Date of Publication :25th September 2023

Abstract: Wireless Sensor Networks are one of the base platform networks highly demanded by various real time applications recently. WSN has gained more attraction for various applications following the advancements of sensor devices and technologies. Agriculture, surveillance monitoring systems, healthcare monitoring, and smart environment are some of the applications fast-growing applications nowadays. Security is one of the major issues that remain in WSNs. Different kinds of malicious attacks are created dynamically anywhere in the WSN at any time. In any network, malicious activities destroy the data transmission process and compromise the other legitimate nodes. Several earlier pieces of research stated that WSN meets security issues because of its restricted infrastructure and physical security. Any sensor can communicate with other sensors, creating an opportunity for various vulnerable attacks affecting legitimate network nodes. Since nodes are tiny, inexpensive, and easy to deploy anywhere without any constraints identifying malicious nodes during the deployment stages is impossible. Various earlier research works have proposed security mechanisms that could not effectively detect malicious threats and attacks. The detailed literature of several proposed methods using machine learning algorithms in the past 10 years is explained here. It is also presented a comparative analysis of different machine learning algorithms used for DoS attack detection in WSN with simulation results. The accuracy comparison is given, and it shows that the SVM model outperforms Random Forest, Logistic Regression, and Decision trees.

Reference :