Ubiquitous ML and Security: Can they co-exist?
TimeTuesday, July 12th1:30pm - 3pm PDT
Location3003, Level 3
Event Type
Special Session (Research)
DescriptionML fundamentally relies on access to data. Models can be reverse-engineered to reveal private
data. Corruption of data can lead to catastrophic results, yet data input to ML models often comes from untrusted sources. How can we overcome these seemingly contradictory requirements of ML and Security? Speakers in this session will outline recent research directions that can address this conundrum. Topics covered will include computation on private encrypted data, Secure federated learning, and secure deployment of distributed analytics.