Deep Transfer Learning-Based Traffic Detection Technology for IoT Devices in Power Grid Integration

With the rapid evolution of Internet of Things (IoT) technology, guaranteeing the secure access of terminal devices to the network has become a core problem that needs to be addressed urgently. To address the current situation of Electric Internet of Things (EIOT) facing attacks due to the complex device environment, large-scale traffic and limited terminal computing power, this paper designs an intrusion detection method based on transfer learning and sustainable evolution classifier. This method extracts and analyzes high-level features from massive IoT network traffic, and then uses a Nearest Class Mean (NCM) classifier to achieve precise pattern recognition and classification. The implementation of freezing and decoupling strategies allows for the fine-tuning of the feature extractor and classifier using a limited amount of sample data,this process enables the model to rapidly and accurately identify abnormal traffic in the dynamic and intricate EIoT environment. Four migration schemes have been systematically designed and comparatively analyzed to select the optimal scheme. Finally, We evaluate the performance of the proposed intrusion detection system under scenarios of few samples, data imbalance, and emerging unknown attacks, thus validating its applicability in EIoT contexts.