A&S Graduate Studies
[PAST EVENT] Qing Yang, Computer Science - Ph.D. Dissertation Defense
Abstract
Smartphone users face increasing security and privacy risks. Power consumption has become a key issue for smartphone security and privacy protection. It creates two problems: power-based security attack, and energy efficiency of privacy protection. In this dissertation, we propose to exploit power for smartphone security, as well as to optimize energy consumption for smartphone privacy.
First, we show that public USB charging stations pose a significant privacy risk to smartphone users even when no data communication exists between the station and the user?s mobile device. We present a side-channel attack that allows a charging station to identify which webpages are loaded while the smartphone is charging. To evaluate this side-channel, we collected power traces of Alexa top 50 websites on multiple smartphones under several conditions, including: varied battery charging level, browser cache enabled/disabled, taps/no taps on the screen, WiFi/LTE, TLS encryption enabled/disabled, different amounts of time elapsed between collection of training and testing data, and various hosting locations of the website being visited. The results of our evaluation show that the attack is highly successful: in many settings, we were able to achieve over 90% accuracy on webpage identification. On the other hand, our experiments also show that this side-channel is sensitive to some of the aforementioned conditions.
Second, we introduce a new attack that allows a malicious charging station to identify which website is being visited by a smartphone user via Tor network. Our attack solely depends on power measurements performed while the user is charging her smartphone and does not require the adversary to observe any network traffic or to transfer data through the USB port. We evaluated the attack by training a machine learning model on power traces from 50 regular webpages and 50 Tor hidden services. We considered realistic constraints such as different Tor circuits types and battery charging levels. In our experiments, we were able to correctly identify webpages visited using the official mobile Tor browser with accuracy of up to 85.7% when the battery was fully charged, and up to 46% when the battery level was between 30% and 50%. Surprisingly, our results show that hidden services can be identified with higher accuracies than regular webpages.
Third, we propose a memory- and energy-efficient garbled circuit evaluation (MEG) mechanism on smartphones by transmitting the circuit data in bursts for secured two-party computation. We measured the energy consumption and execution time of MEG using different burst sizes, and we compared the results with existing circuit evaluation methods. The performance evaluation results show that the energy consumption of MEG is significantly lower than the current pipelining schemes and close to that of the no-pipelining method. MEG also reduces the evaluation time significantly compared to the basic pipelining. We also evaluated two versions of MEG: single-thread, and multi-thread. We compared the energy consumption and time overhead of both versions. We found that the energy consumption of multi-thread MEG and single-thread MEG are close, but the execution time using multi-thread MEG is shorter and more consistent than using single-thread MEG, especially when the burst size is large.
Bio:
Qing Yang is a Ph.D. candidate in the Department of Computer Science at William & Mary since the fall of 2011. He is working with Dr. Gang Zhou in the fields of mobile security and privacy protection. Before joining William & Mary, he received his B.S. in Computer Science from Civil Aviation University of China, and M.Eng. from Institute of Software, Chinese Academy of Sciences.