Navaneethan L·4d agoAI Security: Prompt Injection, Adversarial Inputs, and DefenseThe most dangerous attack on your enterprise AI stack will not target the model. It will target the descriptions of the tools your model…A response icon1A response icon1
Raven·6d agoWindows Post-Exploitation: Registry Persistence, Scheduled Tasks & Access Control AbuseOnce an attacker gains initial access to a Windows machine, the attack does not necessarily end there. In many cases, initial access is…
Hassan Rasyid·6d agoThe Persistence Trap: How DLL Search Order Hijacking Turns a Windows Application into a Red Team…While analyzing the Windows implementation of Krisp version 2.51.3, I came across an interesting behavior related to how the application…
Eng Soon Cheah·Oct 31NVIDIA Garak: the nmap of the LLM worldAs Large Language Models (LLMs) become increasingly integrated into business applications, testing their security is no longer optional…
0xdzubair·Oct 26Session Swap: Exploiting Client-Side Role Cookies in Access Controlwebverselabs-pro.comA response icon1A response icon1
Vin·Oct 24Tattletale: A Small Model Watching a Big OneTeaching a small model to catch what big models get wrongA response icon2A response icon2
Vivekc·Oct 24AI Agent Evals: Right Answer, Wrong PathWhy AI agent evaluation has to check the path, not just the final answer, and how to build evals that catch real regressions.
blackpandalabs·Oct 23You Do Not Have a People Problem. You Have a Verification Problem.An interactive field note on authorized social-engineering simulations.
Abhinav Dobhal·Oct 22Testing AI Apps with DeepEval (Part 7): AI Safety Testing and Red TeamingYour support agent passes every quality test. It answers "Where is my order ORD-1042?" correctly, uses the right tool and stays on topic…
