Conventional volatile memories such as static random-access memory (static RAM) suffer from significant leakage power whereas conventional storage class non-volatile memories (NVM) such as flash memory suffer from higher write energy, poor performance and low endurance. Emerging NVMs such as, spin-transfer torque RAM (STTRAM) and resistive RAM (ReRAM) offer zero leakage, high-density, scalability and high endurance. Due to these promising aspects, emerging NVMs are already being commercialized by several companies. Although NVMs can reap energy and performance benefits they may face new security challenges. This research will expose and address the security and privacy threats to NVMs. This research will (i) identify the vulnerabilities of NVMs by understanding the interaction between device properties and ambient parameters, (ii) identify integrity threats such as tampering and row-hammering, (iii) propose methods of real-time attack sensing, (iv) develop circuits and system-level techniques such as nano-shielding and memory access control to ensure data integrity, (v) identify privacy threats such as unauthorized access and fault injection attack to extract secret data, and (vi) propose quantized memory, intelligent erasure and camouflaging of side channels to ensure data privacy. The success of this research may lead to wider adoption of NVMs in security-critical sectors such as healthcare and banking. With growing acceptance of emerging NVMs in automobiles and solid state drives this work may have more immediate benefits, such as enabling on-chip, NVM-enabled, low-power devices, as well as spreading cybersecurity literacy.
SaTC: STARSS: Small: Assuring Security and Privacy of Emerging Non-Volatile Memories