Dr. Aydin Aysu, North Carolina State University
Intellectual Property (IP) thefts of trained machine learning (ML) models through side-channel attacks on inference engines are becoming a major threat. Indeed, several recent works have shown reverse engineering of the model internals using such attacks, but the research on building defenses is largely unexplored. There is a critical need to efficiently and securely transform those defenses from cryptography to ML frameworks. A common defense technique is called masking, which randomizes all intermediate computations while preserving the same functionality. Although masking is well-known for cryptography its extension to ML is non-trivial. In this talk, I will explain different mechanisms to mask neural networks in hardware and describe related opportunities and challenges. I will first discuss how a straightforward masking adaptation leaks side-channel information on neural networks and how to address this vulnerability. I will then describe a fundamentally new approach that redefines neural networks to make them easier to mask in hardware.
Bookings are closed for this event.