Supervision

PhD Thesis

Student: Yibeltal MOLLA

• Title: XAI-based Long-Term Evolution Design of the V2X Physical Layer.
• Abstract: 6G is envisioned to transcend simple connectivity, evolving toward 'connected intelligence' where native AI serves as a fundamental pillar for end-to-end services. Although significant investments are driving AI adoption, the 'black box' nature of these models raises critical challenges regarding transparency and security, necessitating the development of Explainable AI (XAI) frameworks. However, current literature focuses primarily on theoretical studies rather than practical implementations. Specifically, at the physical layer (PHY), conventional XAI methods struggle to adapt to the high dimensionality and the lack of discriminating features in deep learning models, creating a significant gap in the development of truly deployable solutions. This thesis aims to explore XAI methods tailored to the constraints of physical layer applications, adapt these mechanisms to various deep learning approaches (single, joint, and E2E tasks), and design XAI-optimized Federated Edge Learning (FEL) models to increase the communication efficiency of decentralized V2X communications.
• Keywords: 6G, PHY Layer, Explainable Artificial Intelligence, V2X.
• Establishment: IMT Nord Europe and Paris-East Créteil University (UPEC).
• Laboratory: Center for Digital Systems (CERI-SN).
• Doctoral school: Mathématiques, Sciences du Numérique et de leurs interactions (MADIS).
• Duration: September 2025 - Present.
  

Student: Abdelhak HEROUCHA

• Title: QoE, Explicability and End-to-End trustworthiness among very large heterogeneous IoT Systems: use case of Healthcare applications.
• Abstract: The main objective of the thesis is to see how to install original QoE-based functionalities to guarantee trustiness among all domains followed by an IP traffic among several and heterogeneous IoT Healthcare Systems/Domains under 5G/5G Beyond/6G slicing technologies in the context of New IP architectures and to produce prototypes to evaluate our proposals in the case of Very Large Scale Smart Cities and Health situations. These fields are linked with our past studies on Inter-Autonomous Systems Quality of Service (QoS)/Quality of Experience (QoE)-based Routing, Knowledge/Data plan and Control Plan development. In fact, the flexibility and dynamics of the new generation network infrastructure often based on SDN raise important security issues: sensitive data can easily be transferred between different sites, and false knowledge injected at the right time in the network can lead to violation of some constraints. Furthermore, there is a massive and exponential increase in both the number of nodes and in usage. The question that arises is, therefore, how to design a network that can guarantee the satisfaction of all users even in the case of contradictory objectives. How to ensure traceability at all times, and finally, how to guarantee that each component of the system, software or hardware, is a trusted element.
• Keywords: Networks, Explainable Artificial Intelligence, Routing/Meta Routing, SDN, Big Data, Scalability.
• Establishment: Paris-East Créteil University (UPEC).
• Laboratory: Laboratoire Images, Signaux et Systèmes Intelligents (LISSI).
• Doctoral school: École doctorale Mathématiques, Sciences et Technologies de l’Information et de la Communication.
• Duration: September 2025 - Present.
  

