Soutenance de thèse de doctorat en Cotutelle de Abdelmonom

Membres de Jury :

Abstract :

The rapid advancement and widespread deployment of Vehicle-to-Everything (V2X) communication paradigms have given rise to Vehicular Ad-hoc Networks (VANETs), which form the foundational infrastructure of Intelligent Transportation Systems (ITS). These networks constitute a self-organizing ecosystem that interconnects vehicles, roadside units, and intelligent devices to enable the envisioned cooperative situational awareness and support autonomous driving in highly dynamic traffic environments. While such connectivity promises significant improvements in the operational efficiency of ITS applications, the decentralized and open nature of VANETs introduces substantial communication challenges and exposes the network to various malicious behaviors that can severely compromise the road safety and overall transportation reliability. 

This PhD thesis addresses the dual challenge of efficiency and security in data dissemination within VANETs, with a particular emphasis on achieving rapid emergency message propagation and resilience against False Data Injection Attacks (FDIAs). Initially, we address the issue of dissemination efficiency by proposing a Robust Backbone Network based on Hybrid Selection of Relays (BACKNET-HSR). The proposed scheme combines both sender-oriented and receiver-oriented relay selection strategies into an unified hybrid mechanism that constructs an adaptive backbone aligned with the underlying road topology. After achieving satisfactory dissemination performance in terms of delay and coverage, the focus of this thesis shifts toward ensuring the trustworthiness and integrity of the exchanged emergency messages. To this end, we first propose a spatio-temporal inference framework for traffic state classification and falsification detection. The framework integrates a Graph Convolutional Network (GCN) to capture the spatial dependencies among co-located vehicles, along with a Long Short-Term Memory (LSTM) network to model their temporal dynamics. This design enables a microscopic understanding of traffic conditions across both spatial and temporal dimensions, thereby overcoming the limitations of conventional macroscopic approaches. Lastly, we introduce IDSFormer, a Hierarchical Spatio-Temporal Graph Transformer model designed to mitigate the architectural and perceptual constraints of the GCN-LSTM framework while extending its ability to detect coordinated adversarial behaviors. The effectiveness of the proposed frameworks is validated through extensive simulations, demonstrating significant performance gains over baseline methods.