A Resilience Framework for Sybil Attacks on VANET with Machine Learning Based Classification

In Autonomous IoT networks, the IoT devices are equipped with artificial intelligence, edge computing, cloud computing and autonomous decision making. Applications of IoT networks include Autonomous Vehicles, Smart Cities, Industry 4.0, Precision Agriculture and Healthcare. Their distributed architecture and resource constrained devices make them vulnerable to identity – based threats and attacks. Vehicular Adhoc Networks (VANETs) are crucial for Intelligent Transportation System(ITS) and road safety. The VANETs are deployed in dynamic environments, with changing topologies, varying vehicle speeds and road conditions. The communication architecture used in VANETs are IEEE 802.11p in the PHY and MAC layers. IEEE 1609 standard is used in the upper layers to enable wireless access in Vehicular Adhoc Networks. A Sybil Attack is an attack in which a malicious device forges multiple identities to manipulate routing, trust and collaborative decision making. The effects of Sybil attack include reduced Packet Delivery Ratio, increased routing overhead, higher energy consumption, degraded network lifetime and false traffic congestion reports. Such Sybil attacks compromise safety and reliability of VANETs, since they rely on the exchange of trustworthy information. This paper focusses on the classification of Sybil attacks based on Machine Learning algorithms. The three machine learning algorithms Random Forest, K – Means clustering and Decision Tree are used for the classification of Sybil attacks and the results are compared. With effective Packet Delivery Ratio statistics, the resilience metrics such as PDR Degradation, Throughput Degradation, Packet Loss, Detection Time, Response Time and Recovery Time are calculated. Finally the Resilience Index(RI) is derived. The results show96.15% Accuracy, 98.84% Sybil detection rate, 97.89% F1-score, 50% reduction in Detection Time, 60% reduction in Response Time, Improved PDR and Throughput.