PDSProxy: Trusted IoT Proxies for Confidential Ad-hoc Personalization of AI Services

Abstract

Personal data stores (PDS) typically provide internal confidential processing mechanisms for personalizing AI services. However, these mechanisms cannot be easily applied when AI services are required to run outside a user’s PDS on third-party IoT devices. This paper closes this gap by presenting PDSProxy— an extension for external confidential processing on untrusted devices, newly enabling the secure transmission of personal data over hierarchically operating un-/trusted nodes. Our evaluation shows the feasibility of PDSProxy with reasonable overhead.