A Model for Quantifying Information Leakage

Abstract

We study data privacy in the context of information leakage. As more of our sensitive data gets exposed to merchants, health care providers, employers, social sites and so on, there is a higher chance that an adversary can “connect the dots” and piece together a lot of our information. The more complete the integrated information, the more our privacy is compromised. We present a model that captures this privacy loss (information leakage) relative to a target person, on a continuous scale from 0 (no information about the target is known by the adversary) to 1 (adversary knows everything about the target). The model takes into account the confidence the adversary has for the gathered information (leakage is less if the adversary is not confident), as well as incorrect information (leakage is less if the gathered information does not match the target’s). We compare our information leakage model with existing privacy models, and we propose several interesting problems that can be formulated with our model. We also propose efficient algorithms for computing information leakage and evaluate their performance and scalability.