Aikido's AI pentester found this file-access flaw in Rocket.Chat. A closer look at MongoDB's ObjectId() showed the weak randomness that makes it exploitable.

Our AI pentest agents found a critical phpBB auth bypass (CVE-2026-48611): one unauthenticated request logs you into any account. See the exploit and the fix.

DevArea hosts a freelance developer marketplace backed by several web applications on different stacks. I’ll find a JAR file on an open FTP server that matches one of the web services that uses a vulnerable version of Apache CXF. I’ll abuse its attachment handling to read arbitrary files off the host. Those files leak credentials for a Hoverfly API simulation instance, where a command injection in the middleware feature gives a shell as the first user. From there I’ll pivot through a custom SysWatch monitoring app, forging a session cookie with a secret pulled from a world-readable environment file and slipping a command past a weak input filter to run as the service account. Finally, I’ll exploit a flawed symlink check in a script that account runs as root to read the root login key and get a shell. In Beyond Root I’ll explore why I couldn’t read user.txt from the initial file read, and look at the admin password from the syswatch application.

Node.js is debating whether AI-driven security report volume warrants moving more vulnerability reports into public workflows.

Table of Contents Intro Lore Why "Find All The Vulnerabilities" does not work Minimal Scaffolding That Actually Helps Case Study: Claude Opus 4...

YouTube video by Security Weekly - A CRA Resource

We found a stored XSS in Roundcube's draft attachment endpoint that, chained with a cookie tossing technique, gives an attacker full access to a victim's inbox. Here's how the exploit chain works and ...

Application security teams often deploy AI-powered scanners that use Large Language Models (LLMs) to scan web applications for vulnerabilities. While ...

Aikido's AI pentest agent found three XSS vulnerabilities in Mailcow, one of which let unauthenticated attackers take over administrator accounts. All issues have been patched as of version 2026-03b.

### Summary Bypasses of the attachment view logic and SVG sanitizer make it possible to upload and render an SVG that runs malicious JavaScript. An extension of `.png` with content type of `imag...

Contribute to Syzik/jwt-reauth development by creating an account on GitHub.

Noma Security researchers disclosed GrafanaGhost, a Grafana vulnerability that uses indirect prompt injection to silently exfiltrate sensitive enterprise data — bypassing AI guardrails without a login...

A look at GitHub Actions’ 2026 roadmap, outlining how secure defaults, policy controls, and CI/CD observability harden the software supply chain end to end.

Snapped is a Linux box hosting a static site behind nginx, with an Nginx UI admin panel. I’ll exploit CVE-2026-27944 to decrypt a backup download from the Nginx UI to find bcrypt password hashes in a SQLite database. I’ll crack one to get SSH access. To escalate to root, I’ll exploit CVE-2026-3888, a recent vulnerability in snapd where systemd-tmpfiles deletes snap-confine’s private temp directory, allowing me to win a race condition and replace the dynamic linker with a malicious payload that runs as root.

YesWeHack Dojo #49 writeup: exploiting wildcard argument injection in shell cp and grep commands to access internal secrets

Welcome back! As we near the end of 2025, we are, of course, waiting for the next round of SSLVPN exploitation to occur in January (as it did in 2024 and 2025). Weeeeeeeee. Before then, we want to clear the decks and see how much research we can publish. This

As part of our CVE monitoring, we came across GHSA-pcq9-mq6m-mvmp (CVE-2025-68402), an authentication bypass in FreshRSS, a self-hosted RSS aggregator. It ...

CodeAnt AI found a critical authentication bypass in pac4j-jwt where an attacker can impersonate any user using only the RSA public key. Full PoC and disclosure.

## Impact A stored cross-site scripting (XSS) vulnerability was identified in the email rendering feature of AliasVault Web Client versions 0.25.3 and lower. When viewing received emails on an ali...

Last December, quite unrealistically, I took a solemn oath: I would not write again about AI for at least another year. I was growing tired with the incessant noise, the lack of stability, and the sel...

GitHub Action to alert on security patches before the CVE drops. - spaceraccoon/vulnerability-spoiler-alert-action

A compromised npm publish token was used to push a malicious postinstall script in cline@2.3.0, affecting the popular AI coding agent CLI with 90k wee...

Security researchers report widespread abuse of OpenClaw skills to deliver info-stealing malware, exposing a new supply chain risk as agent ecosystems...

Nearly one in three Meta ads found to point to a scam, phishing or malware

Bamboo offers a Squid HTTP proxy through which I’ll access a PaperCut NG instance. I’ll use Spose to scan through the proxy and discover the print management application. I’ll exploit an authentication bypass vulnerability in PaperCut and use application access to enabling print scripting to get code execution. For privilege escalation, I’ll abuse a root process that runs a script from the papercut user’s home directory.

Dan Moore will show you what JWTs are, how they work and everything you need to know to use them properly in your applications

Imagery hosts a Flask-based image gallery application. I’ll exploit a stored XSS vulnerability in the bug report feature to steal an admin cookie. From the admin panel, I’ll use directory traversal to read the application source code, finding a command injection vulnerability in the image crop feature that requires access as a test user. After reading the database and cracking the test user’s password hash, I’ll exploit the command injection to get a shell. I’ll find an encrypted backup file and brute-force the pyAesCrypt password, getting access to an older backup with additional hashes. After cracking another user’s hash, I’ll pivot to a user that can run a custom backup utility as root via sudo. I’ll show two ways to abuse this. In Beyond Root, I’ll show why SSH is broken and how to get around it.

How We Found Code Execution in Anthropic’s Official Git MCP Server TL;DR What happened: Cyata discovered three security vulnerabilities in mcp-server-git, the official Git MCP server maintained by Anthropic. These flaws can be exploited through prompt injection, meaning an attacker who can influence what an AI assistant reads (a malicious README, a poisoned issue description, […]

I didn't know Harbour even existed as a language when I found this bug. The fun part is that I also ...

HackNet hosts a social media site for hackers built with Django. I’ll find an HTML injection in the username field that, combined with how the likes page renders usernames, leads to server-side template injection. While Django templates are restrictive, I’ll use the SSTI to dump user data including plaintext passwords, finding one user whose email reveals their Linux username. After SSHing in, I’ll discover Django’s FileBasedCache uses pickle serialization with a world-writable cache directory. By replacing cache files with a malicious pickle payload, I’ll get a shell as the web user. From there, I’ll crack a GPG key password to decrypt database backups, finding a password shared in messages that works for root.