Evaluation of Privacy Preserving Algorithms Using Traffic Knowledge Based Adversary Models
By providing location traces of individual vehicles, mobile traffic sensors have quickly emerged as an important data source for traffic applications. In dealing with the privacy issues associated with this, researchers have been proposing different privacy protection algorithms. In this paper, the authors propose traffic-knowledge-based adversary models to attack privacy algorithms. By doing so, they can compare and evaluate different privacy algorithms in terms of both privacy protection and the convenience for traffic modeling. Results show that by having a relatively good privacy performance, the released datasets of both the 3.3 level of confusion entropy and the 0.1 individual likelihood can still be applied for a fine level of traffic applications.