Abstract: We demonstrate 16kB RRAM PUF (Physically Un-clonable Function) arrays with excellent immunity to hightemperature SMT (surface mount technology) -reflow process and resilience against modern Machine-Learning (ML) based sidechannel attacks (SCA). Robust PUF operation is experimentally demonstrated on two optimized 2T-2R CMOS-RRAM designs (28 nm and 130 nm). We exploit forming voltage variability coupled with a dedicated novel programming scheme to extract unique PUF signatures. Fabricated arrays exhibit excellent performance in terms of speed, data retention and memory window. Extracted PUF signatures satisfy NIST (800-90B) tests showing extremely narrow distribution for hamming weight across multiple dies. Resilience against modern ML-SCA is achieved by introducing a secondary low energy HRS (high-resistance state) programming step and RRAM device stack-engineering. Proposed work is one of the first demonstrations for SMT-reflow immunity in context of PUF designs highlighting the high tolerance of the PUF signature to temperatures as high as 200 °C.
Loading