Predictive Analysis of Randomness in 3D RRAM-based Physically Unclonable Security Primitive

This paper investigates quality of randomness in the first generation of 3D analog RRAM Physically Unclonable Function (PUF) primitives using measured and gathered data from fabricated RRAM crossbars. This study is significant as the randomness quality of a PUF directly relates to its resilience against various model-building attacks, including machine learning. Experimental result verifies near perfect (50%) predictability. It confirms the PUF’s potentials for large-scale, yet small and power efficient, implementation of hardware intrinsic security primitives.