Design of a misbehavior detection system for objects based shared perception V2X applications
Published in ‘IEEE Intelligent Transportation Systems Conference’, IEEE 2019
Abstract
The recent trend of integrating vehicular communications with advanced sensors installed on vehicles, enables Connected and Autonomous Vehicles (CAVs) to share their own driving information as well as perception information, such as a list of perceived objects (e.g., dynamic obstacles such as vehicles, pedestrians, and cyclists, and static obstacles). This has the potential to improve driving safety by expanding collective perception of vehicles. However, adversaries may also populate false information to other Connected Vehicles (CVs) via Vehicle-to-Vehicle (V2V) communications. This paper investigates the security aspects of mixed deployment of CAVs, CVs and legacy vehicles, and in particular with regards to misbehavior detection. We provide a generic design framework that is independent from the specific algorithms of the underlying perception system, and can be used to implement a practical Misbehavior Detection System (MDS). We analyze the MDS framework w.r.t. a ghost vehicle attack. While no computing system can be completely secure, we believe this work would help the industry to develop a practical MDS design within a common framework while allowing individual techniques to mature and evolve over time with future academic research.
Key Contributions
- Proposes a generic misbehavior detection framework
- Divides detection into local/collaborative components
- Categorizes four levels of anomaly detection
- Defines Unobservability and Undecidability concepts
- Defines conditions for successful attacks