Sybil Attack Detection in VANET Using CNN Enhanced with Chaotic Maps and Elephant Herding Optimization for Secure Data Transmission

The Vehicular Ad-Hoc Network (VANET) model stands out as a cost-effective and easily deployable solution for traffic management and accident prevention. Within VANET, nodes employ broadcast protocols for disseminating safety information rather than relying on routing protocols. Nonetheless, there exists a vulnerability to malicious activities, such as targeted attacks where a vehicle may intentionally transmit harmful packets to cause harm. Among these, the Sybil attack (SA) poses the most severe threat, wherein the attacker creates multiple identities to impersonate distinct nodes. Detecting and defending against such attacks, particularly when perpetrators operate under genuine identities, presents significant challenges. To mitigate this issue, a deep learning-based intrusion detection system (IDS) has been proposed for effectively identifying SA in VANET. The system employs a clustering algorithm known as Glow Worm Swarm Optimization (Gon SO)-based K-harmonic means (GSOKHM) for vehicle clustering. Subsequently, it utilizes the Floyd-Warshall algorithm (FWA) to designate Cluster Heads (CH) from these clusters. Following CH selection, our advanced CMEHA-CNN algorithm utilizes a combination of Convolutional Neural Network (CNN) and chaotic maps to detect any malicious CH. This entails extracting pertinent features from the CH. Upon confirming the legitimacy of the CH, its information is firmly transmitted to the cloud by means of SHA2-ECC, a fusion of Secure Hashing Algorithm and Elliptic Curve Cryptography. The simulation (NS-2.35) outcomes of our proposed methodology achieves an impressive accuracy rate of 98.9% and ensures a high level of security at 99%, surpassing existing methodologies.