Enhancing IoT Security Through Resource-Efficient Feature Engineering

As a heterogeneous networks, the Internet of Things (loT) is vulnerable to attacks, which brings a threat to human life and property. Intrusion Detection Systems (IDS) based on machine learning, a powerful tool have been developed and applied in network security. However, the deployment of highly complex IDS should consider the resource cost. Data preprocessing and feature engineering are capable of enhancing the performance and resource cost with effectiveness and low cost. Thus, this paper focuses on feature engineering which is divided into 4 steps: elimination of spurious features, data splitting, labeling and normalization, and feature selection. Through the feature reduction, we generate a dataset with different dimensions on the TON IoT dataset. To evaluate the effectiveness and universality of the proposed method, four machine learning methods including gradient boosting machine (GBM), K-Nearest Neighbors (KNN), Random Forests (RF), and MultiLayer Perceptron (MLP) are implemented into the simulation in Python. The comparative results indicate that Random Forest performs optimally in terms of feature engineering, achieving accuracies of 99.96 % and 99.52 % in binary and multiclass experiments, respectively.