A Highly Reliable and Unbiased PUF Based on Differential OTP Memory

Physically unclonable functions (PUF) are essential for hardware identity and security for IoT devices. The PUF consists of a multi-bit string that requires unbiased randomness. A one-time programmable memory (OTPM) uses insulator breakdown in metal-insulator-metal device to switch between insulating (“0”) and conducting (“1”) state. An OTPM based PUF produces an unbiased bit-string if exactly median breakdown voltage (
${V}_{\textsf {BD}}$
) and time are used. The exact application of median
${V}_{\textsf {BD}}$
is challenging in an integrated circuit as a conventional voltage supply tolerance specification is 10%. In this letter, we propose a
differential
OTPM-based PUF with a parallel circuit of two OTPM in series with a resistance. A high voltage stochastically produces a breakdown in one of the two OTPM first, which then reduces the voltage drop across the parallel OTPM circuit to prevent further breakdown events. The experimental results of
differential
OTPM PUF binary string generation show essentially unbiased randomness to demonstrate excellent resistance to supply voltage variation. Ideal uniqueness and reproducibility are observed. Thus, our approach enables PUF implementation with standard voltage supply tolerance and negligible post-processing.