Edge Intelligence-Enabled Network Intrusion Detection in Compute First Networking

The proliferation of smart devices in recent years has fostered unprecedented levels of connectivity, giving rise to a novel paradigm known as Compute First Networking (CFN). This paradigm is designed to optimize and efficiently utilize both network and computing resources. However, the proliferation of network intrusions poses security challenges. Deep learning (DL)-based Network Intrusion Detection Systems (NIDS) can be hardly supported by edge smart gateways. Edge computing enables near-real-time analysis of network intrusions by offloading AI-intensive tasks from edge gateways to cloud servers. Coordinating computing task offloading among multiple edges is challenging due to varying channel conditions and stringent latency requirements of NIDS. To address this challenge, we propose a multi-edge collaborative framework for NIDS. Given the complexity of AI-enabled NIDS, we introduce a delay-aware computational offloading strategy to minimize latency. In addition, inspired by the non-Euclidean nature of network data and its abundant chronological and temporal relations, we design a Node-Level Attention-Based Auto-Encoder Graph Convolutional Network (NAAE-GCN) to capture latent behavioral patterns of evolving network attacks. Experimental results show that the proposed NAAE-GCN achieves new state-of-the-art results in the detection performance with significant improvements over existing approaches.