Adaptive characterization and emulation of delay-based physical unclonable functions using statistical models

It is commonly known that physical unclonable functions (PUFs) are hard to predict and hard to emulate. However, in this paper, we propose to use statistical models to adaptively characterize the delay-based PUFs, and use this as a starting point to emulate a delay-based PUF. The essential idea is that for any challenge C_A of a delay-based PUF A, there is a high probability of finding a paired challenge C_B. When apply C_B to another delay-based PUF B, it can produce the same output as applying C_A on PUF A. Our simulation results indicate more than 99% correctness for the PUF response prediction using characterization and 96% correctness using emulation. Finally, we implement and test the feasibility of our approach on the Xilinx Spartan-6 Field Programmable Gate Array (FPGA).