Project period: 2022-2025

Presentation

Artificial intelligence is increasingly used to protect medical systems and other high-risk environments, but security solutions must remain transparent, traceable, and reviewable. XDGMed explores explainable attack detection by representing complex analyses as meaningful graphs rather than relying only on the opaque outputs of black-box models.

The project develops spectral graph algorithms for explainable security analyses. Its research focuses on dynamic graphs that model evolving environments, with Laplacian analysis as the main approaches. The work targets trustworthy cybersecurity for eHealth and other sensitive systems, while following reproducible-research practices and promoting open data and open-source software.

Partners

Results

  • Graph-based methods for explainable attack detection in dynamic networks.
  • Spectral analysis methods for identifying changes in network structure and detecting suspicious activity.
  • Inductive graph convolutional approaches for anomaly detection in medical and IoT data.

Publications

  • Majed Jaber, Nicolas Boutry, and Pierre Parrend. 2023. “Towards Attack Detection in Traffic Data Based on Spectral Graph Analysis.” Complex Computational Ecosystems (CCE). Best Student Presentation. HAL record.
  • Majed Jaber, Nicolas Boutry, and Pierre Parrend. 2023. “Structural and Spectral Analysis of Dynamic Graphs for Attack Detection.” Rencontre des Jeunes Chercheurs en Inteligence Artificielle (RJCIA). HAL record.
  • Majed Jaber, Nicolas Boutry, and Pierre Parrend. 2024. “Graph-Based Spectral Analysis for Detecting Cyber Attacks.” International Conference on Availability, Reliability and Security (ARES). Paper.
  • Majed Jaber, Pierre Parrend, and Nicolas Boutry. 2025. “Spectral Graph Analysis of Bipartite Graphs for Advanced Attack Detection.” European Interdisciplinary Cybersecurity Conference (EICC). Paper.
  • Majed Jaber, Julien Michel, Nicolas Boutry, and Pierre Parrend. 2025. “Cyberattack Detection through GPML: Graph Processing for Machine Learning.” SoftwareX. Paper.
  • Majed Jaber, Abdul Qadir Khan, Ankush Meshram, Julien Michel, Côme Frappé-Vialatoux, and Pierre Parrend. 2026. “Tool Demo: Topology Analysis with GPML for Detection of Cyberattacks in Water Distribution Networks.” IEEE/IFIP Network Operations and Management Symposium 2026, Rome, France. Paper.

Project page: XDGMed