Cyberattack Detection in Wireless Sensor Networks using a Hybrid Feature Reduction Technique with AI and Machine Learning Methods

Abstract

Security of a Wireless Sensor Network (WSN) is crucial for preventing data sharing from intruders. This paper makes a suggestion for a machine learning-based intelligent hybrid model and AI for identifying cyberattacks. The security of a Wireless Sensor Network (WSN) guards against malevolent hackers cyberattacks on data, networks, and computers. The qualities that are most closely associated to the selected attack categories are also identified using a feature reduction algorithm (SVD and PCA) and machine learning methods. In order to reduce/extract features and rank them, this paper suggests using the K-means clustering model enhanced information gain (KMC-IG). A Synthetic Minority Excessively Technique is also being introduced. Intrusion prevention systems and network traffic categorization are the eventual important stage. The study evaluates the accuracy, precision, recall, and F-measure of a proposed deep learning-based feed-forward neural network algorithm for intrusion detection and classification. Three important datasets, namely NSL-KDD, UNSW-NB 15, and CICIDS 2017, are considered, and the proposed algorithm's performance is assessed for each dataset under two scenarios: full features and reduced features. The study also compares the results of the proposed DLFFNN-KMC-IG with benchmark machine learning approaches. After dimensional reduction and balancing, the proposed algorithm achieves high accuracy, precision, recall, and F-measure for all three datasets. Specifically, for the NSL-KDD dataset in the reduced feature set, the algorithm achieves 99.7% accuracy, 99.8% precision, 97.8% recall, and 98.8% F-measure. Similarly, for the CICIDS2017 dataset, the algorithm achieves 99.8% accuracy, 98.7% precision, 97.7% recall, and 98.7% F-measure. Finally, for the UNSW-NB15 dataset, the algorithm achieves 99.1% accuracy, 98.7% precision, 98.4% recall, and 99.6% F-measure.