Exploiting Memristive Crossbar Memories as Dual-Use Security Primitives in IoT Devices

Abstract

As the internet-of-things (IoT) paradigm emerges, digital system designers are pressed with ever challenging design requirements necessitating smaller, more energy efficient systems. Such requirements for lightweight IoT devices apply first and foremost to the primary functionality of the devices themselves. However, as IoT devices and systems become more prevelant in society, designers also must include strong security measures within a very limited area and power budget. Thus, approaches to lightweight security primitives are needed to address challenges such as reliable device authentication, side-channel analysis, and memory integrity, to name a few. In this paper, we consider nanoelectronic circuit designs that provide robust security with minimal area and power overhead. We focus specifically on memristor based circuits that offer reasonable levels of security with low energy and area overhead. For example, memristive crossbars are attractive in and of themselves as candidates for dense on-chip non-volatile memory in IoT devices. We show how such memristive crossbar structures can be leveraged in the implementation of (1) crossbar-based physical unclonable functions (PUF) useful for authentication and (2) sneak path integrity checking for data stored in the memory itself. These functions also demonstrate the potential dual-use nature of memristive crossbar structures (e.g. use as a memory and as a PUF). Early analysis of the nanoelectronic security primitives considered suggest very low power operation with minimal area footprint, making these solutions reasonable candidates for providing security and a hardware root of trust in emerging IoT devices.