
deleting-the-trace
Experiments for control-token chain-of-thought suppression and parser-leniency attacks on tool-using LLM agents

Experiments for control-token chain-of-thought suppression and parser-leniency attacks on tool-using LLM agents

How Much Can Tool Calls Reveal About You? Privacy Leakage from Tool-Call Traces

Code Implementation of "Pattern Enhanced Multi-Turn Jailbreaking: Exploiting Structural Vulnerabilities in Large Language Models"

GitLab Duo DevSecOps loop where a second AI re-attacks every AI-written fix, saving each successful bypass as a permanent antibody replayed in CI and…

Adversarial image attacks on vision-language web agents, from visual grounding to browser execution

PoC for CVE-2025-62593: unauthenticated RCE in Ray (CISA KEV). Stdlib-only Python.

Public version of the Entropica repo

nas: a zero-dependency Python CLI + MCP server for Synology DSM 7. One-command audit and fix for the C3RB3R / .L0CK3D ransomware campaign; hand it to…

Research code for adding controllable safety experts to LLM layers via two-stage upcycling training, with inference-time safety temperature routing…

Measuring and closing the text–vector integrity gap in offline RAG: margin law, vector-database audits and code-level verification.

MCP server for CVE/config remediation orchestration: SSH inventory, OSV/NVD lookups, CIS/DISA STIG-mapped Ansible playbook suggestions.

Measuring Open Privilege in Agent Defenses

Code and paper source for a NeurIPS EvoRobust study showing simple bag-of-words and metadata features beat most AI control monitors at detecting…

Code for ACL 2026 (main) paper "DeepGuard: Secure Code Generation via Multi-Layer Semantic Aggregation"

Research on persistent prompt injections via memory features in Claude, Gemini, and ChatGPT, with reproduction steps, datasets, and proofs for…

Curated index of deep learning latency, energy-latency & timing (availability) attacks and defenses — companion to the ACM Computing Surveys…

LLM safety: analyze how fine-tuning shifts refusal behavior and evaluate layer-freezing and low-rank defenses.

Knowns 0.30.0: Unauthenticated Header Injection Grants AI Agent Unrestricted Access to Host Filesystem