|Table of Contents|

[1] BEN Wei, MING Xiqin, LI Bing, YIN Guodong, et al. Vehicle intention recognition at signalized intersections based on security-aware inverse reinforcement learning [J]. Journal of Southeast University (English Edition), 2026, 42 (2): 165-172. [doi:10.3969/j.issn.1003-7985.2026.02.003]
Copy

Vehicle intention recognition at signalized intersections based on security-aware inverse reinforcement learning()
基于安全感知逆向强化学习的信号交叉口车辆意图识别

Journal of Southeast University (English Edition)[ISSN:1003-7985/CN:32-1325/N]

Volumn:
42
Issue:
2026 2
Page:
165-172
Research Field:
Publishing date:
2026-05-29

Info

Title:
Vehicle intention recognition at signalized intersections based on security-aware inverse reinforcement learning
基于安全感知逆向强化学习的信号交叉口车辆意图识别
Author(s):
BEN Wei1,2, MING Xiqin2, LI Bing1, YIN Guodong1, JIANG Fei3
1.School of Cyber Science and Engineering, Southeast University, Nanjing 211189, China
2.Nanjing LES Cybersecurity and Information Technology Research Institute Company Limited, Nanjing 210007, China
3.The 28th Research Institute of China Electronics Technology Group Corporation, Nanjing 210023, China
贲伟1,2, 闵溪青2, 李冰1, 殷国栋1, 蒋飞3
1.东南大学网络安全学院, 南京 211189
2.南京莱斯网信技术研究院有限公司, 南京 210007
3.中国电子科技集团公司第二十八研究所, 南京 210023
Keywords:
signalized intersections;  intent recognition;  anomaly detection;  inverse reinforcement learning (IRL);  vehicular ad-hoc networks
信号交叉口;  意图识别;  异常检测;  逆向强化学习(IRL);  车载自组网
PACS:
U495
DOI:
10.3969/j.issn.1003-7985.2026.02.003
Abstract:
To address the challenge of distinguishing subjective aggressive driving (initiated by drivers) from hazardous behaviors caused by external cyberattacks, this study proposes an innovative intent recognition framework named Intent-Decipher. By integrating the information credibility outputted by an intrusion detection system (IDS) into a security-aware inverse reinforcement learning (SA-IRL) model, the framework infers the reward function behind vehicle behaviors and classifies three key driving intents: normal, aggressive, and malicious. Experiments were conducted on a semi-synthetic dataset containing 20 000 trajectories. Results show that Intent-Decipher significantly outperforms baseline methods in classification accuracy, achieving a macro-average F1-score of 0.94. Notably, Intent-Decipher excels at differentiating subjective aggressive driving from attack-induced behaviors: its F1-score for identifying malicious attack-induced (MAI) intent reaches 0.90, an absolute improvement of 0.16 compared with the standard inverse reinforcement learning (IRL) model (which lacks security awareness and only achieves an F1-score of 0.74).
为解决车辆主观激进驾驶与外部网络攻击诱导的危险行为在外部特征上高度相似而难以区分的关键挑战,本文提出了一种名为“意图破译”的创新意图识别框架。该框架将入侵检测系统输出的信息可信度整合到安全感知逆向强化学习(SA-IRL)模型中,通过推断车辆行为背后的奖励函数,实现了对3种关键驾驶意图——正常、攻击性和恶意行为的精准分类。实验在包含20 000条轨迹的半合成数据集上进行,结果显示意图破译在分类准确率上显著优于基线方法,宏平均F1分数达到0.94。在区分主观的攻击性驾驶与客观的恶意攻击诱导行为这一关键任务中表现尤其突出,其对恶意攻击诱导(MAI)意图的识别F1分数达到0.90,相较于缺少安全感知的标准IRL模型(F1分数为0.74)提升了16个百分点。

References:

