Low-cost anti-modeling attack PUF circuit based on configurable APUF

A new low-cost configurable P-2APUF structure is proposed in this paper to solve the predictability problem of Physical Unclonable Functions (PUFs) under machine learning attacks. The structure is composed of Arbitrated PUF (APUF) and Pseudorandom generator (PRNG). By changing the configuration of incentive bits, one APUF generates two different PUF structures, and combines incentive expansion mechanism and output obfuscation mechanism to achieve anti modeling attacks. FPGA experimental results show that the circuit is predicted to be about 52.16% effective against modeling attacks with 20,000 sample sizes using 128 LUTs (look-up tables) and 5 DFFs (D-type flip-flops), and up to 94.31% reliable under different noise environments. Therefore, the P-2APUF structure maintains low overhead and high reliability while achieving resistance to modeling attacks, providing a reliable and secure solution for device authentication and key generation.