Securing AI Systems

The explosive growth of AI systems has created new cybersecurity challenges. Systems that work well on benign prompts can be subverted with adversarial inputs that are easily disguised. The ensuing set of security problems gave rise to a new area of research, called AI security, that now spans thousands of research papers per year in multiple academic conferences. This course will introduce students to the key challenges of AI security, including modern approaches for securing computer systems that rely on AI models. The course content will evolve from year to year based on the latest research in the area.