Ph.D. Dissertation Defense: Md Tanvir Arafin
Wednesday, March 14, 2018
1146 A.V. Williams Building
301 405 3681
ANNOUNCEMENT: Ph.D. Dissertation Defense
Name : Md Tanvir Arafin
Date/ Time : Wednesday, March 14, at 8:30 am
Venue : AVW 1146(ISR)
Title : Hardware-based Authentication for the Internet of Things
Professor Gang Qu, Chair/Advisor
Dr. Dhananjay Anand
Professor Robert Newcomb
Professor Charalampos Papamanthou
Professor Yang Tao, Dean’s Representative
Abstract: Entity authentication is one of the most fundamental problems in computer security. Implementation of any authentication protocol requires the solution of several sub-problems, such as the problems regarding secret sharing, key generation, key storage and key verification. With the advent of the Internet-of-Things (IoT), authentication becomes a central concern in the security of IoT systems. Interconnected components of IoT devices normally contains sensors, actuators, relays, and processing and control equipment that are designed with the limited budget on power, cost, and area. As a result, incorporating security protocols in such resource-constrained IoT components can be challenging. To address this issue, in this dissertation, we design and develop hardware oriented lightweight protocols for the authentication of users, devices, and data. These protocols utilize physical properties of memory components, computing units, and hardware clocks on the IoT device.
Recent works on device authentication using physically unclonable functions can render the problem of entity authentication and verification based on the hardware properties tractable. Our studies reveal that non-linear characteristics of resistive memories can be useful in solving several problems regarding authentication. Therefore, in this dissertation, first, we explore the ideas of secret sharing using threshold circuits and non-volatile memory components. Inspired by the concepts of visual cryptography, we identify the promises of resistive memory based circuits in lightweight secret sharing and multi-user authentication. Furthermore, the additive and monotonic properties of non-volatile memory components can be useful in addressing the challenges of key storage. Overall, in the first part of this dissertation, we present our research on designing low-cost, non-crypto based user authentication schemes using physical properties of a resistive memory based system.
In the second part of the dissertation, we demonstrate that in computational units, the emerging voltage over-scaling (VOS)-based computing leaves a process variation dependent error signature in the approximate results. Current research works in VOS focus on reducing these errors to provide acceptable results from the computation point of view. Interestingly, with extreme VOS, these errors can also reveal significant information about the underlying physical system and random variations therein. As a result, these errors can be methodically profiled to extract information about the process variation in a computational unit. Therefore, in this dissertation, we also employ error profiling techniques along with a basic key-based authentication scheme to create lightweight device authentication protocols.
Finally, intrinsic properties of hardware clocks can provide novel ways of data fingerprinting and authentication. The clock signatures can be used for real-time authentication of electromagnetic signals where some temporal properties of the signal are known. In the last part of this dissertation, we elaborate our studies on data authentication using hardware clocks. As an example, we propose a GPS signature authentication and spoofing detection technique utilizing physical properties such as the frequency skew and drift of hardware clocks in GPS receivers.