CS4324 Adversarial and Secure Machine Learning

With machine learning being used in key components of an increasing number and variety of systems (e.g., cyber-systems and autonomous systems), the security of the whole system depends on the security of the machine learning. Adversarial and secure machine learning, the focus of this class, is the study of how to attack and defend machine learning systems. Students will learn the theory of these methods, how they are implemented in code, how to apply them, and how to evaluate their effectiveness. This course will build upon the material from CS3315.

Prerequisite

CS3315

Lecture Hours

4

Lab Hours

1

Course Learning Outcomes

• Identify, differentiate and explain different evasion, privacy, and backdoor adversarial machine learning attacks and defenses.
• Implement evasion, privacy, and backdoor adversarial machine learning attacks and defenses in Python code using a machine learning library, such as Tensorflow or Pytorch.
• Evaluate the success of adversarial machine learning attacks and defenses.