A Lightweight Sequential AI Framework for Real Time Intrusion Detection in Dynamic Vehicular Networks

Vehicular Ad-Hoc Networks (VANETs) is a type’s of Mobile Ad Hoc Networks (MANETs) designed to facilitate wireless communication among vehicles as well as between vehicles and fixed roadside infrastructure. These networks play a vital role in the development of Intelligent Transportation Systems (ITS), aiming to optimize traffic flow, and provide infotainment services to drivers and passengers. Due to high node mobility, frequent topology changes, and decentralized communication architectures of VANET, making real-time and reliable data exchange a complex challenge. Moreover, open and dynamic nature of VANETs is also susceptible to wide range of security threats. Ensuring secure and efficient communication within VANETs faced more challenging s. To address these challenges, a novel Sequential AI-Powered Lightweight Intrusion Detection System (Seq-AIIDS) is proposed. The main aim of Seq-AIIDS is to enhance the accuracy of intrusion detection with minimal time consumption. The proposed Seq-AIIDS comprises of four major steps namely data acquisition, scrubbing feature selection and classification. Initially, the system performs data acquisition by gathering relevant samples from a dataset. Following this, Rosenthal correlation function is employed for feature selection, identifying the more necessary features for accurate intrusion detection. With these refined features, the system proceeds to the classification phase, where a liquid neural network model is utilized to differentiate between normal and malicious network behavior. The architecture adopts a sequential AI-driven framework capable of detecting and responding to cyber attacks in Real-time detection high-mobility vehicular networks. Fine-tuning the layers in the liquid neural network is an essential step using the multipoint spiral search optimization algorithm thereby minimizing both training and validation errors and increasing the accuracy of the intrusion detection. To evaluate system performance, various metrics are considered, including detection accuracy, precision, recall, F1-score, specificity, and intrusion detection latency relative to the of input data samples. Experimental results demonstrate that the proposed Seq-AIIDS outperforms traditional methods, offering superior accuracy and lower computational latency, making it highly effective for security improvement in Dynamic and Decentralized VANETs.