Vehicle to Everything (V2X) allows vehicles, pedestrians, and infrastructure to
share information for the purpose of enhancing road safety, improving traffic
conditions, and lowering transporation costs. Although V2X messages are
authenticated, their content is not validated. Sensor errors or adversarial
attacks can cause messages to be perturbed increasing the likelihood of traffic
jams, compromising the decision process of other vehicles, and provoking fatal
crashes. In this article, we introduce V2X Core Anomaly Detection System
(VCADS), a system based on the theory presented in [1] and built for the fields provided in the periodic
messages shared across vehicles (i.e., Basic Safety Messages, BSMs). VCADS uses
physics-based models to constrain the values in each field and detect anomalies
by finding the numerical difference between a field and and its derivation using
orthogonal values. VCADS evaluation is performed with four real V2X field
testing datasets and a suite of attack simulations. The results show that VCADS
can constrain a variety of real-world network environments and is able to detect
~85% to ~95% of attacks coming from an adversary capable of perturbing one or
more data fields.
