← Back to projectsView on GitHub →
Open Source · From DARPA research to public education
AI Security Educational Series
12 comprehensive modules
Overview
A comprehensive open-source course covering the full AI attack surface: adversarial attacks on CNNs, audio attacks, LLM security (prompt injection, jailbreaking, guardrail bypass), model-level attacks (DoS, model stealing, embedding inversion), and data poisoning.
Approach
PyTorch + Docker-based interactive challenges
“Bridges PhD-era adversarial ML research with practical, hands-on security education. All modules use local models, no API costs.”
Technologies
PyTorchDockerAdversarial MLLLM SecurityRed Teaming