Verification of Location and Time of Evidential In-Vehicle Dashcam Footage with Anonymity Guarantees

Abstract

This thesis considers the following problem: ‘The police investigating car accidents (i.e., past events) now want to automatically collect and verify locations and times of dashcam videos (DCVs) from vehicles that were near the accidents, while users want to protect their location privacy.’ This is a conflicting condition where existing localization verification techniques cannot ensure such privacy while location privacy techniques are not applicable for verification. In this thesis we present ViewLink, an automated process that verifies whether dashcam videos (DCVs) submitted from anonymous users were truly taken near certain traffic accidents (i.e., containing evidential footage) before actual human review. ViewLink takes a data-oriented approach that only treats DCVs themselves as verification entities instead of user identities. ViewLink instantiates the data-only verification by leveraging dedicated short-range communications (DSRC) broadcast. We demonstrate the feasibility of ViewLink via field experiments on real roads using our DSRC testbed as well as simulations using traffic traces.