Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
This 15-minute talk by Rohit Khera, Distinguished Engineer in Cryptography Engineering at Marvell, explores the intersection of privacy and artificial intelligence through computational cryptography. Discover how universal computation on encrypted data addresses privacy concerns for AI model inputs and outputs during both training and inference phases. Examine state-of-the-art cryptographic methods including MultiParty Computation, Threshold Cryptography, and Fully Homomorphic Encryption techniques that enable private AI systems. Learn about the specific challenges faced when implementing powerful transformer models in the encrypted domain, particularly with operations like softmax and self-attention. Gain insights into how NP-Hard problems form the foundation for privacy-preserving AI computation in this Open Compute Project presentation.
