Comparison of Interaction Profiling Bipartite Graph Mining and Graph Neural Network for Malware-Control Domain Detection

In the rapidly evolving realm of cybersecurity, the detection of malicious domains stands as a critical challenge. Traditional methodologies, reliant on expert-driven feature engineering, are increasingly strained against the dynamic tactics of cyber-criminals. This paper introduces a novel approach utilizing Graph Neural Networks (GNNs) to enhance the detection of malicious domains. By leveraging un-supervised representation learning techniques, such as Deep Graph Infomax, we transform network traffic data into graph data models, thereby reducing reliance on domain expert input for feature identification. Our method demonstrates marked improvements in domain name classification using real-world data. This research contrasts the new data-driven approach with traditional methods, high-lighting its superior adaptability, reduced dependency on expert knowledge, and potential for broader application. The findings underscore the efficacy of GNNs in cybersecurity and open avenues for future research in applying advanced ma-chine learning techniques to cyber threat detection.