Scholarlink

<p>IoT-enabled systems are transforming the way modern transportation works, enabling autonomous vehicles with intelligent decision-making, efficient communication and improved safety. In combination, the capabilities of enhanced interaction between cars and with infrastructure and with users open the doors to approaches to advanced traffic management optimisation and energy-efficient mobility systems. There are important hurdles to overcome if we are to safely and reliably transmit data using autonomous systems. However, these systems heavily depend on the sensor communication platforms, but they are still very vulnerable to jamming attacks, data injection incidents and other cybersecurity threats. These vulnerabilities compromise the data, putting operational safety alongside reliability at extreme risk. The current IDS systems suffer from many critical inadequacies when faced with these security issues. Standard intrusion detection models rooted in rule frameworks and machine learning strategies do not exhibit the capability to adapt to shifting operational conditions. For these systems, which are unable to give clear detection process explanations, detecting anomalies in time-dependent high-dimensional sensor data has been impractical. An Explainable AI-Driven Intrusion Detection System (EAI-IDS) is developed to improve security protection in IoT-enabled autonomous vehicle communication by this research. The system also presents new analytical models which combine the use of LSTM networks to identify temporal patterns in sensor measurements and Random Forest on a feature-level anomaly detection. Integration of Shapley Additive Explanations (SHAP) into the anomaly detection system allows it to explain its way of decision-making by finding features that have a significant effect on deciding to detect anomalies. It shows both anomaly detection capabilities and features-rich transparency in its decision-making systems. In comparison to traditional IDS models, the EAI-IDS framework shows superior real-time anomaly detection with a 94.2% detection accuracy, 15.2&#xa0;ms latency and 2.1&#xa0;J of energy consumption. With 98.5% PDR, the system supports 1000 nodes and provides secure and efficient IoT-enabled autonomous vehicle operations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Explainable AI-driven intrusion detection for securing IoT-enabled autonomous transportation systems

  • 期刊论文

  • 发表日期：

    2025年9月29日

查看全文
  • J. Akshya,
  • M. Sundarrajan,
  • R. Vijayakumar,
  • Rajesh Kumar Dhanaraj,
  • Anand Nayyar

摘要

IoT-enabled systems are transforming the way modern transportation works, enabling autonomous vehicles with intelligent decision-making, efficient communication and improved safety. In combination, the capabilities of enhanced interaction between cars and with infrastructure and with users open the doors to approaches to advanced traffic management optimisation and energy-efficient mobility systems. There are important hurdles to overcome if we are to safely and reliably transmit data using autonomous systems. However, these systems heavily depend on the sensor communication platforms, but they are still very vulnerable to jamming attacks, data injection incidents and other cybersecurity threats. These vulnerabilities compromise the data, putting operational safety alongside reliability at extreme risk. The current IDS systems suffer from many critical inadequacies when faced with these security issues. Standard intrusion detection models rooted in rule frameworks and machine learning strategies do not exhibit the capability to adapt to shifting operational conditions. For these systems, which are unable to give clear detection process explanations, detecting anomalies in time-dependent high-dimensional sensor data has been impractical. An Explainable AI-Driven Intrusion Detection System (EAI-IDS) is developed to improve security protection in IoT-enabled autonomous vehicle communication by this research. The system also presents new analytical models which combine the use of LSTM networks to identify temporal patterns in sensor measurements and Random Forest on a feature-level anomaly detection. Integration of Shapley Additive Explanations (SHAP) into the anomaly detection system allows it to explain its way of decision-making by finding features that have a significant effect on deciding to detect anomalies. It shows both anomaly detection capabilities and features-rich transparency in its decision-making systems. In comparison to traditional IDS models, the EAI-IDS framework shows superior real-time anomaly detection with a 94.2% detection accuracy, 15.2 ms latency and 2.1 J of energy consumption. With 98.5% PDR, the system supports 1000 nodes and provides secure and efficient IoT-enabled autonomous vehicle operations.