TC: Medium: From Statistics to Circuits: Foundations for Future On-chip Fingerprints

点击翻译按钮获取中文摘要

In a world where everybody and everything is electronically connected, identity is essential to support trustworthy transactions. This project investigates novel techniques to design and implement on-chip fingerprints. Such fingerprints establish the hardware identity of an electronic system. The on-chip fingerprints are based on the existing, small and random manufacturing variations of electronic chips. Using a cross-disciplinary approach that combines recent advances in the field of statistics with those in circuit design, this project develops on-chip fingerprint structures that are optimized for stability, implementation cost, and security. Stable on-chip fingerprints are maximally sensitive to random manufacturing variations, and minimally sensitive to other environmental factors such as temperature, voltage, noise, and aging. Low-cost on-chip fingerprints are obtained by using statistical, architectural, and circuit-level techniques that maximize the amount of extracted entropy. Secure on-chip fingerprints are resistant against common attacks such as reverse engineering and model building. Thanks to its cross-disciplinary character, this project establishes a much-needed link between advanced statistical analysis and deep-submicron design for the purpose of circuit identification. This leads to better PUF designs, applicable across a wider range of applications: secure passports, anti-counterfeiting schemes, and security and trust at the endpoints. The project includes strong integration of research and education. For pre-college and entering freshman students, the project offers an introduction to trusted hardware, in the context of existing on-campus programs that involve minorities in engineering. For graduate students, the project offers a team-taught course, shared between the electrical engineering department and the statistics department.