Enhanced SVM and RNN Classifier for Cyberattacks Detection in Underwater Wireless Sensor Networks

Abstract

Researchers have been paying greater attention to Underwater Wireless Sensor Networks (UWSN) lately because of their advancements in ocean surveillance, application deployment, and marine monitoring. However, because of its intrinsic qualities, this sensor network is susceptible to several kinds of cyberattacks, including Active Attacks, Sybil Attack, Denial-of-Service (DoS), Passive Attacks and Traffic Analysis. The Sybil assault is one of the deadliest cyberattacks and causes significant network damage among other attacks. This research proposes an intelligent techniques model-based cyber-attack detection system that combines deep learning and machine learning technologies for identifying cyber-attacks. Additionally, a feature reduction approach using machine learning methods Support Vector Machine (SVM) and Principal Component Analysis (PCA) is used to identify the attributes that are most strongly linked to the chosen attack categories. The study assesses the accuracy of a suggested Recurrent Neural Network (RNN) an algorithm for classifying and detecting intrusions that are based on deep learning. The proposed system achieves (97%) accuracy after dimensional reduction and optimization. This study will help the researchers design the routing protocols to cover the known cyber-attacks and help industries manufacture the devices to observe these cyberattacks, which could reduce the possible attack chances in UWSN communication.