Performance analysis of optimized intelligent intrusion detection system for wireless sensor networks

Abstract

One of the key elements of contemporary electrical and wireless systems is the Wireless Sensor Network (WSN). For the purpose of discovering sensor networks and using functions like data sensing, data processing, and communication, a WSN is made up of several sensor nodes. These networks are extremely important in the field of medical healthcare for transferring and gathering extremely private data from various geographical locations. However, the worry about various attacks on health care data normally grows daily. These attacks may have adverse impacts on the WSN nodes in a relatively short amount of time. Additionally, the current Intrusion Detection System (IDS) has problems with its resource limitations, poor detection rate, significant processing overhead, and increased false alarm rates when identifying various threats. In the first phase of the research, Machine Learning (ML) classifiers like quotSupport Vector Machines (SVM), Decision Tree (DT), K-Nearest Neighbor (KNN), and Artificial Neural Network (ANN)quot are used to detect various WSN attacks in order to address all of the aforementioned problems. It has been shown that ML modules can enhance detection performance. Numerous performance metrics, including quotaccuracy, sensitivity, specificity, and F1-score,quot are generated for comparative study. According to a thorough assessment of these modules, the detection efficiency may be improved even more in terms of performance metrics. The ML classifiers averaged approximately 86% accuracy, 80.2% sensitivity, and 77.3% specificity when attacks increased. However, this architecture is still improvised to manage massive amounts of data. newline