A Systematic Review of AI-Driven Intrusion Detection and Performance Optimization in Wireless Sensor Networks

Abstract

The safety and reliability of Wireless Sensor Networks (WSNs) depend on the crucial function that Intrusion Detection Systems (IDS) perform in their operations. The current security systems face major obstacles because attackers continuously launch cyber operations against increasingly complex networks. Through the development of Artificial Intelligence (AI) technologies, which include Machine Learning (ML) and Deep Learning (DL) methods, IDS systems could now identify both standard and novel cyber threats. The study analyzes the most recent progress in ML and DL methods used to develop IDS that operate in WSNs through analysis of their primary algorithms and algorithmic combinations. The study conducted an extensive literature review by accessing the SCOPUS database to identify relevant studies published between 2021 and 2026. The systematic review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. As revealed by the results, the DL and hybrid approaches are superior to conventional ML algorithms in handling complicated and imbalanced datasets. Some of the accuracy rates observed are 99.94% when using KMeans-SMOTE, and 99.76% when using K-nearest neighbor (KNN). However, Deep Neural Networks (DNN) and Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) networks demonstrate relatively low accuracy levels of 96.23% and 97%, respectively. Most approaches use specific datasets, including WSN-DS and NSL-KDD, which makes the results environment-specific. Besides, issues such as high computational power, data imbalance, absence of standardized datasets, and implementation constraints underscore the need for a scalable and adaptable IDS for WSN.