Reconfigurable PUFs for FPGA-based SoCs

Implementing Physically Unclonable Functions (PUFs) on FPGAs is quite inefficient in terms of resource usage. Many logic and routing resources that could serve as entropy sources remain unused. We introduce a method that uses the partial reconfiguration ability of modern FPGAs as a way to maximize the entropy that can be extracted out of a logic block. Different implementations and types of PUFs can be reprogrammed on the same logic blocks and each of their outputs used as an individual partial key. We show with a first implementation that up to six PUFs can be used on the same logic block on a Xilinx Zynq. The correlation between the PUF outputs remains very small, so that the area needed for the same length of PUF response can be shrunk up to 83%.     «

Implementing Physically Unclonable Functions (PUFs) on FPGAs is quite inefficient in terms of resource usage. Many logic and routing resources that could serve as entropy sources remain unused. We introduce a method that uses the partial reconfiguration ability of modern FPGAs as a way to maximize the entropy that can be extracted out of a logic block. Different implementations and types of PUFs can be reprogrammed on the same logic blocks and each of their outputs used as an individual partial...     »