Doctorate thesis defense of Slim Abbes

Doctorate thesis defense on January 23^th 2025 at 10H00 AM ,in Amphitheater Ibn Kholdoun, SUP'COM 2.

Entitled :Security and Virtualization in the Internet of Vehicles: Models and Applications

Presented by :Slim Abbes

Committee

Abstract

The Internet of Vehicles (IoV) integrates the Internet of Things (IoT) with Intelligent Transportation Systems to enhance vehicular connectivity and intelligence. However, IoV encounters significant security and scalability challenges due to the high mobility of nodes, dynamic network topology, and distributed nature of nodes and services. This thesis aims to advance security and virtualization in the Internet of Vehicles by proposing innovative architectures, models, and analytics to mitigate security threats and attacks on vehicle networks. Additionally, it seeks to optimize the dynamic allocation and utilization of mobile vehicle sensors for cloud-based applications in Intelligent Transportation Systems.

Our contribution is threefold: Firstly, we propose a blockchain-based solution to reinforce security in cloud-based IoV systems by managing vehicle reputations. Our method employs watchdog nodes to supervise vehicles and calculate routing-based metrics, which are continually refreshed to uphold trust within the network. These reputations play a crucial role in issuing security certificates. We utilize distributed and automated processes via smart contracts and integrate homomorphic encryption to guarantee data confidentiality. Secondly, we design a virtual sensor provisioning architecture for IoV aimed at enhancing data collection accuracy while enabling dynamic device reconfiguration, thus allowing a single sensor to serve multiple requests. We leverage a Markov chain to estimate sensor availability based on vehicular movement patterns and develop an utility function for the Sensor Cloud Service Provider (SCSP) to optimize sensor allocation. Additionally, we introduce a sensor selection algorithm to improve allocation efficiency and we simulate its impact on reducing blockage rates. This approach seeks to efficiently manage on-demand, scalable services, maximizing resource utilization, and addressing mobility challenges in IoV. Thirdly, we enhance the intelligent provisioning of virtual vehicle sensors through the development of a reinforcement learning model. This model dynamically selects the most suitable physical sensor for each service period, considering factors such as mobility, availability and allocation cost. By leveraging Q-learning and SARSA algorithms, we efficiently manage sensor allocation, leading to enhanced operational costs and improved stability in sensor service delivery.

Keywords

IoV, Virtualization, Virtual sensor, Cloud/Fog Computing, Blockchain, Security attacks, Reinforcement learning