Ensemble agent based machine learning approach for energy efficient attack detection and prevention in MANETs

TL;DR

This paper introduces an energy-efficient, agent-based machine learning system to detect and prevent attacks in mobile ad hoc networks, improving security and performance.

Contribution

A novel agent-based attack detection system combined with energy-efficient clustering and machine learning for MANETs.

Findings

The proposed approach achieved a throughput of 93 Kbps, an 8% improvement over existing methods.

The system effectively detects blackhole, wormhole, and suspicious nodes in real-time.

Ensemble clustering optimization enhances energy efficiency and data aggregation reliability.

Abstract

Mobile Ad hoc Networks (MANETs) represent a decentralized and self-tuning network paradigm that relies on routing protocols to transmit data from source to destination. However, the absence of a fixed infrastructure makes MANETs vulnerable to various security threats, including blackhole and gray hole attacks. Addressing these vulnerabilities is critical to ensuring the reliability and security of MANETs. The paper proposes an agent-based approach for effectively identifying and preventing such attacks within the MANET environment. Unlike existing static or centralized models, agent-based approach deploys dedicated agent nodes in each cluster for real-time monitoring and classification of malicious behaviour. Furthermore, the paper introduces an energy-efficient optimum clustering method, leveraging ensemble clustering-based optimization techniques, to select cluster heads responsible for data aggregation. The combination of optimal clustering and agent-based attack detection enhances the overall security and performance of the MANET. This also significantly improving energy efficiency and data aggregation reliability. Each cluster in the proposed model is equipped with an attack detection agent node, which plays a critical role in identifying suspicious, blackhole, wormhole, and normal nodes within the incoming traffic. This proactive detection mechanism ensures timely response and mitigation of potential security threats. The development of an ensemble-based clustering optimization technique to enhance energy efficiency and improve data aggregation. In addition to the detection mechanism, the study performs a comprehensive comparison of multiple machine learning algorithms. This comparison aims to determine the most effective models for accurate attack identification and trust score generation for network nodes. This determines the most effective algorithms based on accuracy and computational cost by enabling more accurate threat identification and trust-based routing decision. By combining agent-based attack detection, energy-efficient clustering, and intelligent machine learning models, this research work offers a comprehensive and robust solution to enhance the security and reliability of MANETs. Experimental results on simulated MANET environments demonstrate that the proposed approach significantly improves detection accuracy and enhances network lifetime and throughput compared to existing methods. Specifically, proposed approach achieved a throughput of 93 Kbps. This shows approx. 8% improvement over existing approach. The results demonstrate the effectiveness of the proposed approach in providing valuable insights for future research in securing MANETs.

Figures