Security and Resilience in Autonomous Vehicles: A Proactive Design Approach

Chieh Tsai; Murad Mehrab Abrar; Salim Hariri

cs.CRApr 14, 2026

Security and Resilience in Autonomous Vehicles: A Proactive Design Approach

Authors: Chieh Tsai, Murad Mehrab Abrar, Salim Hariri

Abstract

Autonomous vehicles (AVs) promise efficient, clean and cost-effective transportation systems, but their reliance on sensors, wireless communications, and decision-making systems makes them vulnerable to cyberattacks and physical threats. This chapter presents novel design techniques to strengthen the security and resilience of AVs. We first provide a taxonomy of potential attacks across different architectural layers, from perception and control manipulation to Vehicle-to-Any (V2X) communication exploits and software supply chain compromises. Building on this analysis, we present an AV Resilient architecture that integrates redundancy, diversity, and adaptive reconfiguration strategies, supported by anomaly- and hash-based intrusion detection techniques. Experimental validation on the Quanser QCar platform demonstrates the effectiveness of these methods in detecting depth camera blinding attacks and software tampering of perception modules. The results highlight how fast anomaly detection combined with fallback and backup mechanisms ensures operational continuity, even under adversarial conditions. By linking layered threat modeling with practical defense implementations, this work advances AV resilience strategies for safer and more trustworthy autonomous vehicles.

Figures & tables

Layer	Attack Surface (AS)	Impact	Mitigation Strategy
Driver Interface and Control System	Malicious driver input injection; compromised dashboard/override signals	Unauthorized override of system behavior; driver confusion leading to unsafe maneuvers	Authenticate commands, secure UI design, redundant safety checks for manual override
Control Layer	Command injection in steering/braking; kernel-level rootkits	Silent manipulation of vehicle dynamics, risk of collisions	Runtime integrity monitoring, secure boot, anomaly detection in actuation signals
Decision-Making Layer	Tampered ML models; adversarial inputs to planning module	Unsafe maneuvers, rule violations, unsafe path selection	Model validation, adversarial robustness testing, consensus-based decision redundancy
Knowledge Processing Layer	Data poisoning; sensor fusion manipulation	Distorted environmental model, leading to misinformed decisions	Trusted data provenance, anomaly detection in fused data, hash-based detection, back up and fall back
Perception Layer	Sensor spoofing (LiDAR, radar, camera); adversarial image perturbations	Incorrect object detection, phantom obstacles, missed hazards	Sensor redundancy, adversarial training, anomaly detection
Communication Framework	V2V/V2I spoofing, replay, DoS, packet manipulation	Loss of situational awareness, degraded coordination between vehicles	Encryption, authentication protocols, rate limiting, intrusion detection systems

Table 1: Threat Modeling for the proposed AV architecture. Each row summarizes the attack surface, safety impact, and candidate mitigation strategy.

Figure 1: AVR architecture integrated across AV functional layers and coordinated by a Cross-Layer Coordinator.

Figure 2: Two operational scenarios for the backup/fallback mechanism: (1) happy state with primary execution, (2) failure state with automatic switchover to verified fallback.

Figure 3: Intrusion Detection Techniques: anomaly-based (perception) and hash-based (software/system integrity).

Figure 4: Simplified Autonomous Vehicle Behavior Model

Figure 5: Anomaly-based Intrusion Detection Methodology

Figure 6: Setup of the depth camera blinding attack experiment on QCar

Figure 7: Software-based attack: ECU perception module compromise on QCar2.

Metrics	RF	LR	XGB	KNN	SVC	MLP	NB
Precision	0.801	0.800	0.751	0.793	0.742	0.786	0.692
Recall	0.894	0.872	0.876	0.877	0.913	0.874	0.923
F1 Score	0.845	0.834	0.809	0.833	0.819	0.828	0.791
Accuracy	0.838	0.829	0.795	0.826	0.800	0.820	0.759

Table 2: Depth Camera Blinding Attack Detection Performance of Different Machine Learning Models

Figure 8: Distribution of Random Forest abnormal-class probability scores for normal and attack data. The evaluation fold consists of 3,893 normal samples and 13,902 attack samples. An effective detection threshold lies in the range $\mathbb{T}=0.4$ – $0.5$ .

