Advanced Persistent Threat Detection Using an Improved Graph Convolutional Network with Big Data Network Log Analysis

In the modern era of big data, network security systems face significant challenges in detecting sophisticated cyber threats such as Advanced Persistent Threats (APTs). Traditional intrusion detection systems often fail to capture complex relationships and multi-stage attack patterns present in large-scale network environments. This study proposes an enhanced network security framework based on Relational Graph Convolutional Networks (RGCN) for big data analysis. The system transforms network logs into graph structures representing interactions among entities such as users, hosts, and processes. By integrating subgraph sampling, ensemble learning, and adaptive thresholding, the proposed model improves both detection accuracy and computational efficiency. Experimental results on the DARPA TC dataset demonstrate high detection accuracy exceeding 94% at both graph and node levels, with an average detection time of approximately 19 seconds. The framework provides a scalable, efficient, and interpretable solution for real-time APT detection and strengthens modern cybersecurity systems.