Awesome AI Security Reference

Skill by ara.so — Security Skills collection.

This skill helps you navigate the Awesome AI Security learning journey, a curated collection of resources covering ML security fundamentals, attack vectors (prompt injection, adversarial examples, poisoning, privacy attacks), defensive tools, and AI-powered penetration testing.

What This Project Provides

Awesome AI Security is a comprehensive, annotated roadmap covering:

• ML Foundations: Essential courses and books (Stanford CS229, fast.ai, Deep Learning Specialization)
• Deep Learning: Neural networks, transformers, and modern architectures
• Prompt Injection: Attacks manipulating LLM behavior and defenses
• Adversarial Attacks: Methods to fool neural networks with crafted inputs
• Poisoning & Backdoors: Data poisoning and neural network backdoors
• Privacy & Extraction: Membership inference, model extraction, data leakage
• Tools & Frameworks: Security testing tools (Garak, Rebuff, Counterfit, NeMo Guardrails)
• AI Pentesting: Using AI agents for automated security assessments
• Vulnerability Detection: AI-powered code analysis and bug detection