Program
Program
TX4Nets PROGRAM
May 24, 2026
Please find the detailed program of the workshop on IFIP Networking official website: https://networking.ifip.org/2026/index.php/program/detailed-program
Speaker: Dr. Behnam Shariati
Abstract
The modern AI revolution was not sparked by algorithms alone, but by data at scale. Most famously with ImageNet, whose introduction around 2010 and the subsequent deep learning breakthrough in 2012 demonstrated that access to large, well-structured datasets can fundamentally transform what is technically possible. Since 2020, Fraunhofer HHI has been conducting sustained research and making concrete advances toward turning data sharing into a practical reality for the optical networking community. We developed NOBS, a data-sovereign platform for comprehensive monitoring of multi-vendor infrastructure; produced multiple well-documented reference datasets; pioneered policy-enforceable governance frameworks for network data sharing; demonstrated privacy-preserving AI solutions, including federated learning, together with differential privacy and secure multi-party computation, for open and disaggregated ecosystems; and introduced the Optical Testbed Data Space (OTDS) as a key enabler for regulated cross-stakeholder data sharing.
While the community remains far from having large, well-structured datasets that would make AI for telecom networks truly trustworthy and carrier-grade, remarkable progress has been achieved. Our progress has been strongly accelerated by contributions from peer institutions, focused workshops, and growing engagement in standardization bodies and industry forums. In this talk, I will review these developments from a unified perspective, examine the technical foundations of the emerging tools, and highlight the critical challenges and open gaps that must be addressed by the research community, industry, and standardization bodies to move from promising demonstrations to operational, trustworthy AI-driven networks.
Biography
Behnam Shariati (B.Sc., M.Sc., Ph.D.) is Deputy Head of the Data Analytics and Signal Processing group within the department of Photonic Networks and Systems at Fraunhofer HHI. He leads the topical area AI for Photonics, pioneering state-of-the-art AI-assisted automation solutions for packet-optical networks.
He has been a strong advocate of public dataset sharing in the optical networking community, and his team has developed multiple solutions to lower the barrier to cross-stakeholder data sharing in the telecommunications industry, while addressing confidentiality, regulatory, and business-critical requirements of the problem space.
Dr. Shariati has (co-)authored over 100 peer-reviewed publications in leading journals and conferences and has served on the Technical Program Committees of major international venues, including Optical Fiber Communication Conference, European Conference on Optical Communication, APC, and Optical Network Design and Modeling. He represents Fraunhofer HHI in ETSI standardization activities and serves as a key delegate to the Telecom Infra Project (TIP). He is a Senior Member of Optica and a Member of IEEE.
Speaker: Prof. Alfredo Nascita
Abstract
As modern Internet traffic becomes increasingly dynamic, encrypted, and heterogeneous, traditional approaches to network monitoring and cybersecurity are becoming progressively less effective. Detecting cyber threats, anomalies, and malicious activities in real-world environments requires more adaptive and intelligent solutions. Deep Learning (DL) has emerged as one of the most powerful strategies for traffic classification, intrusion detection, and cyber-attack identification, thanks to its ability to capture complex patterns in large-scale network data.
However, despite its strong performance, the adoption of DL in operational cybersecurity systems remains limited by a critical issue: trustworthiness. Security analysts and network operators often struggle to rely on models whose decisions cannot be clearly understood. The black-box nature of DL models reduces confidence in automated detection systems and creates significant barriers to deployment in critical environments. In this context, Explainable Artificial Intelligence (XAI) becomes not simply an additional capability but a fundamental requirement for trustworthy AI-driven cybersecurity solutions. Indeed, building such solutions requires moving beyond accuracy alone—toward models that security professionals can interpret, validate, and trust in real operational environments.
This keynote explores how explainability can move from post-hoc interpretation to a core design principle for next-generation network security systems. Through applications in traffic classification, intrusion detection, and cyber-attack identification, the talk discusses how XAI improves model transparency, reliability, and robustness to ultimately design trustworthy AI models for network security. Particular attention is given to feature relevance analysis, confidence estimation, and concept localization inside deep models, enabling a deeper understanding of model behavior and supporting targeted model refinement without costly retraining.
Biography
Alfredo Nascita is an Assistant Professor at the Department of Electrical Engineering and Information Technologies (DIETI) of the University of Naples Federico II, where he is a member of the Traffic research group led by Antonio Pescapè. He received his Ph.D. in Information Technology and Electrical Engineering in February 2025 and his M.S. degree in Computer Engineering in March 2021, both from the University of Naples Federico II.
His research focuses on Internet traffic analysis, cybersecurity, machine learning, deep learning, and Explainable Artificial Intelligence (XAI), with particular attention to traffic classification, anomaly detection, cyber-attack identification, and trustworthy AI for network security. His work investigates how explainability can improve the transparency, reliability, and operational adoption of AI-driven network monitoring systems.
He has co-authored more than 25 publications in international journals and conference proceedings in the fields of network intelligence, cybersecurity, and trustworthy AI.