Cluster based Malicious Node Detection in Wireless Multimedia Sensor Networks

Abstract

Wireless Sensor Networks (WSN) consist of a group of small, newlinelow-power sensors that can be deployed over vast areas for various newlineapplications. With the inclusion of CMOS cameras and microphones, the newlinesensors have advanced to form Wireless Multimedia Sensor Networks newline(WMSN). This technology allows for the collection and wireless newlinetransmission of multimedia data, such as audio, images, and videos, to a newlinebase station for analysis and processing. The introduction of WMSN has newlinerevolutionized the conventional WSN, making it more versatile for newlineapplications in environmental monitoring, healthcare, and surveillance newlinesystems. Despite its advantages, transmitting significant amounts of newlinemultimedia data in real-time over a low-speed wireless connection poses newlineseveral challenges, including the risk of malicious nodes in the sensor newlinenetwork. newlineTo safeguard the privacy and reliability of wireless network newlinetransmissions, it is paramount to proactively prevent malicious nodes newlinefrom causing interference. These nodes, susceptible to cyber-attacks or newlinephysical tampering, can compromise data confidentiality and network newlineperformance. Hence, implementing strong security measures to detect and newlineneutralize such nodes is critical. Doing so ensures the network s newlinetrustworthiness and security, shielding against any potential privacy newlinebreaches or unauthorized access to sensitive information. newlineThe objective of this research is to detect malicious nodes that newlinecould affect the network performance and to propose three methods that newlinecan enhance network efficiency and conserve energy. The first approach newlinevi newlineentails utilizing a genetically optimized feature index to identify and newlineneutralize malicious nodes. Next, the feature set is classified using an newlineadvanced LeNET deep learning approach. However, since this can be newlinetime-consuming, the suggested technique replaces the dense layers of newlineLeNET architecture with the fuzzy C-means (FCM) algorithm while newlineadjusting the inner layers.