Security Defense of Transportation Networks Against Cyberattacks: A Physics-Informed AI Approach
archive: archived pipeline: cataloged verified
Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)
Summary
This study addresses the cybersecurity vulnerabilities of connected and automated vehicles (CAVs), which rely on digital information streams for decision-making and are susceptible to cyberattacks that manipulate perceived traffic states. Such manipulation can subtly alter vehicle behavior, propagating disturbances through interactive traffic environments. The research focuses on lane-changing maneuvers, a safety-critical tactical action sensitive to observation perturbations, with two primary objectives: developing a framework for detecting abnormal lane-changing under cyberattacks and enabling robust decision-making under adversarial conditions. To achieve these goals, the authors developed two complementary modeling components grounded in physics-informed AI. For detection, they created a physics-guided neural network (PGNN) that integrates a game-theoretic lane-changing model with a Long Short-Term Memory (LSTM) network. The game-theoretic component captures interaction mechanisms and gap-acceptance structures, while the LSTM captures temporal dependencies in trajectory data. The system detects anomalies by monitoring cumulative discrepancies between predicted decision probabilities and observed maneuver execution. For robust control, the study proposes a hierarchical adversarial reinforcement learning (HARL) framework. This approach explicitly models adversarial observation perturbations, using an upper-level policy to determine maneuver decisions under worst-case bounded sensor disturbances and a lower-level Stackelberg execution module to resolve timing and longitudinal control. The PGNN was evaluated using the I-24 MOTION dataset under simulated false data injection (FDI) and denial-of-service (DoS) attacks. Results demonstrated improved prediction accuracy compared to purely data-driven or physics-based models and effective detection of falsified lane-changing behaviors. The HARL framework was tested via simulations, showing improved traffic efficiency while maintaining safety under worst-case perturbations. The robust decision-making model successfully balanced safety, self-efficiency, and flow-level stability despite adversarial conditions. The findings provide a modeling foundation for cybersecurity-aware monitoring and robust tactical decision-making in CAV systems. By embedding physical interaction principles into learning-based models, the study offers interpretable methods for distinguishing malicious deviations from normal traffic variability. This approach enhances the resilience of CAV operations against cyber threats, ensuring that vehicle behavior remains safe and efficient even when perception systems are compromised.
Key finding
A physics-guided neural network integrating game theory with LSTM effectively detects falsified lane-changing behaviors under cyberattacks, while hierarchical adversarial reinforcement learning enables robust and efficient lane-changing decisions despite adversarial sensor perturbations.
Methodology
modeling
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (5 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | rosap | — | — | 2 | 2026-05-23 |
| archive | success | — | — | — | 1 | 2026-05-23 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | — | — | — | 1 | 2026-06-01 |
| chunk | success | — | — | — | 1 | 2026-06-01 |
| embed | success | — | — | — | 1 | 2026-06-02 |
| enrich | success | — | — | — | 1 | 2026-05-23 |
| promote | success | — | — | — | 1 | 2026-05-23 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 3 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 23 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Theoretical Contribution: computational model