Lightweight (Reverse) Fuzzy Extractor with Multiple Referenced PUF Responses

The physical unclonable function (PUF), alike fingerprint of human being, exploits manufacturing randomness to bind each physical item with a unique identifier. One primary PUF application is for securely cryptographic key derivation via fuzzy extractor that comprises two sequential procedures: error correction and entropy extractor. Despite the entropy extractor can be very lightweight, the overhead of the error correction logic responsible for reconciling natural fuzzy PUF responses is, nonetheless, very expensive. This runs into hardness when the PUF targets to secure resource-constraint Internet of Things (IoT) objects e.g., with limited computational power and battery lifetime. In this work, we recognize that there is an approximately linear relationship between the response unreliability and operating condition (e.g., voltage and temperature) under which the PUF works. For the first time, we exploit such an {\it important} but {\it inadvertent} fact. Instead of enrolling only a single response under the nominal operating condition during the PUF key provisioning phase, we propose to lodge multiple referenced responses (MRR), subject to the same challenge, produced under multiple discrete operating conditions. As an immediate application, we combine MRR with the reverse fuzzy extractor (RFE) to realize a MRR based RFE (MR$^3$FE) that fits lightweight mutual authentication well attributing to the greatly reduced overhead. To examine the MRR's generalization, it is adopted to the FE case, termed as MR$^2$FE. Software implementation of both MR$^3$FE and MR$^2$FE is comprehensively carried on the computational radio frequency identification (CRFID) device that is battery-less and resource-restrict.