Intrusion Detection Method for Power Internet of Things Based on Data Hybrid Detection Model Considering the Spatio-temporal Characteristics

The wide application of the Internet of Things brings convenience to people, but also causes many security problems, so it is urgent to establish a complete and stable system to ensure the security of the Internet of Things. Intrusion detection system has become the key technology to protect the security of the Internet of Things. Unlike the traditional intrusion detection mechanism, intelligent intrusion detection technology can fully extract data features and has higher detection efficiency, but the requirements for data sample labels are also higher. Therefore, we propose a hybrid data detection model considering the spatio-temporal features. In that process of data sampling, based on the ADASYN algorithm, appropriate minority class data are synthesized according to the feature space condition of the data set. Then, the density distribution of the feature space is improved through the generation of the synthesized data. Finally, we use the Temporal Convolutional Network (TCN) to extract the spatial characteristics of data traffic. The Bidrection Gated Recurrent Unit (BiGRU) model is also used to extract the time characteristics of data traffic, and finally to detect the intrusion anomaly data. The experiment is verified by the NB15-IoT data set, Bot-IoT data set and NSL-KDD data set, with the results showing that the proposed model has higher accuracy, precision and F1 value. Thus, it has strong generalization ability and good compatibility with various data.