Chaire Cyber CNI

Chaire Cyber CNI – Cybersecurity for Critical Networked Infrastructures

The “Management of Complex Threats” workshop at IEEE/IFIP NOMS 2026 was a full success

Great discussions today at the “Management of Complex Threats” workshop at IEEE/IFIP NOMS 2026 in Rome!

Together with Pierre Parrend and colleagues from IMT Atlantique and EPITA, and with the support of the CyberSecDome Horizon Europe project, we brought together around 20 researchers and practitioners from across Europe and Asia to discuss the future of AI-driven cybersecurity and the management of complex threats.

The workshop showcased an impressive breadth of topics:
• uncertainty-aware traffic classification from researchers at Czech Technical University in Prague
• adaptive IoT anomaly detection from National Institute of Informatics
• federated security gateways for LLM agents from Hochschule Furtwangen University
• autonomous AI-driven penetration testing from University of Calabria
• LLM-based honeypots from researchers in France
• cyber-physical attacks against cooperative robot systems from Birla Institute of Technology and Science Pilani
• and topology-based cyberattack detection for water distribution systems involving University of Strasbourg, Karlsruhe Institute of Technology, and EPITA.

I was also particularly happy to see strong contributions from the broader Cyber CNI chair ecosystem at IMT Atlantique:

Mathis Durand (Cyber CNI, CNRS UMR IRISA, IMT Atlantique) presented joint work with Yvon Kermarrec and Marc-Oliver Pahl on SSH honeypot realism and fingerprinting resistance. Their work analyzed how ethical hackers detect deception systems in practice and derived a taxonomy of effective honeypot detection techniques.

Mohammed Mezaouli (CNRS Lab-STICC, IMT Atlantique) presented joint work with Yehya Nasser, Samir Saoudi, and Marc-Oliver Pahl on AI-based real-time anomaly detection for embedded C functions using instruction-level traces and current measurements — an exciting step toward protecting resource-constrained IoT systems against zero-day style attacks.

One recurring theme across the workshop was clear: AI is both the attack surface and the defense mechanism. From semantic attacks against LLMs to autonomous penetration testing and trustworthy AI-driven defense systems, future 6G, IoT, and industrial infrastructures will require fundamentally new approaches to resilience and operational security.

A big thank you to all speakers, authors, and participants for the technically deep discussions and excellent atmosphere throughout the day!

Industrial Networking Beyond Connectivity: Key Scientific and Cybersecurity Challenges Discussed at NOMS 2026

Industrial networking is reaching a turning point.

We are moving from isolated PLC-centric systems toward distributed AI-assisted cloud-edge infrastructures running industrial protocols over IP networks.

But here is the uncomfortable question:

Can probabilistic IT and cloud infrastructures ever truly satisfy deterministic industrial requirements?

At our NOMS 2026 panel on *“Next Steps of Industrial Communications and Networks”*, we debated topics such as:

* industrial protocols over Layer 3/IP,
* strict SLA enforcement,
* virtualized PLCs,
* AI-driven operational management,
* cloud-edge continuums,
* and the growing cybersecurity risks of converged IT/OT infrastructures.

One insight from the discussion stayed with me:

**Risk = Probability × Impact**

In industrial systems:

* the probability of failures or attacks on complex IP-based infrastructures is non-negligible,
* while the impact can be catastrophic.

That fundamentally changes the networking equation.

My conclusion:
Industrial networking is no longer merely a connectivity problem.
It is becoming a **trustworthy autonomy and resilience problem**.

And perhaps the most controversial question:
Will AI become the only realistic way to manage the operational complexity we are creating ourselves?

I summarized the key scientific debates, disagreements, and cybersecurity implications from the panel in a longer article below.

I would genuinely love to hear different opinions from academia and industry:
Are today’s IP/cloud-native architectures sufficient for future industrial systems — or do we need fundamentally new networking paradigms?

#NOMS2026 #IndustrialNetworking #Industry40 #Industry50 #Cybersecurity #IndustrialAI #CloudNative #EdgeComputing #OTITConvergence #Resilience #AI

With:Xinze Li (Southeast University, China), Renwei “Richard” Li (Southeast University, China), Luca Foschini (University of Bologna, Italy), Marc-Oliver Pahl (IMT Atlantique, France), Stefano Salsano (University of Rome Tor Vergata, Italy), Jose Fontalvo-Hernandez (Siemens AG, Germany), Xipeng Xiao (Huawei Germany, Germany), Luis Miguel Contreras Murillo (Telefónica, Spain)

AIMLOps 2026 Workshop at IEEE/IFIP NOMS 2026 a Great Success

🚀 AI is rapidly transforming from a support tool into the operational core of future communication systems — and this was exactly the focus of the 1st AIMLOps Workshop at IEEE/IFIP Network Operations and Management Symposium 2026 in Rome.

Co-organized by Marc-Oliver Pahl, Hanan Lutfiyya, and Stuart Clayman, the workshop brought together researchers and practitioners from across Europe, North America, and Asia to discuss the future of autonomous and AI-driven network and service management.

The workshop saw strong engagement throughout the entire day with more than 100 participants distributed over the four sessions.

What stood out most was the exceptional quality of the papers and the depth of the discussions. Topics ranged from:
🔹 AI-driven cybersecurity and anomaly detection
🔹 Explainable AI (XAI) for trustworthy operations
🔹 MLOps and LLMOps for AI lifecycle management
🔹 Multimodal LLMs for troubleshooting and network management
🔹 Autonomous protocol translation for IoT interoperability
🔹 AI assurance for distributed AI infrastructures
🔹 Sovereign European cloud and AI infrastructures

A particularly strong contribution from the [Cyber CNI Chaire](https://cybercni.fr?utm_source=chatgpt.com) ecosystem came from Mohammed Mezaouli (CNRS Lab-STICC, IMT Atlantique) together with Yehya Nasser, Samir Saoudi, and Marc-Oliver Pahl. Their presentation, *“Closing the Loop in Embedded Security: Evolution of an AIOps Framework for Threat Hunting,”* explored how AIMLOps methodologies can secure resource-constrained embedded systems in future 6G and IoT environments through adaptive AI pipelines, side-channel telemetry analysis, and continuous model evolution.

The keynote sessions perfectly complemented the technical program.

🎤 Deep Medhi (National Science Foundation) delivered the keynote *“When Networks Think for Themselves: Who Is in-charge?”* and challenged the community to rethink trust, explainability, and control in future autonomous and zero-touch networks where AI systems increasingly make operational decisions themselves.

🎤 Hans Torben Löfflad and Manuel Hoffmann from STACKIT presented *“A European Data & AI Strategy – A Cloud Provider’s View.”* Their keynote addressed sovereign European cloud infrastructures, large-scale AI model serving, Kubernetes-based AI platforms, and Europe’s path toward digital sovereignty.

This discussion was particularly timely as STACKIT was recently selected by the European Commission as one of the providers for the EU Sovereign Cloud procurement initiative — a major milestone for resilient and sovereign European digital infrastructures and highly relevant for research on cyber-secure critical infrastructures.

A huge thank you to all authors, keynote speakers, participants, and organizers for making AIMLOps 2026 such an engaging and inspiring event. The workshop clearly demonstrated that AI lifecycle management, operational trustworthiness, resilience, and sovereignty are becoming defining challenges for the future of networking and distributed systems.

#AIMLOps #NOMS2026 #AI #AIOps #MLOps #LLMOps #AutonomousNetworks #Cybersecurity #DigitalSovereignty #SovereignCloud #6G #IoT #NetworkManagement #GenerativeAI #CloudComputing #CyberCNI

View More