Automated Attribribution of Multi-Layered Threat Actor Infrastructure Using Hybrid Graph Neural Networks and Adaptive Hyperparameter Optimization

**Abstract:** This paper introduces a novel framework for automated attribution of sophisticated, multi-layered threat actor infrastructure. Existing attribution methodologies often struggle with the dynamic nature and obfuscation tactics employed by modern threat actors, particularly those operating across decentralized platforms. Our system, utilizing a hybrid Graph Neural Network (GNN) incorporating both Structural and Relational message passing, coupled with adaptive hyperparameter optimization based on Bayesian optimization, achieves a significantly improved accuracy and speed in identifying key actors and their roles within complex threat networks. The system's architecture is designed for immediate commercialization within cybersecurity threat intelligence platforms, offering quantifiable improvements in proactive defense capabilities.