Frontiers in Machine Learning: Security and Machine Learning

Machine learning has enabled many advances in processing visual, language, and other digital data signals and, as a result, is quickly becoming integrated in a variety of real-world systems with important societal and business purposes. However, as with any computer technology deployed at scale or in critical domains, ML systems face motivated adversaries who might wish to cause undesired behavior or violate security restrictions. In this session, participants will discuss the security challenges of today’s AI-driven systems and opportunities to mitigate adversarial attacks for more robust systems.

Session Lead: Emre Kiciman, Microsoft

Speaker: Aleksander Mądry, Massachusetts Institute of Technology
Talk Title: What Do Our Models Learn?

Speaker: Dawn Song, University of California, Berkeley
Talk Title: AI & Security: Challenges, Lessons & Future Directions

Speaker: Jerry Li, Microsoft
Talk Title: Algorithmic Aspects of Secure Machine Learning

Q&A panel with all 3 speakers

Date:
Haut-parleurs:
Aleksander Mądry, Dawn Song, Jerry Li
Affiliation:
Massachusetts Institute of Technology, University of California Berkeley, Microsoft Research