Digital Encryption is typically performed by specialized circuits to ensure confidentiality and integrity of data. While encryption is mathematically robust, the circuits encrypting data may leak information via the amount of the power drawn from the supply, and the amount of electromagnetic (EM) radiation that emanates from the circuit. This is known as side-channel leakage. An attacker may be able to unravel the secret cryptographic information by analyzing the side-channel leakage, thereby compromising security. Newer analysis techniques based on machine-learning make the attack easier. This proposal will study how these attacks are performed to develop means of protection against such attacks. Machine learning (ML) based side-channel attack (SCA) increases the attack surface of secure hardware, as an attacker can potentially compromise a device using a few power and EM traces. This proposal will provide a comprehensive analysis for new attack models and countermeasures through: (1) analysis and development of the best possible Deep Learning based SCA attack (on power and EM). (2) Design and demonstrate low-overhead countermeasures to enable "critical" crypto signature attenuation and reduce the signal-to-noise ratio by a factor of 500. The goal of this proposal is to develop end-to-end models to build defense mechanisms for both protected and unprotected Advanced Encryption Standard implementations. Results from this project will be disseminated through papers and articles, which will be made publicly available. Results from this project will be incorporated into the courses taught by the investigators. Investigators will seek to work with undergraduate students providing hands-on experience on cryptography and side-channel attacks and analysis. The data generated from this project will be in the form of simulation results and models, software tools and hardware measurements. The developed models and benchmarking software will be uploaded to GitHub at: https://github.com/anupamgolder/mlsca This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
EAGER: Collaborative: Machine-Learning based Side-Channel Attack and Hardware Countermeasures