Maintaining Privacy and Utility in IoT System Analytics

Privacy risks can hinder the future of IoT applications. Users should know and trust that they are only sharing or exposing specific portions of their data that will be used for data analysis on cloud services. However, this is not a trivial problem, given that any IoT sensor can potentially capture much more information than the original purpose. This data can potentially contain unknown micro-signatures of sensitive information that can be mined by attackers using highly available and powerful AI tools. Indiscriminately sharing all data for remote storage or processing is risky. In this paper, we share our vision of how IoT systems can be built to preserve both privacy and utility. We show that some unique open challenges and opportunities arise in designing utility-preserving privacy techniques in resource-constrained IoT devices.