Intelligent Identification of Security Threats in Long-Distance Communication Networks Based on Embedded AI Perception

In the context of network security, when threats exhibit spatial similarities — such as attacks originating from the same or nearby geographic locations or using similar methods within a specific region — it can complicate the identification of security threats in long-distance communication networks. To address this challenge and improve both the efficiency and accuracy of threat identification, we propose an intelligent method for detecting long-distance communication network security threats based on embedded AI perception. This method takes spatial similarities into account by comparing sensing data between nodes within the network to assess their real-time status. Using a group decision-making mechanism, we correlate the data collection points of vulnerable network nodes to gather security log data. The network logs are then preprocessed using the Grubbs statistical algorithm to extract key security threat features. By incorporating machine learning algorithms from the field of AI, and leveraging the edge computing capabilities of the NVIDIA Jetson series as the embedded AI platform, we construct an intelligent perception model to detect security threats in real-time. Experimental results demonstrate that the proposed method can effectively detect the actual state of different nodes within the network, with a maximum state deviation of only 0.2. Additionally, the method achieved an [Formula: see text]1 score of 0.986, a recall value of 0.979, and a precision value of 0.981, indicating strong robustness and comprehensive threat identification performance.