Artificial Intelligence (AI) has been the key force in transforming our lives in the era of machine intelligence and automation. In the coming years, AI is expected to involve nearly every technology, e.g., Superintelligence (ChatGPT), AI for healthcare, manufacturing, autonomous vehicles and transportation systems, AI for agriculture, and environmental monitoring. However, AI creates new headaches for humans. For example, AI can be abused in disinformation campaigns or mishandled for harmful purposes, e.g., Deepfake, AI-empowered weapons, AI-empowered surveillance, and Cybercrime and hacking. This course will cover fundamental knowledge about AI security and attack/defense techniques on AI-empowered applications. Specifically, the introduction topics consist of (1) Basic applied AI/ML models; (2) Common threats/attacks in AI/ML (deep fake, adversarial attacks, data poisoning); (3) AI for threat hunting and attack defense; (4) AI tools for DevSecOps; (5) Security risks of superintelligence. Besides, AI techniques for solving some common tasks (e.g., checking bugs/security vulnerabilities, writing secure programs) are also introduced in this course. Finally, the principles of developing Responsible AI models to benefit humans are also discussed.
Mid-term/final exam: Open book, but no electric device is allowed.