RPUF: Physical Unclonable Function with Randomized Challenge to resist modeling attack

Abstract

The Physical Unclonable Function (PUF) has broad application prospects in the field of hardware security. The strong PUFs with numerous Challenge-Response Pairs (CRPs), such as various arbiter PUFs, mirror current PUF, and voltage transfer PUF, are severely threatened by the machine learning based modeling attacks. To handle this issue, we propose the Physical Unclonable Function with Randomized challenge (RPUF). Challenges are randomized by a Random Number Generator (RNG) before inputting to the strong PUF, so to prevent attackers from collecting effective training set for conducting modeling attacks. Experiments on both simulations and FPGAs prove the effectiveness of RPUF in resisting modeling attack, with negligible effects on uniformity, uniqueness, and reliability.