Impact of multi-class stochastic cyberattacks on vehicle dynamics and rear-end collision risks for heterogeneous traffic.

Cooperative adaptive cruise control (CACC) has been regarded as a promising approach to improving traffic stability and safety with vehicle-to-vehicle communication. However, current communication system is vulnerable to all kinds of cyberattacks, bringing potential security risks to the transportation system. Although cyberattacks have been well documented recently, realistic impact of cyberattacks is not replicated by reason of: First, real-world cyberattacks can take various attack forms and cause diverse damages, but there has been limited research in generalizing their effects on inter-vehicle communication or providing a comprehensive investigation into the resulting response in vehicle dynamics. Second, some adaptive traffic operations, e.g., platooning under a connected environment and dynamic controls under disturbance, were usually neglected. Third, cyberattacks were thought to have constant properties in terms of attack targets and attack severity, which goes against the stochastic nature of cyberattacks. To this end, the objective of this study is to address these existing gaps by generalizing multi-class stochastic cyberattacks and examining their impact in a more realistic traffic stream. Specifically, we categorized different cyberattacks as bogus messages, delay/replay, and malicious messages, and proposed their general forms with stochastic noise or failures on control variables, i.e., location, velocity, and acceleration. Moreover, the research scenario was expanded to a heterogeneous traffic comprising of CACC vehicles, ACC vehicles, and manually driven vehicles, incorporating adaptive traffic operations such as dynamic controls, vehicle degradation, and platooning. Finally, we analysed​ the rear-end collision risks in the presence of multi-class cyberattacks with surrogated safety measures and identified the most critical circumstances under the cyberattack combinations. The findings provide evidences for the adverse effects of cyberattacks on velocity oscillation and collision risks, based on our general modelling of cyberattacks in a more practical scenario. • Propose a framework for the general modelling of multi-class stochastic cyberattacks. • Extend the stochasticity of cyberattacks to stochastic targets and stochastic severity. • Expand the cyberattack scenarios to heterogeneous traffic with dynamical controls. • Assess the impact of cyberattacks on both velocity oscillation and collision risks. [ABSTRACT FROM AUTHOR]