everything.study
E C E 763

TRUSTWORTHY ARTIFICIAL INTELLIGENCE

3 credits.

Explore security and privacy aspects of trustworthy artificial intelligence. Three core subjects will be considered: differential privacy and algorithmic fairness; adversarial machine learning; and end-to-end trustworthy systems. A selection of more advanced topics may be covered such as additional notions of privacy, language-based security, and robust optimization. Knowledge of probability/statistics (such as MATH 431), cryptography (such as MATH 435), security (such as COMP SCI 642 ), and modern machine learning (such as M E/COMP SCI/E C E 539 or 540 ) is required.

Prerequisites & course information

Requisites: Graduate/professional standing Course Designation: Grad 50% - Counts toward 50% graduate coursework requirement Repeatable for Credit: No Last Taught: Spring 2026 Learning Outcomes: 1. Evaluate machine learning and AI systems from an adversarial, security and privacy mindset. Audience: Graduate 2. Identify common pitfalls and problems in ensuring security and privacy for AI. Audience: Graduate 3. Summarize the commonalities and differences between notions of security and privacy (e.g., the difference between privacy and cryptographic security). Audience: Graduate 4. Explain the strengths and limitations of candidate definitions of robustness, security, privacy, and fairness properties in AI. Audience: Graduate 5. Apply useful primitives from end-to-end trustworthiness to machine learning and AI systems. Audience: Graduate 6. Use modern tools to design attacks and implement defensive measures. Audience: Graduate Learning Outcomes: 1. Evaluate machine learning and AI systems from an adversarial, security and privacy mindset. Audience: Graduate 2. Identify common pitfalls and problems in ensuring security and privacy for AI. Audience: Graduate 3. Summarize the commonalities and differences between notions of security and privacy (e.g., the difference between privacy and cryptographic security). Audience: Graduate 4. Explain the strengths and limitations of candidate definitions of robustness, security, privacy, and fairness properties in AI. Audience: Graduate 5. Apply useful primitives from end-to-end trustworthiness to machine learning and AI systems. Audience: Graduate 6. Use modern tools to design attacks and implement defensive measures. Audience: Graduate

Sources
Catalogue variants