Events Calendar

Security and Privacy of Non-Volatile Memories- Vunerabilities, Attack Models and Preventions

Speaker: Swaroop Ghosh, Professor, University of South Florida

Talk Title: Security and Privacy of Non-Volatile Memories- Vulnerabilities, Attack Models and Preventions

Abstract: Non-volatile memories (NVM) such as Spin-Transfer Torque RAM (STTRAM), Resistive RAM and Domain Wall Memory have drawn significant attention due to complete elimination of bitcell leakage. In addition to plethora of benefits such as density, non-volatility, low-power and high-speed, majority of NVMs are also compatible with CMOS technology enabling easy integration. NVMs are particularly interesting for a class of Internet-of-Things (IoT) that are normally OFF but require instant ON experience. Although promising, I will show that NVMs bring new security and privacy challenges that were absent in their conventional volatile memory counterparts. Assuring data integrity and privacy against malicious attacks is particularly critical on deployed systems that are hard to maintain and enforce physical security. I will present two aspects to NVM security in Last Level Cache (LLC) using STTRAM as test case:
(i) Data integrity which pertains to data corruption by malicious attack with the intention to launch denial-of-service. Such attacks exploit the fact that NVMs are fundamentally susceptible to ambient parameters such as magnetic field and temperature. I will describe these vulnerabilities and attack models, and, propose two micro-architectural techniques to assure data integrity under attack namely, cache bypassing and checkpointing. These techniques allow seamless computation in presence of attack at minimal design overhead.
(ii) Data privacy which pertains to sensitive data such as keys and passwords being compromised. Storage such as Hard Disk Drive (HDD) has been the non-volatile part of memory system traditionally protected by encryption. Although effective, the latency associated with encryption makes it non-trivial for application in higher levels of memory stack such as LLC. I will present the vulnerabilities and attack models, and, propose two low-overhead techniques to maintain data privacy namely, Semi Non-Volatile Memory which is similar to NVM but with very low retention time so that the data vanishes after power is turned OFF, and, irreversible erasure of data at power down using residual charge from power rail.

Biography: Swaroop Ghosh (S'04, SM'13) received his B.E. (Hons.) from IIT, Roorkee (2000), M.S. from University of Cincinnati (2004) and Ph.D. from Purdue University (2008). He joined USF in Fall 2012. Dr. Ghosh was senior research and development engineer in Advanced Design, Intel Corp from 2008 to 2012 where pioneered 32nm and 22nm SRAM and eDRAM designs. His research interests lie at the intersection of circuits, micro-architecture and hardware security. He is a senior member of IEEE.
Dr. Ghosh is serving as Associate Editor of IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS-I and Senior Editorial Board member of IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS. He has served in the technical program committees of DAC, DATE, ICCAD, ISLPED, HOST, Nanoarch, VLSI Design, ISQED, ASQED, and VLSI-SOC. He is a recipient of DARPA Young Faculty Award (2015), ACM SIGDA Outstanding New Faculty Award (2016), USF Outstanding Research Achievement Award (2015) and USF College of Engineering Outstanding Research Achievement Award (2015).

Host: Professor Murali Annavaram

Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248

Audiences: Everyone Is Invited

Contact: Suzanne Wong