Graph Neural Networks for Botnet Detection

Abstract

Bots and botnets are among the most prevalent cybersecurity risks, a topic of significant concern for security communities, and a persistent threat that continues to impact digital systems. The detection of botnets is crucial to ensure the security and reliability of these systems, as various detection methods and research have been conducted to mitigate their risks. In this master’s thesis, we have proposed a new architecture based on Graph Neural Networks (GNNs) for the detection of botnets, focusing on their temporal network traffic behavior. Our architecture effectively captures both the temporal dynamics of network traffic and the structural relationships between machines in a network. Experiments on a real-world dataset show that our proposed architecture is efficient and achieves 81.25% accuracy, which is superior results compared to other state-of-the-art approaches.