Seminars‎ > ‎

Information Theoretic Analysis of Inference Privacy for Sensor Data

Supriyo Chakraborty 


Time: 12 noon

Place: Faraday room 67-124, ENG IV


In this talk, we study the competing goals of utility and privacy as they arise when a user shares personal sensor data with untrusted third-party application providers. Instead of the traditional notion of identity privacy in a multi-user setting, we formulate our goals in more general terms of inferences which can be made using the data shared by a single user. We propose metrics for the utility and the privacy goals, and present preliminary results for the specific case of the problem when the shared samples are iid. For the above scenario, we derive the feasible region, provide information leakage bounds and outline constructive schemes for achieving some of those bounds.