Enhancing Intrusion Detection in Manets Using Canonical Correlation Analysis and Fuzzy Cognitive Adaptive System

Abstract

Ensuring strong cybersecurity in Mobile Ad Hoc Networks (MANETs) is essential due to their dynamic and decentralised characteristics, rendering them susceptible to sophisticated cyber Threats. This research paper presents an enhanced intrusion detection framework leveraging Canonical Correlation Analysis (CCA) for feature selection and a Fuzzy Cognitive Adaptive System (FCAS) for classification to address class imbalance and improve detection accuracy. The methodology includes critical data pre-processing steps, such as one-hot encoding, data normalization, and class balancing, used the Multi-Step Cyber-Attack Dataset (MSCAD). Performance analysis reveals that integrating CCA significantly boosts precision and accuracy across models, with FCAS achieving superior results. Notably, FCAS improved precision from 0.92 to 0.94 and accuracy to 96%, outperforming alternatives like Random Forest (RF) (91%) and Multilayer Perceptron (MLP) (93%). This framework demonstrated exceptional efficacy in identifying uncommon assault modalities such as "HTTP_DDoS" and "ICMP_Flood," reducing false positives and enhancing class separation. The findings underscore the efficacy of the suggested method, confirming its reliability as a solution for intrusion detection in dynamic and imbalanced environments.