Differentially private streaming to untrusted edge servers in intelligent transportation system

This paper considers the privacy issues in the intelligent transportation system, in which the data is largely communicated based upon vehicle-to-infrastructure and vehicle-to-vehicle protocols. The sensory data communicated by the vehicles contain sensitive information, such as location and speed, which could violate the driver's privacy if they are leaked with no perturbation. Recent studies suggested mechanisms for randomizing the stream of vehicular data to ensure individuals' privacy. Although the past works on differential privacy provide a strong privacy guarantee, they are limited to applications where communication parties are trusted and/or data is limited to a few types. In this paper, we address this gap by proposing a differentially private mechanism that adds noise in the user side rather than the server. Also, our mechanism is able to perturb various types of data as pointed out by the dedicated short-range communication protocols in the automotive industry. The proposed mechanism is data adaptive and scales the noise with respect to the data type and distribution. Our extensive experiments show the accuracy of our mechanism compared to the recent approaches.