Privacy Analysis of Online and Offline Systems
Abstract
How to protect people's privacy when our life are banded together with smart devices online and offline? For offline systems like smartphones, we often have a passcode to prevent others accessing to our personal data. Shoulder-surfing attacks to predict the passcode by humans are shown to not be accurate. We thus propose an automated algorithm to accurately predict the passcode entered by a victim on her smartphone by recording the video. Our proposed algorithm is able to predict over 92% of numbers entered in fewer than 75 seconds with training performed once. For online systems like surfing on Internet, anonymous communications networks like Tor can help encrypting the traffic data to reduce the possibility of losing our privacy. Each Tor client telescopically builds a circuit by choosing three Tor relays and then uses that circuit to connect to a server. The Tor relay selection algorithm makes sure that no two relays with the same /16 IP address or Autonomous System (AS) are chosen. Our objective is to determine the popularity of Tor relays when building circuits. With over 44 vantage points and over 145,000 circuits built, we found that some Tor relays are chosen more often than others. Although a completely balanced selection algorithm is not possible, analysis of our dataset shows that some Tor relays are over 3 times more likely to be chosen than others. An adversary could potentially eavesdrop or correlate more Tor traffic. Further more, the effectiveness of website fingerprinting (WF) has been shown to have an accuracy of over 90% when using Tor as the anonymity network. The common assumption in previous work is that a victim is visiting one website at a time and has access to the complete network trace of that website. Our main concern about website fingerprinting is its practicality. Victims could visit another website in the middle of visiting one website (overlapping visits). Or an adversary may only get an incomplete network traffic trace. When two website visits are overlapping, the website fingerprinting accuracy falls dramatically. Using our proposed "sectioning" algorithm, the accuracy for predicting the website in overlapping visits improves from 22.80% to 70%. When part of the network trace is missing (either the beginning or the end), the accuracy when using our sectioning algorithm increases from 20% to over 60%.
Collections
- OSU Dissertations [11222]