A Word Selection Algorithm for Reliable Enhanced Latency-Based DRAM Physical Unclonable Function

Dynamic Random-Access Memory Physical Unclonable Functions (DRAM PUFs) are an interesting and emerging area in hardware security, especially for the ubiquitous Internet of Things (IoT) devices in the era of rapid globalization. Since DRAM is present in most of the embedded devices, implementing DRAM PUFs does not require a lot of additional hardware components. In addition, DRAM consumes less power than SRAM and DRAM's size is generally larger, which leads to larger challenge-response pair (CRP) space, enhancing security and robustness. This paper presents a word-based PUF generation for the DRAM Latency PUF. It demonstrates that constructing a PUF using words instead of individual bits is not only feasible but also maintains performance across key metrics such as uniqueness, reliability, and randomness. Additionally, generating words significantly reduces processing time and storage cost compared to extracting PUFs bit by bit. The experimental results show that the proposed PUFs achieved ∼50% bit flips indicating good randomness and intra Hamming distance as low as 0-2 bits indicating a good stability while NIST tests results reflect that the proposed PUFs are sufficiently robust in randomness and bias.