Association for Computing Machinery (ACM), ACM Transactions on Embedded Computing Systems, 1(15), p. 1-20, 2016
DOI: 10.1145/2815621
Full text: Unavailable
Physically Unclonable Function (PUF) circuits are often vulnerable to mathematical model-building attacks . We theoretically quantify the advantage provided to an adversary by any training dataset expansion technique along the lines of security analysis of cryptographic hash functions. We present an algorithm to enumerate certain sets of delay constraints for the widely studied Arbiter PUF (APUF) circuit, then demonstrate how these delay constraints can be utilized to expand the set of known Challenge--Response Pairs (CRPs), thus facilitating model-building attacks. We provide experimental results for Field Programmable Gate Array (FPGA)--based APUF to establish the effectiveness of the proposed attack.