PrivacyProxy: Leveraging Crowdsourcing and In Situ Traffic Analysis to Detect and Mitigate Information Leakage

Smartphone apps often transmit personally identifiable information (PII) without the user's knowledge. To address this issue, we present PrivacyProxy, a system that monitors outbound network traffic and generates app-specific signatures to represent sensitive data being shared, all without any modifications to the OS. We use a crowdsourcing based approach to detect likely PII in an adaptive and scalable manner by anonymously combining signatures from different users of the same app. Our system design is itself privacy sensitive as we do not observe users' network traffic and instead rely on cryptographically hashed signatures. We present the design and implementation of PrivacyProxy and evaluate its effectiveness in detecting various PII through a lab study and a field deployment. Our field study shows that even without any user labeling of PII, PrivacyProxy automatically detects various forms of PII with a precision of 81%. PrivacyProxy also achieves a precision of 85.3% in our controlled experiment using 300 apps. We also show that the performance overhead of PrivacyProxy is under 13% and majority of the users report no perceptible impact on battery life or the network.