Publication

Abstract : In recent years, there has been a proliferation of IoT (Internet of Things) devices and its enabling technologies in industries, product flow management, healthcare, transportation and other smart environments. The provision of IoT devices with IP (Internet Protocol) address allows for communication between these cyber-physical systems without any intervention. Lack of security on these end devices has led to many attacks like denial-of-service, Botnets, identity theft and data theft attacks. Mirai, Torii, Emotet, Dridex, Trickbot, Glupteba and QBots are all example of IoT vulnerabilities. Artificial intelligence enables to build strong secure cyber-physical systems. Machine learning and Deep learning techniques combats cyber-attacks by detection and prevention of these Botnets. In this paper, we explore on Botnet attacks that is prevalent in IoT devices because of lack of standards in security when manufacturing or illiteracy of security awareness by the users. Machine learning and Deep learning are prospective solutions to detect and prevent cyber intrusions on IoT devices using anomaly detection. In this paper, firstly we present a framework for Botnet detection to detect an attack on real time traffic. Secondly, we compare and contrast different machine learning and deep learning algorithms for detection of Botnets on the standard features using the Aposemat IoT-23 dataset. Among the deep learning models Gated Recurrent Unit (GRU) efficiently detects Botnets with 99.87% accuracy. Thirdly, the raw packet captured (pcap) files form Aposemat IoT-23 dataset are analyzed for attacks using Wireshark tool. Then deep learning model GRU is used to detect the malware attacks with accuracy of 99.89 and better time complexity. Finally, we compare the models based on performance metrics like accuracy, precision, recall, F1-Score and time complexity to deduce the best model on both feature based and featureless IoT traffic.