Attack Detection in Wireless Sensor Networks using Novel Artificial Intelligence Algorithm

The Wireless Sensor Network (WSN) is a concept of an Internet of Things network (IoT) with limited energy resources. When many sensors are used despite their limitations, security becomes a significant concern in the network, necessitating the practical design of detection and mitigation techniques. Using Artificial Intelligence (AI) algorithms is a very effective way to develop cyber-attack detection systems in WSN. Squirrel Search Optimized Probabilistic Neural Network (SSOPNN), a unique classification approach for the detection of cyberattacks in WSN, is presented in this research. Gaussian Naïve Bayes (Gaussian NB), K-Nearest Neighbors (KNN), and Random Forest (R.F.) are three well-known methods against which their results are compared. The study utilizes Pearson's correlation and mutual information to choose features for the ensemble and to reduce the number of dimensions. T Pipeline optimization technique with a tree structure is used to fine-tune hyperparameters. Specifically, one-of-a-kind WSNDS dataset is used, which provides adequate material for four distinct attacks (Grayhole, Blackhole, Flooding, and TDMA scheduling). The study compares three approaches using a variety of metrics, including detection rate, false positive rate, false negative rate, average prediction time per sample, and memory and processing time requirements.