Student: Rafik DERRADJI

• Title: Implementation of an on-board Zero-Touch adaptive and explainable Trustworthiness QoE-centered decision support system for mission-critical operations.
• Abstract: The evolution of technologies driven by the significant increase in data exchange capacities with the upcoming arrival of 6G, the numerous applications stemming from IoT, the multiplication of information sources and flows, and the availability of increasingly advanced and diverse devices and accessories necessitate rethinking decision-support tools. In the field targeted by this thesis, namely video surveillance, administrations, organizations, and security agents in the field and command centers are, and will increasingly be, faced with a considerable amount of information. The ''intelligent'' exploitation of this information could enable disruptive advancements in the effectiveness and efficiency of their missions. This thesis builds on the work carried out by Airbus and the LISSI laboratory, which led to the creation of an initial decision-support platform based on the user experience paradigm (Quality of Experience, QoE). It aims to address new challenges by integrating explainable, adaptive, and frugal artificial intelligence technologies to enhance the effectiveness of video surveillance systems and support field agents. The main objectives of this research include the integration of explainability mechanisms to enable end-users to understand system decisions, thereby strengthening their trust and operational efficiency. The project also seeks to develop a scalable system based on the zero-touch paradigm, capable of learning in real time from new data and responding to unknown situations. By leveraging reinforcement learning, the aim is to optimize an agent's actions in dynamic and uncertain environments, with a deep understanding of the surroundings to maximize long-term performance. Human-machine interaction is studied to design cooperative systems that integrate up-to-date, multi-scale data to optimize decision-making. Additionally, frugal solutions are explored to ensure reliable operation even in the event of disruptions, such as communication losses with the control center. Finally, the integration of Trust QoE will enable the design of a system ensuring a trustworthy user experience by guaranteeing security, reliability, and rapid decision-making, particularly in high-risk situations. This future system will stand out through its ability to support first responders during their missions with accessible visual explanations, continuously adapt to field conditions using data generated by generative AI techniques, and optimize resource management to reduce energy consumption while maintaining high performance levels.
• Keywords: Trustworthy QoE, Trustworthy AI, Zero-touch, Explainability, Generative AI, Real-time.
• Establishment: Paris-East Créteil University (UPEC).
• Laboratory: Laboratoire Images, Signaux et Systèmes Intelligents (LISSI).
• Doctoral school: École doctorale Mathématiques, Sciences et Technologies de l’Information et de la Communication.

• Duration: September 2025 - Present.

  
  

Internships - M1 - M2

Student: Muhammad Hashir (M2)

• Title: Explainable AI for Real-Time QoE Prediction for Modern Communication Systems.
• Establishment: UPEC - France.
• Duration: May 2026 - October 2026.

Student: Ali Saleh (M2)

• Title: Explainable AI for Optimized Edge Federated Learning.
• Establishment: EFREI & UPEC - France.
• Duration: April 2026 - October 2026.

Student: Stéphane KOUAME (M2)

• Title: XAI-assisted Deep Receiver Design in Wireless Communications.
• Establishment: SogetiLabs & IMT Nord Europe - France
• Duration: April 2025 - August 2025.

Student: Ali Saleh (M1)

• Title: Explainable AI for Remote Sensing and Image Segmentation.
• Establishment: SogetiLabs & GEOAI Group - France/Lebanon.
• Duration: March 2025 - August 2025.

Student: Romain GRANGE (M2)

• Title: SDN Network Automation Using Graph Neural Networks (GNNs).
• Establishment: SogetiLabs & Université Claude Bernard Lyon 1- France
• Duration: April 2024 - August 2024.

Student: Ilyes GHOMARI (M1)

• Title: SDN Traffic Classification and Rerouting Using Reinforcement Learning (RL).
• Establishment: SogetiLabs & Université UVSQ - France
• Duration: April 2024 - August 2024.

Student: Mouhamed SY (M1)

• Title: Optimizing Semantic Segmentation Models with Explainable AI (XAI).
• Establishment: SogetiLabs & Université PSL- France
• Duration: Jan 2024 - August 2025.

Student: Reem Hammoud (Internship)

• Title: XAI Evaluation Strategies and Metrics for Semantic Segmentation.
• Establishment: SogetiLabs & GEOAI Group - France/Lebanon.
• Duration: Jan 2024 - August 2024.

Student: Mohamed QASSIR (M2)

• Title: SDN Network end-to-end Implementation Using MiniNet Emulator.
• Establishment: SogetiLabs & ESIEE Paris - France
• Duration: December 2023 - August 2024.

Student: Hossein Shreim (M1)

• Title: Perturbation-based XAI Schemes for Remote Sensing.
• Establishment: IMT Nord Europe & GEOAI Group - France/Lebanon.
• Duration: March 2023 - August 2023.

Student: Nada RAKRAKY (M2)

• Title: Infrastructure design, security, and virtualization with VMware GNS3 simulator.
• Establishment: ENSEA Cergy - France
• Duration: June 2021 - September 2021.