A MULTI-LAYER PERCEPTRON BASED FUZZY LOGIC MODEL FOR DETECTION AND CLASSIFICATION OF BLACKHOLE ATTACKS IN MOBILE AD HOC NETWORKS

Abstract

In the rapidly evolving field of network security, the characteristics of Mobile Ad Hoc Networks (MANETs) that makes it advantageous has created some challenges and vulnerabilities. Its dynamic nature poses great challenges in implementing any security mechanism because there is no fixed topology and it has no central server. This research presents a hybrid detection model based on the combination of Multilayer Perceptron (MLP) and Fuzzy Logic to detect and classify blackhole attacks Manets. They are widely being deployed in critical applications such as military operations, disaster recovery etc. that require a decentralized and dynamic network structure; however, these same features render MANET susceptible to various cyber security threats like blackhole attacks due to limited bandwidth, resource constraints and the absence of centralized authority-hardened resources. The traditional detection methods, which include signature-based, anomaly-based and rule-based defence solutions, commonly provide a high false positive rate with not much adaptation to sample changes and are slow in catering the dynamic behaviour of MANETs. The proposed model integrates MLP's learning capabilities to recognize attack patterns with Fuzzy Logic's ability to manage uncertainty in network conditions, making the detection system more adaptable and accurate. The developed hybrid model will solve this issue and meet the needs of different network traffic data conditions by integrating the MLP part, which trains intricate and non-linear associations, and fuzzy logic, which addresses uncertainty and promotes interpretability, specifically in the unclear network settings. The NSL-KDD data set, which is commonly used in the study of intrusion detection, was used to test the model. The reason for selecting the NSL-KDD dataset is its comprehensive representation of network traffic features. The dataset contains 41 traffic attributes describing network connections. The results of the experiments prove that the proposed MLP-Fuzzy Logic model has an accuracy of 0.96, precision of 0.95, recall of 0.96, and an F1-score of 0.96 and has a false positive rate of 0.02. As the analysis shows, the application of the hybrid model is highly effective compared to single-technology applications like vanilla MLP And SVM, providing the sound and trustworthy solution to the problem of the blackhole attack detection in the MANET networks with resource limitations. However, the limitation of this study is that it is restricted to only blackhole attacks rather than other attack types present in MANETS like wormhole, grayhole and sinkhole attacks. For further enhancement of the security of MANETs in different threat scenarios, future work should try to extend the proposed hybrid approach to detect and classify a wider set of MANET attack types.