FLAIR: Storing unbounded data streams on mobile devices to unlock user privacy at the edge

Anonymous

Keywords: mobile, storage, location, privacy

TL;DR: We propose a solution to efficiently store time series on memory-constrained devices; we demonstrate it by implementing points-of-interest search and associated protections on smartphones, to evaluate location privacy without any data sharing.

Abstract:

Mobile devices are producing larger and larger data streams, such as location streams, which are consumed by location-based services to deliver personalized content to end users. Such data streams are generally uploaded and centralized to be processed by third parties, potentially exposing sensitive personal information. In this context, existing protection mechanisms, such as Location Privacy Protection Mechanisms (LPPM), have been investigated. Alas, none of them have actually been implemented, nor deployed in real-life, in mobile devices to enforce user privacy at the edge. We believe that the effective deployment of LPPM on mobile devices faces a major challenge: the storage of unbounded data streams.
This paper introduces FLAIR, a storage system based on a new piece-wise linear approximation technique that increases the storage capacity of mobile devices by relying on data modeling. Beyond the FLAIR storage layer, we also introduce Divide & Stay, a new privacy-preserving technique to execute Points Of Interest (POI) inference. Finally, we deploy both of them on Android and iOS to demonstrate that a real deployment of LPPM is now possible.

Area: Systems for ML and ML for systems

Type: Solution

Revision?: No