Cooperative Location Privacy in Vehicular Networks: Why Simple Mix-zones are not Enough

TL;DR

This paper demonstrates that simple cryptographic mix-zones are insufficient for vehicular location privacy and introduces a cooperative scheme with relaying vehicles to significantly improve pseudonym unlinkability with minimal overhead.

Contribution

The paper proposes a novel cooperative mix-zone scheme with relaying vehicles that enhances location privacy against inference attacks in vehicular networks.

Findings

Linkability of pseudonyms is high with standard mix-zones, especially during non-rush hours.

The cooperative scheme reduces pseudonym linking probability from 68% to 18%.

The scheme incurs low additional computation and communication overhead.

Abstract

Vehicular communications disclose rich information about the vehicles and their whereabouts. Pseudonymous authentication secures communication while enhancing user privacy. To enhance location privacy, cryptographic mix-zones were proposed to facilitate vehicles covertly transition to new ephemeral credentials. The resilience to (syntactic and semantic) pseudonym linking (attacks) highly depends on the geometry of the mix-zones, mobility patterns, vehicle density, and arrival rates. We introduce a tracking algorithm for linking pseudonyms before and after a cryptographically protected mix-zone. Our experimental results show that an eavesdropper, leveraging standardized vehicular communication messages and road layout, could successfully link 73% of pseudonyms during non-rush hours and 62% of pseudonyms during rush hours after vehicles change their pseudonyms in a mix-zone. To mitigate such inference attacks, we present a novel cooperative mix-zone scheme that enhances user privacy regardless of the vehicle mobility patterns, vehicle density, and arrival rate to the mix-zone. A subset of vehicles, termed relaying vehicles, are selected to be responsible for emulating non-existing vehicles. Such vehicles cooperatively disseminate decoy traffic without affecting safety-critical operations: with 50% of vehicles as relaying vehicles, the probability of linking pseudonyms (for the entire interval) drops from 68% to 18%. On average, this imposes 28 ms extra computation overhead, per second, on the Roadside Units (RSUs) and 4.67 ms extra computation overhead, per second, on the (relaying) vehicle side; it also introduces 1.46 KB/sec extra communication overhead by (relaying) vehicles and 45 KB/sec by RSUs for the dissemination of decoy traffic. Thus, user privacy is enhanced at the cost of low computation and communication overhead.