Privacy-Preserving Localization Techniques for Autonomous Vehicle Navigation Systems

Abstract

In this paper, we present several privacy-preserving techniques that can be used for vehicle localization in autonomous vehicles. These techniques use plausibly deniable encryption and randomness in initial seeds to hide the location trajectory of a vehicle while it is moving. We evaluate our techniques and demonstrate that it is possible to achieve the same level of accuracy (in terms of localization) and save power on the computing platform in the autonomous vehicle, as compared to using the basic distributed (or centralized) Monte Carlo localization algorithms. Our privacy-preserving algorithms also operate about the same amount of time as Monte Carlo localization, while the vehicle is moving on a given trajectory. We also demonstrate that an adversarial observer cannot retrieve the last known location or the true position trajectory of an autonomous vehicle that uses our privacy-preserving Monte Carlo localization algorithm while it is moving. Our work is the first published evidence that privacy-preserving vehicle localization is practical.