Patrick Schaumont is working with Leyla Nazhandali to create physical unclonable functions (PUFs) based on the random variations that occur during chip manufacturing. Read more in “On-chip fingerprints.”
Unlocking the secrets of the Spartan boards
Researchers in the Secure Embedded Systems Laboratory are capitalizing on the Spartan 3E development kits used by all ECE undergraduates to make a groundbreaking study of how variances and the submicron level can affect computations.
The Spartan boards have an FPGA chip that varies slightly when compared to the same model chip on another board. Manufacturers have been aware of such die-to-die variations for more than a decade, and designers have been investigating techniques that take such variations into account during the design process, according to Patrick Schaumont, who leads an effort to investigate circuit variability from die-to-die, as well as within each chip.
As semiconductor process technology reaches ever smaller feature sizes, the impact of within-die variations has become significant, and that has begun to attract research attention, he says.
Schaumont’s team is studying the population of more than 250 Spartan boards owned by ECE students. Previous research has studied at most 36 FPGAs, limiting the precision and detail needed to accurately characterize a large population of deployed chips. The team also hopes to show that it is feasible to recognize each individual chip from the entire population, which has important applications in secure hardware design.
To encourage students to bring in their Spartan boards for testing, the research team is running a contest to give away a game console for every 50 students who come in. The group recently awarded a Nintendo Wii to the first winner, CPE student Jingyao Zhang. The next winner will receive a Sony Playstation 3.
The project is funded by a $110,000 NSF grant, and will run until 2012. Schaumont is spending one year designing and implementing experiments to measure die-to-die variability, the second year creating experiments to measure variability within the die, and the final grant year on improving the experimental method and analyzing the collected data. The anonymized, collected data will be made available to the research community for further investigation of process manufacturing variations.