Physical Unclonable Function With Enlarged String Current Variation in NAND Flash Array by WL Selection Scheme

In this work, we present a nand flash-based physical unclonable function (PUF) system. This system utilizes the intrinsic device variations of nand flash cells as an entropy source, effectively addressing the limited variation distribution inherent in conventional nand flash due to its serially connected structure. We experimentally validated the enhancement in the randomness of the entropy source using a fabricated $32\times 32$ nand flash array. The variation in bitline current is analyzed as a function of the number of selected wordlines ( ${N} _{\text {WL}}$ ), and a large variation is achieved by limiting ${N} _{\text {WL}}$ . Based on the measured data, we confirm that the bit-error rate (BER) of the 3-D nand flash based PUF system exhibits a negligible increase of just 0.073% per $10~^{\circ }$ C, demonstrating its high stability against temperature variation. Additionally, we evaluate the randomness metrics of the PUF system, such as uniformity, diffuseness, and uniqueness, by utilizing the challenge-response pair (CRP) space. The security of the PUF system is also assessed against machine learning (ML) attacks.