Edge Computing for IoT Security: Integrating Machine Learning with Key Agreement

The Internet of Things (IoT) is one of the most widely used technologies today, the number of IoT connections growing by 18% to 14.4 Billion active endpoints globally. Thus, with such a large scale IoT application, security becomes the focus. The issue of intrusion and interception has been a controversial and much disputed subject within the field of IoT security. The main limitation of IoT security, however, is a single security technology cannot ensure security of the IoT system. The Internet of Things requires diverse technologies applied at different layers to work together to ensure security. Based on these problems, proposed a security solution with integrity is necessary. Edge computing is a great way to solve security problems for resource-constrained devices. This paper proposes a machine learning and cryptography combined IoT security scheme at edge computing, achieves monitoring and detection abnormal behaviors of IoT system and automatically responds at an early stage. Designed a lightweight key management scheme without third-party involvement. The combination of active detection and passive protection for IoT security is achieved. Meanwhile, data for this study were collected using Edge-IIoTset dataset for the prediction of IoT security threats in edge computing networks. In this paper, experiments offers some important insights into the model selection for the prediction of time series data. The models selected include Support Vector Machine(SVM), K-Nearest Neighbor(KNN), and Long Short-Term Memory(LSTM) models. The significance of the comparison of several models is to assess their accuracy in detecting IoT communication data. To compares the performance of models, the following performance evaluation parameters were used: Precision, Recall and F1-Score. This study has identified LSTM has better precision in time series data such as communication data, 10% more precision than the other models used in this paper.