IoT NETWORK TRAFFIC ANALYSIS USING MACHINE LEARNING BASED TECHNIQUES

Abstract

Despite the advantages of the IoT, a rising amount of malware designed specifically for IoT devices poses a serious danger to the Internet's environment. These malware attacks have created the need to evaluate the IoT system’s security and the need to create defenses against potential threats. It is crucial to stop IoT malwares from spreading. Furthermore, for educating people and evaluating the accuracy of cyber security the gathering and researching of data from many sources of IoT data is essential. Regarding this, we need to analyze the network traffic. Previous research papers suggested some models, which is based on Long Short-Term Memory (LSTM), successfully completes two tasks 1) Identifying the benign nature of the given traffic and 2) Identifying the sort of malware to look for in malicious network data. For this, there is need for a sizable amount of traffic data from the number of files of both good and bad traffic which can be gathered from different distinct IoT devices. Flowrelated, traffic flag related, and packet payload related characteristics were the three modalities into which the features that were retrieved from the datasets at the feature and modality levels, a feature selection technique was used, and the best modalities and features were applied for performance improvement. After we apply a number of Machine Learning algorithm for analyzing the traffic to find if it is benign or malicious.