Prediction Models to Effectively Detect Malware Patterns in the IoT Systems

Abstract

The Widespread use of the Internet of Things (IoT) has influenced many domains including smart cities, cameras, wearables, smart industrial equipment, and other aspects of our daily lives. On the other hand, the IoT environment deals with a massive volume of data that needs to be kept secure from tampering or theft. Detection of security attacks against IoT context requires intelligent techniques rather than relying on signature matching. Machine learning (ML) and Deep Learning (DL) approaches are efficient to detect these attacks and predicting intrusion behavior based on unknown patterns. This study proposes the application of five deep and ML techniques for identifying malware in network traffic based on the IoT-23 dataset. Random Forest, Catboost, XGBoost, Convolutional Neural Network, and Long Short-Term Memory (LSTM) models are among the classifiers utilized. These algorithms have been selected to provide lightweight security systems to be deployed in the IoT devices rather than a centralized approach. The dataset was preprocessed to remove unnecessary or missing data, and then the most significant features were extracted using a feature engineering technique. The highest overall accuracy achieved was 96% by applying all classifiers except LSTM which recorded a lower accuracy.