Nowhere to hide? Mix-Zones for Private Pseudonym Change using Chaff Vehicles

摘要

In vehicular communication systems, cooperative awareness messages provide contextual information required for transportation safety and efficiency applications. However, without the appropriate design, these messages introduce a new attack vector to compromise passenger privacy. The use of ephemeral credentials - pseudonyms - was therefore proposed, essentially to split a journey into unlinkable segments. To protect segment transitions, encrypted mix-zones provide regions where vehicles can covertly change their pseudonyms. While previous work focused on the placement, shape, and protocols for mix-zones, attacks that correlate vehicles entering and existing these zones still remain a problem. Furthermore, existing schemes have only considered homogeneous traffic, disregarding variations in vehicle density due to differences in driver population, road layout, and time of day. Without realistic experimental results, any conclusion on real-world applicability is precarious. In this paper, we address this challenge and present a novel scheme that works independent of vehicles' mobility patterns. More precisely, our system generates fictive chaff vehicles when needed and broadcasts their traces, while it remains unobtrusive if sufficiently many vehicles are present. This greatly improves privacy protection in situations with inherently low traffic density, e.g., suburban areas, and during low traffic periods. Our scheme ensure that an external attacker cannot distinguish between real and chaff vehicles, while legitimate vehicles can recognize chaff messages; this is important, because chaff vehicles (and messages) must not affect the operation of safety applications. In our evaluation, we compare our chaff-based approach with an existing cryptographic mix-zone scheme. Our results under realistic traffic conditions show that by introducing fictive vehicles, traffic flow variations can be smoothed and privacy protection can be enhanced up to 76%.