An official website of the United States government
Panoptic perception models in autonomous driving use deep learning models to interpret their surroundings and make real-time decisions. However, these models are susceptible, carefully designed noise can fool models all while being imperceptible to humans. In this work, we investigate the impact of blackbox adversarial noise attacks on three core perception tasks: drivable area recognition, lane line segmentation, and object detection. Unlike white-box attacks, black-box attacks assume no knowledge of the model’s internal parameters making them a more realistic and challenging threat scenario. Our goal is to evaluate how such an attack affects the model’s predictions and explore countermeasures towards such attacks. In response to our implemented attack, we have tested various defense methods. With each defense method, we have assessed the recovery on prediction accuracy. This research aims to provide valuable insights into the vulnerabilities of panoptic perception models and highlights strategies for enhancing their resilience against adversarial manipulation within real-world scenarios. All our attacks are performed against images from the BDD100K dataset.
more »
« less
Beyond the Glow: Understanding Luminescent Marker Behavior Against Autonomous Vehicle Perception Systems
Autonomous driving (AD) systems rely heavily on accurate lane marker detection for safe navigation, particularly during nighttime or low-light conditions. While luminescent lane markers have been introduced to improve visibility and enhance road safety in these scenarios, they also introduce potential vulnerabilities. This paper investigates these risks by introducing novel luminescent adversarial attacks that exploit the lane detection models used in autonomous vehicles (AVs). We demonstrate how these attacks, targeting deep neural network-based perception models, can manipulate the textural properties of the markers to cause misdetection of lanes, leading to safety violations. Through comprehensive experiments in both digital and physical domains, we systematically expose the vulnerabilities of state-of-the-art lane detection models to adversarial luminescent markers. In our digital experiments, we observe complete model failure in the worst cases and a failure rate of approximately 33% in the best cases. Physical experiments using a device running Openpilot further confirm these risks, underscoring a significant safety threat posed by luminescent adversarial attacks. Our findings emphasize the need for robust defenses to protect AVs from such adversarial threats.
more »
« less
- Award ID(s):
- 2443252
- PAR ID:
- 10677561
- Publisher / Repository:
- 3rd USENIX Symposium on Vehicle Security and Privacy
- Date Published:
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Multi-sensor fusion systems (MSFs) play a vital role as the perception module in modern autonomous vehicles (AVs). Therefore, ensuring their robustness against common and realistic adversarial semantic transformations, such as rotation and shifting in the physical world, is crucial for the safety of AVs. While empirical evidence suggests that MSFs exhibit improved robustness compared to single-modal models, they are still vulnerable to adversarial semantic transformations. In addition, although many empirical defenses have been proposed, several works show that these defenses can be further attacked by new adaptive attacks. So far, there is no certified defense proposed for MSFs. In this work, we propose the first robustness certification framework COMMIT to certify the robustness of multi-sensor fusion systems against semantic attacks. In particular, we propose a practical anisotropic noise mechanism that leverages randomized smoothing on multi-modal data and performs a grid-based splitting method to characterize complex semantic transformations. We also propose efficient algorithms to compute the certification in terms of object detection accuracy and IoU for large-scale MSF models. Empirically, we evaluate the efficacy of COMMIT in different settings and provide a comprehensive benchmark of certified robustness for different MSF models using the CARLA simulation platform. We show that the certification for MSF models is at most 48.39% higher than that of single-modal models, which validates the advantages of MSF models. We believe our certification framework and benchmark will contribute an important step towards certifiably robust AVs in practice.more » « less
-
Autonomous Vehicles (AVs) are revolutionizing transportation, but their reliance on interconnected cyber-physical systems exposes them to unprecedented cybersecurity risks. This study addresses the critical challenge of detecting real-time cyber intrusions in self-driving vehicles by leveraging a dataset from the Udacity self-driving car project. We simulate four high-impact attack vectors, Denial of Service (DoS), spoofing, replay, and fuzzy attacks, by injecting noise into spatial features (e.g., bounding box coordinates) to replicate adversarial scenarios. We develop and evaluate two lightweight neural network architectures (NN-1 and NN-2) alongside a logistic regression baseline (LG-1) for intrusion detection. The models achieve exceptional performance, with NN-2 attaining an AUC score of 93.15% and 93.15% accuracy, demonstrating their suitability for edge deployment in AV environments. Through explainable AI techniques, we uncover unique forensic fingerprints of each attack type, such as spatial corruption in fuzzy attacks and temporal anomalies in replay attacks, offering actionable insights for feature engineering and proactive defense. Visual analytics, including confusion matrices, ROC curves, and feature importance plots, validate the models' robustness and interpretability. This research sets a new benchmark for AV cybersecurity, delivering a scalable, field-ready toolkit for Original Equipment Manufacturers (OEMs) and policymakers. By aligning intrusion fingerprints with SAE J3061 automotive security standards, we provide a pathway for integrating machine learning into safety-critical AV systems. Our findings underscore the urgent need for security-by-design AI, ensuring that AVs not only drive autonomously but also defend autonomously. This work bridges the gap between theoretical cybersecurity and life-preserving engineering, offering a leap toward safer, more secure autonomous transportation.more » « less
-
Autonomous vehicles rely on deep neural networks (DNNs) for traffic sign recognition, lane centering, and vehicle detection, yet these models are vulnerable to attacks that induce misclassification and threaten safety. Existing defenses (e.g., adversarial training) often fail to generalize and degrade clean accuracy. We introduce Vehicle Vision–Language Models (V2LMs), fine-tuned Vision Language Models (VLMs) special- ized for AV perception, and show that they are inherently more robust to unseen attacks without adversarial training, maintain- ing substantially higher adversarial accuracy than conventional DNNs. We study two deployments: Solo (task-specific V2LMs) and Tandem (a single V2LM for all three tasks). Under attacks, DNNs drop 33%–74%, whereas V2LMs decline by under 8% on average. Tandem achieves comparable robustness to Solo while being more memory-efficient. We also explore integrating V2LMs in parallel with existing perception stacks to enhance resilience. Our results suggest V2LMs are a promising path to- ward secure, robust AV perception.more » « less
-
Large language models trained for safety and harmlessness remain susceptible to adversarial misuse, as evidenced by the prevalence of "jailbreak" attacks on early releases of ChatGPT that elicit undesired behavior. Going beyond recognition of the issue, we investigate why such attacks succeed and how they can be created. We hypothesize two failure modes of safety training: competing objectives and mismatched generalization. Competing objectives arise when a model's capabilities and safety goals conflict, while mismatched generalization occurs when safety training fails to generalize to a domain for which capabilities exist. We use these failure modes to guide jailbreak design and then evaluate state-of-the-art models, including OpenAI's GPT-4 and Anthropic's Claude v1.3, against both existing and newly designed attacks. We find that vulnerabilities persist despite the extensive red-teaming and safety-training efforts behind these models. Notably, new attacks utilizing our failure modes succeed on every prompt in a collection of unsafe requests from the models' red-teaming evaluation sets and outperform existing ad hoc jailbreaks. Our analysis emphasizes the need for safety-capability parity -- that safety mechanisms should be as sophisticated as the underlying model -- and argues against the idea that scaling alone can resolve these safety failure modes.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript
- Conference Paper:
- The DOI is not currently available.
