Machine learning-based ERA model for detecting Sybil attacks on mobile ad hoc networks

Abstract

Mobile ad hoc networks provide a substantial security threat because they lack central management and sufficient resources. These networks function autonomously without any central authority regulating the inclusion or removal of nodes. Nodes have the autonomy to choose when to join or quit. Dynamic multi-hop networks, either stationary or mobile, provide quick and simple access to data. Predicting the evolution of MANET can be challenging due to the network’s dispersion and self-organization, as well as its unpredictable and constantly changing topology. The independent organization of nodes in MANETs, coupled with their dispersion, may complicate the prediction of the network’s future growth due to its unstable and constantly changing structure. A Sybil attack, a deceptive tactic, involves a small number of individuals creating multiple counterfeit identities to gain dominance over a substantial portion of the system. To deceive legitimate users into believing that their system is utilizing their identities, the malicious attacker node adopts numerous identities. An attacker aims to gather a substantial number of node IDs, potentially generated at random, to appear and function as distinct nodes. Within the peer-to-peer overlay, the enemy can approach a single object or a group of objects by adopting different identities. Mobile ad hoc networks are intrinsically less secure than wired networks due to inherent security vulnerabilities and limited energy resources. To enhance detection accuracy, it is recommended to employ an ensemble regression arboretum model, which is a type of machine learning prediction model. This study proposes a machine learning-based method for detecting Sybil attacks in MANETs by collecting network metrics such as traffic characteristics, communication patterns, and node behaviors.