Cross-network Intrusion Detection Based On Selfattention Mechanism Hybrid Neural Network

With the rapid development of the Internet of Things(IoT), the network environment is becoming increasingly complex, and attack methods are becoming more diverse, posing a huge challenge to cyberspace security. Existing research on network intrusion detection has the problems of incomplete feature extraction, reliance on outdated datasets, and a lack of cross-network intrusion detection. A cross-network intrusion detection method based on a self-attention mechanism Hybrid Neural Network is proposed. The method consists of five parts: the Data Awareness Layer, Feature Extraction Layer, Sequential Information Fusion Layer, Global Context Attention Layer, and Intelligent Decision Layer. To address the problem of incomplete feature extraction, this paper uses a convolutional neural network in the Feature Extraction Layer to extract the spatial features of traffic data; in the Sequential Information Fusion Layer, a bidirectional long short-term memory network is used to capture the contextual temporal dependencies of traffic data in both directions; and a self-attention mechanism is introduced into the Global Context Attention Layer to achieve focused attention on key features. In view of the problems of relying on outdated datasets and the lack of cross-network intrusion detection, this paper trains and tests the model using traditional Internet datasets and IoT datasets that were released relatively late. The experimental results show that this method is robust when faced with different types of datasets and performs well in various Evaluation Metrics.