Cooperative vehicle sensing vulnerable to new stealth feature attacks
Investigating Adversarial Robustness of Heterogeneous Cooperative Perception
Computer Vision and Pattern RecognitionCryptography and SecurityMultiagent SystemsRobotics
Summary
Connected vehicles share sensing data to better detect objects around them, even though their sensors can be different. It was believed that having different types of sensors would protect these vehicles from attacks where bad actors send fake data to hide real objects. This work shows that attackers can still fool these systems by smartly crafting fake data, even without knowing the victim’s exact setup. The authors created a fast method to produce these fake signals and also developed a simple defense that checks consistency over time to detect and reduce these attacks.
What this means in practice
- •For automotive cybersecurity teams: Protect vehicle networks by detecting and mitigating feature-level attacks against shared sensor data during cooperative perception.
- •For autonomous vehicle system engineers: Improve fusion modules by incorporating consistency checks to maintain reliable object detection despite adversarial manipulation.
Authors
Chenyi Wang, Yutong Liu, Qingzhao Zhang, Ming F. Li
Abstract
Heterogeneous cooperative perception (CP) enables connected vehicles with diverse sensor setups to share spatial awareness via compact feature maps, where receivers reconcile these maps using learned translation modules for fusion and inference. Prior attacks against CP in a homogeneous setting reveal that the data exchange introduces a critical attack surface: a single malicious agent can transmit crafted features that erase real objects from a neighbor's fused scene. Yet, it is widely hypothesized that heterogeneity naturally defends against these attacks, as the attacker lacks knowledge of the victim's detector and the translation module scrambles adversarial gradients. We demonstrate that this protection is largely an illusion. Using a matched-objective harness to standardize the perturbation budget, objective, and forward path, we show that properly tuned iterative attacks close or reverse the apparent robustness gap. However, these optimization-based attacks require ground-truth labels and iterative backpropagation, meaning they do not represent a practical field threat running in real-time. To bridge this gap, we introduce HetPoison, a learned generator that crafts a removal perturbation in a single, label-free forward pass. HetPoison transfers across major heterogeneous designs without requiring access to the victim's detector, matching or exceeding the effectiveness of expensive optimizer-based attacks. Since heterogeneity itself is not a defense, we propose HetShield, a lightweight trust layer that validates the spatiotemporal consistency across features, recovering 83--95% of the accuracy degraded by attacks, outperforming prior art.