[1]ALMEHDHAR M, ALBASEER A, KHAN M A, et al. Deep learning in the fast lane: A survey on advanced intrusion detection systems for intelligent vehicle networks[J]. IEEE Open Journal of Vehicular Technology, 2024, 5: 869-906.
[2]CHU R S, SUN L S, KONG D W, et al. Knowledge-mapping visual analysis of formation research in connected and autonomous vehicles[J]. Journal of Southeast University (English Edition), 2025, 41(4): 525-542.
[3]BHATTACHARJEE S, SAHA S, NANDI S. A multi- modal explainable framework for adversarial intention identification in autonomous vehicles[J]. IEEE Transactions on Intelligent Transportation Systems, 2023, 24(7): 7230-7244.
[4]WANG C S, REN W L, XU C C, et al. Rear-end conflicts prediction of vehicle platoon on highways integrating data-driven stability analysis[J]. Journal of Southeast University (Natural Science Edition), 2025, 55(1): 239-245. (in Chinese)
[5]BAO Q, WU Z, CHEN J M, et al. Comprehensive evaluation of traffic operation risk on road sections near intersections based on vehicle trajectory prediction and macro-micro index aggregation[J]. Journal of Southeast University (Natural Science Edition), 2025, 55(1): 230-238. (in Chinese)
[6]BOHRA N, KUMARI A, MISHRA V K, et al. Intelligence-based strategies with vehicle-to-everything network: A review[J]. Future Internet, 2025, 17(2): 79.
[7]GHRAIZI D. Design of integrated decision-making and trajectory planning architectures for autonomous vehicles using deep reinforcement learning and risk assessment[D]. Compiègne, France: Université de Technologie de Compiègne, 2025.
[8]RAJAMANICKAM N, VISHNURAM P, ABRAHAM D S, et al. Review of authentication, blockchain, driver ID systems, economic aspects, and communication technologies in DWC for EVs in smart cities applications[J]. Smart Cities, 2024, 7(6): 3121-3164.
[9]CHEN S Z, HU X H, ZHAO J H, et al. A review of decision-making and planning for autonomous vehicles in intersection environments[J]. World Electric Vehicle Journal, 2024, 15(3): 99.
[10]HU L, ZHOU X Q, ZHANG X, et al. A review on key challenges in intelligent vehicles: Safety and driver-oriented features[J]. IET Intelligent Transport Systems, 2021, 15(9): 1093-1105.
[11]RAJAPAKSHA S, GURUSAMY M. Protecting vehicles from cyberattacks: context aware AI-based intrusion detection for vehicle CAN bus security[C]//Proceedings of the 2024 International Conference on Cybersecurity in Intelligent Transport. Singapore, 2024: 112-125.
[12]WANG Y, LI Z, XU C, et al. A security-aware intention prediction framework for connected vehicles using inverse reinforcement learning[J]. IEEE Transactions on Intelligent Vehicles, 2023, 8(2): 1562-1575.
[13]XU Q, ZHANG L, LIU Y X. Enhancing trust management system for connected autonomous vehicles using machine learning methods: A survey[EB/OL]. (2025-05-10)[2025-10-12]. https: //arxiv. org/abs/2505. 07882.
[14]KRISTIANTO E, LIN P C, HWANG R H. Sustainable and lightweight domain-based intrusion detection system for in-vehicle network[J]. Sustainable Computing: Informatics and Systems, 2024, 41: 100936.
[15]ZHANG L, LIU H, WANG K, et al. Integrating V2X communication reliability into driving behavior prediction: A hybrid deep learning approach[J]. Transportation Research Part C: Emerging Technologies, 2023, 146: 103991.
[16]PAN R Y, JIE L H, ZHANG X Y, et al. A V2P collision risk warning method based on LSTM in IOV[J]. Security and Communication Networks, 2022, 2022(1): 7507573.
[17]CHEN D, ZHU M X, YANG H, et al. Data-driven traffic simulation: A comprehensive review[J]. IEEE Transactions on Intelligent Vehicles, 2024, 9(4): 4730-4748.
[18]KUDERER M, GULATI S, BURGARD W. Learning driving styles for autonomous vehicles from demonstration[C]//2015 IEEE International Conference on Robotics and Automation (ICRA). Seattle, WA, USA, 2015: 2641-2646.

Memo

Memo:
Received: 2025-10-16; Revised: 2026-01-23.
Biographies: BEN Wei (1987—), male, Ph.D.candidate; YIN Guodong (corresponding author), male, doctor, professor, ygd@seu.edu.cn.
Foundation item: The National Key Research and Development Program of China (No.2022YFB4300304).
Citation: BEN Wei, MING Xiqin, LI Bing, et al. Vehicle intention recognition at signalized intersections based on security-aware inverse reinforcement learning[J]. Journal of Southeast University (English Edition), 2026, 42(2): 165-172. DOI: 10. 3969/j. issn. 1003-7985. 2026. 02. 003.
Last Update: 2026-06-20