Anomaly Traffic Detection in Smart Home Using Insight from Exploratory Data Analysis

The use of IoT devices is widespread across various industries. Yet, they often fall victim to cyberattacks such as creating botnets, hacked to obtain sensitive information about the lifestyle of the user, etc. To adequately safeguard them, understanding their behavioural patterns is crucial in detecting and preventing these attacks. This project utilises exploratory data analysis to analyse the network traffic of smart home IoT devices to gain insight into the fundamental network traffic characteristics and behavioural patterns of these devices. Several data visualisation methods were explored to present the result of the analysis. The insights from the EDA were used to inform the statistical attributes extracted from the network traffic to generate flow data. Three novelty detection models were trained using this data, with the Local Outlier Factor model being selected as the top performer, boasting a remarkable 99.87% overall accuracy and zero false negative rate. The main contributions of this study are threefold. Firstly, it provides valuable insights into the typical behavioural patterns of smart home network traffic. Secondly, it sheds light on the features that can be extracted from network traffic to build a robust machine learning or deep learning model for an intrusion detection system. Finally, it developed a novel detection model for intrusion detection systems.