
Visco-Attack
Research pipeline implementing a visual-contextual jailbreak against multimodal LLMs: T2I prompt extraction, SD3.5 image rendering, multi-round spoof…

Research pipeline implementing a visual-contextual jailbreak against multimodal LLMs: T2I prompt extraction, SD3.5 image rendering, multi-round spoof…

Curated catalog of 2026 command-and-control frameworks for red team operations, with per-tool breakdowns of implants, listeners, payloads, and…

Reproduction harness, patches and analysis tools for auditing a learned cyber-defense benchmark: captures agent traces, runs red-agent ablations, and…

Benchmark harness that evaluates LLM agent tool-calling security under long-horizon tasks, generating prompt-injection and tool-poisoning attacks and…

A post exploitation framework / C2 for advanced Red Team operations

Local test range for agent memory poisoning - watch a planted memory steer an agent, then prove the repair. Synthetic targets, scripted deterministic…

Proof-of-concept and full writeup for CVE-2026-47644: Edge Copilot prompt injection via attachment tag breakout leading to cross-tab data…

Deterministic benchmark and evaluation harness for persistent-memory poisoning in LLM agents, with paired clean/poisoned cases, LangChain/LlamaIndex…

Code for "Systematic Evaluation of Adversarial Jailbreaking Attacks and Defenses Across Seven Large Language Models"

Benchmark and library for evaluating model-extraction attacks and ownership defenses (watermarking, output perturbation, query detection) on graph…

Beyond the Prompt: Jailbreaking Function-Calling LLMs via Simulated Moderation Traces

PyTorch reference implementation of a fast adversarial training method that prevents catastrophic overfitting via coupled objectives and an EMA-based…

Research framework for detecting and removing backdoors in language models, combining feature-based anomaly detection with LoRA-based unlearning…

Reversible HTML and JavaScript prompt decoys with opaque per-insertion keys.

Code and results for the QAI 2026 paper on FGSM attack transferability from simulation to IBM quantum hardware.

Lightweight evaluation harness for RAG poisoning defenses that builds four-condition clean/poisoned records and reports prevented targets, net…

Hardens binaries, Electron ASAR, JS, WASM, Python, and model files against AI reverse-engineering agents using decoy metadata, budget exhaustion, and…

SQLite relational data model for MITRE ATT&CK plus an adversary emulation planning tool, enriched with Red Canary ATOMIC tests and APT3 emulation…