PhD Position: Applications of Graph Foundation Models in Cybersecurity

Inria is seeking a doctoral candidate to join the PIRAT research team at the University of Rennes to explore the potential of Graph Foundation Models (GFMs) in cybersecurity. The research focuses on leveraging pre-trained models to address critical security challenges, including network flow classification, system log anomaly detection, and malicious binary identification. The successful applicant will work on designing effective inference strategies, fine-tuning existing models, and potentially developing new architectures to improve performance in security-specific tasks. This position offers the opportunity to contribute to the intersection of machine learning and cybersecurity, aiming to democratize advanced ML tools for security professionals.

• Conduct research on applying Graph Foundation Models to cybersecurity tasks like network flow classification and log anomaly detection.
• Identify and implement efficient methods for graph construction and task formulation for GFM applications.
• Develop inference strategies to maximize GFM performance on specific security-related tasks.
• Fine-tune existing GFM architectures or design new ones for specialized cybersecurity applications.
• Optimize models for computational efficiency through techniques such as distillation or parameter quantization.