ML-Based Security Analysis and Mitigation of Homomorphic Encryption Side-Channels, CAEML Core Project 5A4
Power-based side-channel attacks are a major threat to cryptographic systems. This research evaluates the vulnerability of next-generation homomorphic encryption systems to such attacks and the efficacy of state-of-the-art machine-learning classifiers to carry them out. Results suggest that ML classifiers can be used to extract secret-key information from NC State's homomorphic encryption system, as evidenced by its intrinsic correlation with the device's power consumption.
Sponsor
University of Illinois - Urbana-Champaign
The grant—running from January 1, 2021 to July 31, 2022—is for a total of $51,000.
Principle Investigators
Aydin Aysu
Paul D. Franzon
More Details
Power-based side-channel attacks are a fundamental threat for cryptographic systems because they can extract secret-key information from its intrinsic correlation to the power consumption of the underlying device. This research analyzes the vulnerability of next-generation cryptographic systems, called homomorphic encryption, against power-based side-channel attacks and explores the efficiency of the state-of-the-art machine-learning (ML) classifiers to carry out the attack.