Publish or Drop Traffic Event Alerts? Quality-aware Decision Making in Participatory Sensing-based Vehicular CPS

Abstract

Vehicular cyber-physical systems (VCPS), among several other applications, may help address an ever-increasing challenge of traffic congestion in large cities. Nevertheless, VCPS can be hindered by information falsification problem, resulting due to the wrong perception of a traffic event or deliberate faking by the participating vehicles. Such information fabrication causes the re-routing of vehicles and artificial congestion, leading to economic, safety, environmental, and health hazards. Thus, it is imperative to infer truthful traffic information in real-time to restore the operational reliability of the VCPS. In this work, we propose a novel reputation scoring and decision support framework, called Spoofed and False Report Eradicator (SAFE) , which offers a cost-effective and efficient solution to handle information falsification problem in the VCPS domain. The framework includes humans in the sensing loop by exploiting the paradigm of participatory sensing , a concept of a mobile security agent (MSA) to nullify the effects of deliberate false contribution, and a variant of the distance bounding mechanism to thwart location-spoofing attacks. A regression-based model integrates these effects to generate the expected truthfulness of a participant’s contribution. To determine if any contribution is true or false, a generalized linear model is used to transform the expected truthfulness into a Quality of Contribution (QoC) score. The QoC of different reports is aggregated to compute user reputation. Such reputation enables classification of different participation behaviors. Finally, an Expected Utility Theory (EUT) -based decision model is proposed that utilizes the reputation score to determine if event-specific information should be published or dropped. To evaluate the SAFE framework through experimental study, we used both simulated and real data to compare its reputation-based user segregation performance with state-of-the-art frameworks. Experimental results exhibit that SAFE captures the fine differences in participants’ behavior through the quality and quantity of participation, and the accuracy of their informed location. It also significantly improves operational reliability through publishing the information of only legitimate events.