NSF Award
2219678
This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).<br/><br/>Designers of modern Integrated Circuit (IC) take advantage of pre-designed Intellectual Property (IP) components produced by third parties to reduce costs and time to market. Although this approach decreases the IC production costs, a malicious adversary could tamper with the manufacturing process by inserting Hardware Trojans (HT) into the final product. For example, according to a well-known report by an anonymous US defense contractor, hidden kill switches were found in some commercial processors. This proposal targets a new security threat originating from hardware-infected thermal sensors. Thermal sensors used for dynamic power and thermal management can be compromised by HTs, which in turn, seriously impacts the target IC's security, reliability, and performance. Addressing such a security vulnerability has positive impacts on a wide range of computing systems. The large spectrum of potential victim systems, e.g., edge computing devices, mobile smartphones, wearable systems, data centers, and supercomputers, indicates the utmost need to address this new threat.<br/> <br/>HT detection methods working at only one stage of the IC development life-cycle (e.g., design-, test-, or runtime) have major limitations. Having billions of gates in an IC makes design-time HT detection methods impractical due to scalability issues. Test-time methods rely heavily on a statistical analysis of the IC power and thermal profiles. But, they cannot detect small-size HTs due to the small footprints. Run-time methods require an always-running monitor to discriminate sensor readings, which could have large overheads. This project develops an ensemble of novel countermeasures that address the problem cooperatively at test-time and run-time to attain the required security and reliability of ICs. If test-time methods fail in detecting the HT infected sensor(s) due to a negligible footprint of an HT, then run-time methods can detect and isolate the compromised sensor(s). Furthermore, the introduced methods can accurately estimate the temperature of the spot assigned to the isolated thermal sensor so that the thermal and power management routine scan continues to function properly.<br/><br/>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.