Detection Margin	Normal Data Misclassified	Attack Data Misclassified	FP Rate	FN Rate
0.4 – 0.5	155	0	0.0398	0
0.3 – 0.5	293	0	0.0752	0
0.4 – 0.6	155	26	0.0398	0.00187
0.3 – 0.6	293	26	0.0752	0.00187

Table 3: False Positive (FP) and False Negative (FN) Rates for Different Detection Margins

Figure 9: Success rate of hash-based integrity validation across vehicle speeds (0.5–1.75 m/s) and hash update intervals (1 s, 3 s, and 5 s), showing reliable detection of all software modifications.

Figure 10: System response timeline under software tampering. The attack is detected via hash mismatch within 0.030 s after onset, and the trusted configuration is restored within 0.002 s, preventing unsafe vehicle behavior.

Figure 11: QCar speed profile during ECU tampering. After the attack is injected at 10 s, the vehicle continues operating under the fallback perception module and correctly stops at a STOP sign at 20 s, demonstrating preserved safety behavior without an emergency halt.

Autonomous Driving

Autonomous driving develops vehicles that perceive road scenes, predict interactions, and plan and control safe movement with limited or no human input.

See all 536 papers

Resilient Control Loops in Autonomous Vehicles Under Adversarial Jamming via Spectral Perception and Network-Layer Failover
Sep 4, 2026Luis Barajas, Colin Jeardoe, Jaewon Kim +1Wireless Communications Fault-Tolerant Control
iSTAR Laboratory · Department of Engineering Technology and Industrial Distribution, Texas A&M University, College Station, TX, USA · Global Cyber Research Institute (GCRI)
Autonomous Cyber Defense in Connected Vehicles: A Multi-Agent Approach to V2X Security
Aug 19, 2026Krishna Teja MedamAutonomous Cyber Defense V2X Communication
Dept. of Electrical Engineering & Computer Engineering, and Computer Science University of Detroit Mercy Detroit, MI, USA
Adversarial Robustness Analysis of Cloud-Assisted Autonomous Driving Systems
Apr 6, 2026Maher Al Islam, Amr S. El-WakeelAdversarial Attacks on Object Detectors Adversarial Robustness
Lane Department of Computer Science and Electrical Engineering, West Virginia University, WV, USA

Sep 4, 2026J/K move · Enter open · S save

Resilient Control Loops in Autonomous Vehicles Under Adversarial Jamming via Spectral Perception and Network-Layer Failover

Luis Barajas, Colin Jeardoe, Jaewon Kim, Eman Hammad

iSTAR Laboratory · Department of Engineering Technology and Industrial Distribution, Texas A&M University, College Station, TX, USA · Global Cyber Research Institute (GCRI)

Wireless Communications Fault-Tolerant Control Mobile Robotics Robust Control

Open paper AskPDF arXiv

The operational integrity of autonomous mobile robots relies on the continuous availability of wireless control loops, making them highly attractive targets for adversarial intentional electromagnetic interference. This paper introduces a resilient, cross-layer framework that combines physical-layer spectral perception with network-layer routing optimization to protect middleware stability, such as ROS2, during intentional electromagnetic interference. Utilizing a software-defined radio front-end, the system extracts dynamic spectral descriptors, including spectral entropy and channel occupancy, to inform a Random Forest classifier that establishes adaptive environmental baselines. To ensure uninterrupted data flow, the architecture maintains dual pre-authenticated physical interfaces in a hot-standby configuration, enabling instantaneous failover through automated network routing table updates. Empirical validation on a physical ROS2 mobile robot testbed demonstrates that this adaptive hardware-assisted architecture optimizes communication recovery to an average of 141ms. This sub-second restoration translates directly into a 78.9% reduction in pooled root-mean-square path tracking error compared to software re-association, successfully securing system-level mission integrity.

Abstract

Autonomous Driving

Explore similar work

Resilient Control Loops in Autonomous Vehicles Under Adversarial Jamming via Spectral Perception and Network-Layer Failover

Autonomous Cyber Defense in Connected Vehicles: A Multi-Agent Approach to V2X Security

Adversarial Robustness Analysis of Cloud-Assisted Autonomous Driving Systems