llmai

Prompt Engineering for Security Agents - A Measurable Approach with GEPA

One of the things that has been winding me up about working with LLMs is how unmeasurable prompt modifications can be. You know how it goes, you task your agent to perform a task, and eventually it starts to deviate from the objective. So you head into your AGENTS.md to correct course, and hope that by shouting at the AI in just the right incantation, you fix the problem. GEPA is one established solution to this problem.

Read More »
claudellmmcpvulnerabilitycve

An Evening with Claude (Code)

A deep dive into discovering CVE-2025-64755, a vulnerability in Claude Code v2.0.25. This post walks through the process of reversing the obfuscated Claude Code JavaScript, and exploiting weak regex expressions to achieve code execution unprompted.

Read More »
windowsredteamreversing

Administrator Protection Review

Microsoft will be introducing Administrator Protection into Windows 11, so I wanted to have an understanding of how this technology works and how it interacts with existing offensive tooling. While this technology is just a thin wrapper around a separate account, there are a few nuances such as who is permitted to access these accounts, as well as existing UAC bypasses which are still effective against the new "backdoorless" Administrator Protection. This post explores these nuances in detail.

Read More »
llmmachine-learning

Tokenization Confusion

In this post we look at the new Prompt Guard 2 model from Meta, and introduce a concept I've been calling "Tokenization Confusion" which aims to confuse Unigram tokenization into generating tokens which will result in the misclassification of malicious prompts. We'll also look at why building up our ML knowledge will lead to better findings when assessing LLM API’s, as I discovered during a flight across the Atlantic.

Read More »
sqlserverredteam

The SQL Server Crypto Detour

One of the things that I love about my role at SpecterOps is getting to dig into various technologies and seeing the resulting research being used in real-time. This post will explore one such story of how I was able to go from a simple request of recovering credentials from a database backup, to reverse engineering how SQL Server encryption works, finding some new methods of brute-forcing database encryption keys.. and finally identifying a mistake in ManageEngine’s ADSelfService product which allows encrypted database backups to reveal privileged credentials.

Read More »
idpredteamoktapingoneloginentraidazuread

Identity Providers for RedTeamers

Originally presented at SOCON-2024, and continuing the series into post-exploitation techniques against Identity Providers, in this blog post we'll look at Ping, OneLogin and Entra ID. I'll discuss how post-exploitation techniques effective against Okta apply to other providers, release new tools for post-exploitation, and look at what proves to be effective when critical assets lie beyond an Identity Provider portal.

Read More »
macosexploit

MacOS "DirtyNIB" Vulnerability

While looking for avenues of injecting code into platform binaries back in macOS Monterey, I was able to identify a vulnerability which allowed the hijacking of Apple application entitlements. Recently I decided to revisit this vulnerability after a long time of trying to have it patched, and was surprised to see that it still works. There are some caveats introduced with later versions of macOS which we will explore, but in this post we’ll look at a vulnerability in macOS Sonoma which has been around for a long time, and remains an 0day, urm, to this day.

Read More »
oktaredteam

Okta for Red Teamers

In this blog post, I'll discuss some of the post-exploitation techniques that I've found to be useful against Okta. Specifically, this post will look at how to use delegated authentication to our advantage, silver tickets, Okta AD agent spoofing, and finally how to deploy a fake SAML provider.

Read More »
Next Page