Sneak path enabled authentication for memristive crossbar memories

Abstract

The memristive crossbar has emerged as a promising candidate for future memory technologies. In this paper, we explore a sneak path enabled authentication method for memristive crossbar memory. We show that, due to data dependent sneak paths currents, memristive crossbar memory can inherently produce authenticating tags for stored data. A reserved row of memory cells is also used in the proposed protocol which is reconfigured randomly during every write to the memory. We perform a numerical probabilistic analysis for evaluating the security of the proposed protocol for different sizes of crossbar and generated tags. Security is measured in terms of minimum number of trials needed to get a collision among data tags. Simulation results can be extrapolated for predicting the security of the system with larger data tags. Since the memory itself with minimal additional circuitry is used as the tag generating function, the proposed protocol exhibits low overhead compared to more conventional methods.