Build AI-powered security tools. 50+ hands-on labs covering ML, LLMs, RAG, threat detection, DFIR, and red teaming. Includes Colab notebooks, Docker environment, and CTF challenges.

Noise Injection Techniques provides a comprehensive exploration of methods to make machine learning models more robust to real-world bad data. This repository explains and demonstrates Gaussian noise, dropout, mixup, masking, adversarial noise, and label smoothing, with intuitive explanations, theory, and practical code examples.

Comprehensive taxonomy of AI security vulnerabilities, LLM adversarial attacks, prompt injection techniques, and machine learning security research. Covers 71+ attack vectors including model poisoning, agentic AI exploits, and privacy breaches.

An application to demonstrate stealing an AI model through knowledge distillation.

🤖 Test and secure AI systems with advanced techniques for Large Language Models, including jailbreaks and automated vulnerability scanners.

Reproducible security benchmarking for the Deconvolute SDK and AI system integrity against adversarial attacks.

Bug bounty report demonstrating prompt injection and command execution vulnerabilities in Meta AI's Instagram Group Chat

Complete 90-day learning path for AI security: ML fundamentals → LLM internals → AI threats → Detection engineering. Built from first principles with NumPy implementations, Jupyter notebooks, and production-ready detection systems.

Master's students in NCCU SoSLab maintaining a cleaned and restructured version of INCITE (based on PyCT).

🛡️ Discover and analyze critical vulnerabilities in Meta AI's Instagram Group Chat, ensuring robust security through comprehensive testing and reporting.

A curated list of awesome resources for AI system security.

Adversarial ML Scanner for threat detection and ML backdoor attcaks

Adversarial Machine Learning Toolkit - Model extraction, adversarial examples, neural network probing, and defense evaluation in Julia

Security Vulnerabilities and Defensive Mechanisms in CLI/Terminal-Based Large Language Model Deployments - A Comprehensive Research Synthesis (Technical Report, November 2025)

Automatically generate YARA rules from adversarial and benign text samples. Built for detecting indirect prompt injection attacks on RAG pipelines.

An experiment in backdooring a shell safety classifier by planting a hidden trigger in its training data.

Autonomous adversarial agents that debate and debug code before you see it.

Hybrid Threat Intelligence Engine with Explainable AI (XAI) and Automated Triage.

A collection of resources documenting my research and learning journey in AI System Security.

Solves AI security research information overload. Autonomous system monitors ArXiv, filters relevant papers, and synthesizes threat intelligence using multi-agent LLM workflow. Features: two-stage filtering, agentic pipeline, forensic logging, air-gapped processing, real-time dashboard.