Privacy-preserving resource management in LEO satellite internet-of-things

• Main Author: Shen, Bowen
• Other Authors: Lam Kwok Yan
• Format: Thesis-Master by Research
• Language: English
• Published: Nanyang Technological University 2025
• Subjects: Computer and Information Science; Artificial intelligence
• Online Access: https://hdl.handle.net/10356/182525
• Institution: Nanyang Technological University

Low Earth Orbit (LEO) satellite systems play a significant role in next generation communication networks due to their capability to provide extensive global coverage with guaranteed communications in remote areas and isolated areas where base stations cannot be cost-efficiently deployed. With the pervasive adoption of LEO satellite systems, especially in the LEO Internet-of-Things (IoT) scenarios, their spectrum resource management requirements have raised more privacy-related issues. First, the high bandwidth demand of the massive service requests and various types of application services results in concerns about data leakage during transmission. This is because, as bandwidth and type of application services increase, more user data are transmitted over the open network, hence higher risk of exposure. Second, data leakage is more difficult to control as user data are transmitted over multiple satellites, due to the rapidly changing satellite spectrum availability. Rapid changes in spectrum allocation involve the use of multiple satellites to transmit fragmented data, often requiring frequent switching between satellites operated by different operators. Such frequent changes in spectrum may result in limited or intermittent bandwidth availability, forcing the data to be split into smaller fragments to utilize available channels efficiently. Since each operator may implement distinct protocol designs, including varying encryption methods, this leads to inconsistencies in how data privacy is enforced. These differences in encryption standards across operators can result in inconsistent levels of data protection. Therefore, the privacy of the transmitted data may be compromised due to the inconsistent privacy level across the network. Third, typically satellite spectrum management systems would adopt some optimized intelligent allocation strategy which will require collection of user data for machine learning purposes, and the privacy of these data will be at risk. Safeguarding sensitive information while being able to guarantee the efficiency of the optimization process through machine learning is a big challenge. In this thesis, we mainly focus on the schemes for privacy-preserving resource management in LEO satellite IoT. First, we propose a hybrid privacy-aware spectrum pricing and power control framework for LEO IoT by combining blockchain technology and FL. We first design a local deep reinforcement learning algorithm for LEO satellite systems to learn a revenue-maximizing pricing and power control scheme. Then the agents collaborate to form an FL system. We also propose a reputation-based blockchain which is used in the global model aggregation phase of FL. Based on the reputation mechanism, a node is selected for each global training round to perform model aggregation and block generation, which can further enhance the decentralization of the network and guarantee trust. Second, we introduce a federated dynamic spectrum access scheme. Bi-level KMeans clustering is utilized to cluster the terrestrial users (TUs) to improve the training efficiency in federated learning (FL). To address the interference caused by frequent changes in the location of TUs, we then design a federated PPO-based algorithm to obtain an optimal strategy to work even under an interactive environment for efficient resource allocation. FL contributes to efficient aggregation while preserving the privacy of users' data during the training. Third, we present the design of a satellite-enabled system for efficient and certified service providing. Taking advantage of the LEO satellite communications, our system can accommodate a wide variety of sensors including wind speed sensors, temperature sensors, cameras, etc. operating simultaneously and acquiring their information in real time. Besides, to enhance the long-term maintenance of communication services and the collaboration of each device, a federated spectrum learning approach for dynamic spectrum access is proposed. In addition, blockchain technology is utilized in the global model aggregation in federated learning to strengthen the decentralization of the architecture for security purposes. The feasibility and efficiency of the system are proved with the details of the information collected by the sensors shown in the proposed functional architecture. First, we propose a hybrid privacy-aware spectrum pricing and power control framework for LEO IoT by combining blockchain technology and FL. We first design a local deep reinforcement learning algorithm for LEO satellite systems to learn a revenue-maximizing pricing and power control scheme. Then the agents collaborate to form an FL system. We also propose a reputation-based blockchain which is used in the global model aggregation phase of FL. Based on the reputation mechanism, a node is selected for each global training round to perform model aggregation and block generation, which can further enhance the decentralization of the network and guarantee trust.