XRay: Enhancing the Web's Transparency with Differential Correlation

Abstract

Today's Web services - such as Google, Amazon, and Facebook - leverage user\ndata for varied purposes, including personalizing recommendations, targeting\nadvertisements, and adjusting prices. At present, users have little insight\ninto how their data is being used. Hence, they cannot make informed choices\nabout the services they choose. To increase transparency, we developed XRay,\nthe first fine-grained, robust, and scalable personal data tracking system for\nthe Web. XRay predicts which data in an arbitrary Web account (such as emails,\nsearches, or viewed products) is being used to target which outputs (such as\nads, recommended products, or prices). XRay's core functions are service\nagnostic and easy to instantiate for new services, and they can track data\nwithin and across services. To make predictions independent of the audited\nservice, XRay relies on the following insight: by comparing outputs from\ndifferent accounts with similar, but not identical, subsets of data, one can\npinpoint targeting through correlation. We show both theoretically, and through\nexperiments on Gmail, Amazon, and YouTube, that XRay achieves high precision\nand recall by correlating data from a surprisingly small number of extra\naccounts.\n