new rasterization backend, Graphite, in Chrome on Apple Silicon Macs. Graphite is instrumental in helping Chrome achieve exceptional scores on Motionmark 1.3 and is key to unlocking a ton of future improvements in Chrome Graphics. A brief history of Skia in Chrome since the very beginning. Skia eventually ran into performance issues as the web evolved and became more complex, which led Chrome and Skia to invest in a GPU accelerated rasterization backend called Ganesh. Over the years, Ganesh matured into a solid highly performant rasterization backend and GPU rasterization launched on all platforms in Chrome on top of GL (via ANGLE on Windows D3D9/11). However, Ganesh always had a GL-centric design with too many speciali…  ( 16 min )

  • Open

    CVE-2025-32023 Redis 漏洞分析
    <link rel="stylesheet" type="text/css" hr  ( 6 min )

  • Open

    A non-anthropomorphized view of LLMs
    In many discussions where questions of "alignment" or "AI safety" crop up, I am baffled by seriously intelligent people imbuing almost magical human-like powers to something that - in my mind - is just MatMul with interspersed nonlinearities. In one of these discussions, somebody correctly called me out on the simplistic nature of this argument - "a brain is just some proteins and currents". I felt like I should explain my argument a bit more, because it feels less simplistic to me: The space of words The tokenization and embedding step maps individual words (or tokens) to some \(\mathbb{R}^n\) vectors. So let us imagine for a second that we have \(\mathbb{R}^n\) in front of us. A piece of text is then a path through this space - going from word to word to word, tracing a (possibly convolu…  ( 14 min )

  • Open

    CVE-2025-36463 Sudo_chroot Elevation of Privilege 漏洞分析
    <link rel="stylesheet" type="text/css" hr  ( 9 min )

  • Open

    LuaJIT trickery
    In this blog-post I will bring together two cool things: ELF DSO injection into runtime processes and LuaJIT. Lua itself is a very nice scripting language and LuaJIT integrates a very decent FFI module which contains a cdecl parser and other interesting things. I do not need the speedup of LuaJIT over Lua so much, but the C-binding and -inter-operability is "awesome". I wrote my first ELF injection tools >20y ago, so my actual injectso project needed some polishing first to work with current distros at all. Not only did the ld.so impl and glibc pathnames change, CPUs also feature new extensions like SSE2 and compilers make use of it for string and memory related functions like memset(). The drawback in this particular case was that SSE2 instructions add 16byte alignment requirements for the address operands (in this case indirectly the stack) which made injectso fail on distros that make use of it. Now that was fixed and injectso runs on most glibc based distros again, it was time to add LuaJIT support to allow loading of Lua scripts into "foreign" processes in order to do similar things that you know from frida: accessing memory locations, hooking functions, dumping parameters etc. but with much less bloat. Here is how it would look in Lua: It was necessary to add peek/poke primitives to LuaJIT as well as a trap mechanism for functions in order to hook/unhook functions and dump their parameters or otherwise access memory at this point. Changing of function parameters on entry is not yet supported, but would be easy. When you hook a python interpreter with this script, it would look like this: Is that cool? Running Lua code when Python calls into a C function to do some things. Some fun that AI assisted coding will never give you. You can try out yourself here.  ( 7 min )
2026-06-22T05:13:12.365Z osmosfeed 1.15.1