Machine learning has enabled dramatic advances in many areas, including cybersecurity. But, it also raises important security and privacy concerns. Malicious actors can fool machine learning algorithms. Attackers can poison an entire training process by corrupting one item in a large data set. Models can leak the underlying training data.
In this webinar, Professor Dan Boneh will discuss recent work at the intersection of cybersecurity and machine learning. Specifically, he will explore an area known as “adversarial machine learning” which looks at the stability of machine learning models in the presence of adversarial behavior.
- What recent research on adversarial behavior tells us about machine learning models
- How to protect classification and training processes from attacks
- Ways to insure the privacy of underlying training data
- Item 4