# Compromised ACC vehicles can degrade current mixed-autonomy traffic performance while remaining stealthy against detection

**Authors:** George Gunter, Huichen Li, Avesta Hojjati, Matthew Nice, Matthew Bunting, Carl A. Gunter, Bo Li, Jonathan Sprinkle, Daniel Work  
**arXiv:** [2112.11986](https://arxiv.org/abs/2112.11986)  
**Date:** 2021-12-22  
**Summary page:** https://www.summarizepaper.com/en/paper/2112.11986v1/

## Key points

- The paper presents a study on the impact of compromised adaptive cruise control (ACC) vehicles on mixed-autonomy traffic performance.
- Supply-chain level compromise of ACC capability can significantly degrade system-level performance, causing congestion waves that decrease average speed and network throughput.
- Random deceleration attacks (RDAs) can have varying effects depending on underlying traffic congestion levels and what percentage of ACC vehicles can be compromised.
- In moderate congestion regimes, mean commuter speed can be degraded by over 7%, while overall network throughput can be reduced by up to 3% in high-density regimes. Commuters may also incur costs exceeding 300 USD/km hr in moderate to high congestion regimes.
- Anomaly detection technique using GPS traces is unable to distinguish between compromised ACCs and benign/normal drivers, making these attacks stealthy to detection.
- Stronger attacks can be accurately labeled as malicious, indicating that there is a limit to how impactful attacks can be before they are no longer stealthy.
- RDA attack was experimentally executed on a real commercially available ACC vehicle demonstrating its possible real-world feasibility.
- Findings highlight potential vulnerabilities in current mixed-autonomy traffic systems and underscore the need for improved security measures at both supply-chain and system levels.

## Summary

The paper presents a study on the impact of compromised adaptive cruise control (ACC) vehicles on mixed-autonomy traffic performance. The authors demonstrate that a supply-chain level compromise of the ACC capability can significantly degrade system-level performance, causing congestion waves that decrease average speed and network throughput. Using a realistic traffic simulation environment, the authors quantify the impacts of random deceleration attacks (RDAs) on a model of a high-volume freeway in the United States. The results show that the effect of the attack depends on both underlying traffic congestion levels and what percentage of ACC vehicles can be compromised. In moderate congestion regimes, mean commuter speed can be degraded by over 7%, while overall network throughput can be reduced by up to 3% in high-density regimes. Commuters may also incur costs exceeding 300 USD/km hr in moderate to high congestion regimes. The authors also develop an anomaly detection technique using GPS traces to identify malicious/compromised vehicles; however, they find that this technique is unable to distinguish between compromised ACCs and benign/normal drivers, making these attacks stealthy to detection. Stronger attacks can be accurately labeled as malicious, indicating that there is a limit to how impactful attacks can be before they are no longer stealthy. Finally, the authors experimentally execute the RDA attack on a real commercially available ACC vehicle demonstrating its possible real-world feasibility. These findings highlight potential vulnerabilities in current mixed-autonomy traffic systems and underscore the need for improved security measures at both supply-chain and system levels.