True Random Numbers from Pairs of SRAM Devices

Static random-access memory (SRAM) physically unclonable functions (PUFs) have been used to design true random number generators capable of passing the National Institute of Standards and Technology (NIST) statistical test suite for randomness. These SRAM PUF-based true random number generators (TRNGs) rely on unstable cells that randomly awake as a ‘0’ or a ‘1’ when the power is cycled. The three issues with the previously proposed SRAM TRNGs that we aim to address are the limited number of challenge-response pairs (CRPs), the overhead required to select “flaky” cells, and the restriction of a device to support either TRNG or authentication, but not both. We present an SRAM PUF-based TRNG that passes NIST randomness tests designed with two SRAM devices and a lightweight post-processing method to enhance randomness. This design benefits from a much higher number of CRPs, does not need to identify unstable cells, and supports authentication and true random number generation. The work presented uses commercially available development boards and SRAM devices, a simple custom circuit board to connect the two, lightweight post-process, and a combination of C++ and python3 to drive the system.