Design and Performance Analysis of Data Security Algorithms in Iot Networks

Abstract

In recent years, the internet of things (IoT), 5G, and artificial intelligence (AI) continue to converge the various security incidents and occurrences. Since more devices are connected to IoT networks, security is becoming a significant concern. Conventional intrusion detection systems (IDS) are not well suited for use in the complex lightweight IoT message queuing telemetry transport (MQTT) environment. The homogenous IoT-MQTT device-based Raspberry pi MQTT dataset (RPIMQTTSET) is developed to record regular and attack features for three scenarios: no attacks were involved in both subscriber or broker, subscriber under attack, and broker under attack. An improved critical feature selection algorithm (ICFS) and an ensemble learning algorithm (EL) is proposed for detecting attacks in real time. The accuracy of 99% and 92.68% were achieved using random forest (RF) on training and testing, respectively, which is 1.1% greater when compared with the literature method. The ICFS algorithm has been applied to XIIoTDataset to reduce the features and the outcomes of performance metrics have been achieved better when compared to the original dataset newline