COSC 89.33 The Dark Side of AI/ML: Machine Learning Security, Privacy, Fairness, and Interpretability
The goal of this course is to equip students to responsibly deploy machine learning tools in an unfair and adversarial world. We will survey the vulnerabilities of mainstream machine learning models and algorithms to manipulation, privacy leakage, and unfairness. We will then assess the conditions under which we can understand, interpret, or measure our models’ predictions with respect to each of these vulnerabilities. Finally, we will address the feasibility and limitations of appropriate countermeasures. Our course will build towards a final project assignment where students are given the opportunity to develop their own research projects in this emerging field.
Prerequisite
Before taking this course, all students are expected to have completed an advanced undergraduate machine learning course such as COSC 74, 78, 83 or the graduate course equivalent.