Search Results

This course features hands-on and project-based approaches to the understanding of the robustness and vulnerability of current state-of-the-art deep learning systems, particularly in the context of realworld security applications. Lectures will cover the theoretical foundation and algorithmic details of white/black-box adversarial attacks, data poisoning attacks, and appropriate defenses for multiple machine learning tasks, including image classification, object detection, natural language processing, graph neural networks, etc. More generally, the idea of adversarial machine learning is crucial for expanding learning capabilities, ensuring trustworthy decision-making, and enhancing the generalizability of deep learning methods.

College/Department: College of Computing and Informatics
Repeat Status: Not repeatable for credit
Prerequisites: CS 613 [Min Grade: C]