Network Traffic Classification for Internet of Things Based on Deep Learning Models

<p>Internet of Things (IoT) is a system of interconnected computing devices. The continuous growth of the number of IoT devices leads to expansive traffic. It is crucial to study the behaviour of network flows for Internet Service Providers (ISPs) to manage the performance of the IoT network. The Network Traffic Classifier (NTC) is the foremost essential tool in finding the network flows and behavioural aspects of a network such as network latency, volume, bandwidth consumption and many more. The success of deep learning models extended to the NTC as well. The current deep learning based solutions for the NTC contributed to considerable success. However, the current so- lutions are proposed to classify the flows that are captured in monitored network. In the real world, IoT traffic is diverse and heterogeneous in nature. Therefore in the ever-evolving IoT world, it is challenging to keep up the classification model trained with flows that are captured in controlled environment. Hence, in this research the effort is made to design the model that can classify the traffic flows from real world. A supervised deep learning method is proposed to classify network traffic and chi-square algorithm is used to select the features that can provide best information about the flows. The proposed method achieves 70% accuracy. The time distribution wrapper over the Convolutional Neural Network (CNN) is employed to extract the network features. The Long-Short Term Memory (LSTM) layer is applied to classify the network flows. The thesis explains a detailed study of feature engineering, successful deep learning models, and the research results. </p>