The charming city of Prague recently hosted a vibrant exchange of ideas at the second meeting of the Special Interest Group on Artificial Intelligence for NRENs (SIG-AI), held on the premises of CESNET, the Czech National Research and Education Network. This gathering, which preceded the GÉANT Security Days, centred on the increasingly critical intersection of Artificial Intelligence and Cybersecurity.
The day commenced with an engaging Open Brainstorming session on AI, facilitated in the morning before the official SIG-AI meeting began. During this hour, attendees actively exchanged information and explored new avenues for the SIG-AI to investigate. Key topics included the necessity for a more strategic approach to AI initiatives within the GÉANT project, a comprehensive mapping of AI use cases currently underway in participating organisations, and the development of mechanisms to foster and support collaborative efforts. Data sharing emerged as a significant challenge for NRENs in adopting AI, prompting discussions around creating a trusted environment for data exchange. Furthermore, the community highlighted the importance of addressing regulations, ethics, AI standards for quality assurance, skills development, and potential risks such as gender bias in data.
Following the brainstorming, the official SIG-AI meeting kicked off with a warm welcome from Leonie Schäfer (DFN), the coordinator of the SIG-AI, and Daniela Brauner (GÉANT), the chair of the steering committee, who outlined the objectives and presented the roadmap of planned activities and meetings for 2025. The participants were then formally welcomed by Jakub Papírník, the CEO of CESNET.
The agenda featured insightful presentations, beginning with an overview of AI projects from the host, CESNET. Martin Zadnik (CESNET) showcased diverse applications of machine learning, ranging from creating mitigation rules for DDoS attacks to identifying crypto-mining flows for GPU management. Jan Kohlrausch (DFN-CERT) provided a crucial overview of AI-driven security threats. Magdalena Rzaca (GÉANT) underscored the importance of European AI Act compliance even in open-source development or testing affecting real users, and introduced the concept of an AI Sandbox for the GÉANT community to test AI solutions for research and education. Mary Grammatikou (NTUA, Greece) presented a compelling example of federated learning for handling data that cannot be moved from national sites, emphasising the concept of explainability.
The dynamic lightning talks session further enriched the discussions. Maarten Meijers (Intern at GÉANT and University of Twente) introduced Graph Neural Networks as a promising model for anomaly detection and flow categorisation, while Jakub Kwiatkowski and Maksymilian Marcinowski (PSNC) presented Red Teaming LLM Vulnerabilities.
Subsequently, Nicole van der Meulen and Albert Hankel from SURF provided insightful points to steer the ensuing breakout room discussions, where attendees delved deeper into specific aspects of AI and cybersecurity. The groups then reconvened to share their findings. The SIG-AI meeting concluded with the customary Mentimeter survey, followed by a summary of the day’s key discussions delivered by Daniela Brauner. The productive day culminated in a Joint Dinner at Kolkovna Celnice, a traditional restaurant situated in the historic heart of Prague.
SIG-AI session at the GÉANT Security Days
A significant highlight following the SIG-AI meeting was the dedicated session organised by the SIG-AI within the GÉANT Security Days. This session continued the focus on the critical intersection of AI and Cybersecurity in the form of a panel discussion.
Looking Ahead
The next opportunity for the SIG-AI community to convene will be at TNC25 in Brighton, with the next SIG-AI meeting scheduled on 9 June 2025 at 14:00-17:30 BST in the Renaissance Suite. The discussions initiated in Prague are sure to continue and evolve in this upcoming forum.
