# Secure Estimation and Attack Isolation for Connected and Automated Driving in the Presence of Malicious Vehicles

**Authors:** Tianci Yang, Chen Lv  
**arXiv:** [2103.00878](https://arxiv.org/abs/2103.00878)  
**Date:** 2021-03-01  
**Summary page:** https://www.summarizepaper.com/en/paper/2103.00878v1/

## Key points

- Connected and Automated Vehicles (CAVs) rely on accurate position and vehicle kinematics information for driving tasks
- Malicious vehicles can send false sensor information, leading to traffic disruptions or loss of human lives
- Authors propose using cloud-computing to enhance resilience against malicious vehicles
- Vehicles share local sensor information in the cloud to create redundancy
- Sensor fusion algorithm implemented in the cloud aims to provide robust state estimation of all vehicles
- Algorithm identifies and isolates malicious vehicles by combining data from multiple sources
- Numerical examples demonstrate successful detection and isolation of malicious vehicles while maintaining accurate state estimation for other vehicles
- Research presents a novel solution for enhancing security and reliability of CAVs through cloud-computing and information redundancy.

## Summary

Connected and Automated Vehicles (CAVs) rely on accurate position and vehicle kinematics information to perform various driving tasks, including vehicle following, lane change, and collision avoidance. However, there is a risk of malicious vehicles intentionally or unintentionally sending false sensor information to other vehicles, which can lead to traffic disruptions or even loss of human lives. To address this issue, the authors propose leveraging cloud-computing to enhance the resilience of CAVs against malicious vehicles. The idea is for each vehicle to share its local sensor information with other vehicles, creating redundancy in the cloud. This redundancy is then exploited through a sensor fusion algorithm implemented in the cloud. The proposed sensor fusion algorithm aims to provide a robust state estimation of all vehicles in the cloud, assuming that the number of malicious information is relatively small. By combining data from multiple sources and cross-referencing them, the algorithm can identify and isolate malicious vehicles. To demonstrate the effectiveness of their approach, numerical examples are provided which showcase how the proposed estimator successfully detects and isolates malicious vehicles while maintaining accurate state estimation for all other vehicles. Overall, this research presents a novel solution for enhancing the security and reliability of CAVs by leveraging cloud-computing and information redundancy. By effectively detecting and isolating malicious vehicles, this approach has significant potential in ensuring safer and more efficient autonomous driving systems.