| With the proliferation of mobile devices and mobile services, preserving user privacy has become a great challenge. While other studies directly focus on user information leakage from particular services (social networks or advertising), we call attention to the privacy leakage in mobile network traffic. This concern is motivated by two factors. First, nowadays people can access personal online services from fingertips, hence more sensitive information can and will be left in the mobile traffic. Second, traffic observers, who are able to access user traffic, can connect user traffic to user real-world identities because of the personalized online services such as social networks. These factors pose a serious threat to user privacy as they enable an adversary to attribute significant portions of data traffic including the ones with NO identity leaks to network users' real-world identities.;In this thesis, I focus on several aspects related to user identifiable information and user attributes from mobile traffic in design and functioning of network systems. First, I propose the Mosaic system along with the Tessellation methodology to demonstrate the feasibility from the traffic observer side. The methodology shows that up to 50% of the traffic can be attributed to the names of users. Second, I propose and design GeoEcho to demonstrate the potential value of user traffic. GeoEcho extracts and analyses a wealth of latitude-longitude geotag reports from user mobile traffic to uncover real-world locational interests. Third, I propose Messup to protect mobile users from traffic observers. Messup is a client-side universal solution for protecting user identities from potential traffic observers.;One of our key contributions is the demonstration of vulnerability of user privacy from user mobile traffic. More specifically, by applying Tessellation to analyze user mobile traffic, we find that: ( i) traffic observer can automatically bring together isolated user information to their owners, even for traffic with NO obvious identity leaks. (ii) The combination of information from both network and web profiling reveals far greater knowledge of users than what can be obtained individually. Another key contribution of this thesis is, I propose and evaluate the Messup system to protect user privacy from traffic observers, which is a universal client-side privacy-preserving solution. Messup automates the detection of leaking user identifiable information from mobile apps. The traffic analysis results from top Android apps suggest most Android apps have the leaking user ID issue, and over 50% HTTP traffic can be attribute to the owners. To protect user privacy, Messup generates synthetic requests to poison the traffic collected by the traffic observers, while without disturbing normal user activities.;We believe our work show noteworthy promise for further user privacy protection, clearing the way for future advances in understanding user identifiable information and user attributes, mobile application traffic, mobile social networking, mobile advertising, etc. |
