http://hdl.handle.net/10453/19978 2026-06-06T02:09:45Z http://hdl.handle.net/10453/195186 Title: Enabling Secure and Reliable Wireless Services With Intelligent Reflecting Surfaces Authors: Abughalwa, Monir Abstract: The forthcoming sixth-generation (6G) mobile networks are expected to deliver ultra-fast, low-latency, and highly reliable connectivity, supporting applications in smart cities, healthcare, transportation, and industry. However, 6G systems face major challenges, including extremely low latency, high reliability, and security requirements. Intelligent reflecting surfaces (IRS) have emerged as a promising technology to improve user data rates and secrecy by enhancing the received signal of legitimate users while limiting eavesdroppers’ interception. This thesis investigates IRS-assisted secure communication through joint optimization of the transmitter beamforming vector and the IRS programmable reflecting elements (PREs), considering practical constraints such as low-resolution IRS, imperfect channel state information (CSI), and different eavesdropper CSI assumptions (perfect, imperfect, or unknown). First, we study data rate maximization in a downlink IRS-aided system under the finite blocklength regime (FBR), where a base station serves multiple single-antenna users. Since achievable rates in the FBR are intricate functions of beamforming and IRS phase shifts, we propose a joint optimization framework that maximizes the geometric mean of user rates. A computationally efficient algorithm based on closed-form approximations is developed, and simulations confirm its effectiveness. Second, we address users’ secrecy in long blocklength (LBR) IRS-aided systems with low-resolution IRS. The objective is to maximize the minimum secrecy rate among all users under different CSI conditions. The resulting nonconvex problem is tackled by linearizing its objective function and then decomposing it into a series of tractable subproblems. For imperfect CSI, we use the successive convex approximation (SCA) method, and S-procedure to tackle the problem. Extensive simulations under practical settings validate the efficacy of the proposed framework. Finally, we extend the study to users’ secrecy in FBR-IRS-aided systems, where enhancing users’ secrecy is more challenging due to latency and reliability FBR constraints. We formulate problems to maximize both the minimum and sum secrecy rates while satisfying FBR constraints, by jointly optimizing the beamform and the IRS PREs. We address the nonconvex problems using linearization and the SCA technique. Extensive simulations under practical conditions demonstrate that, when it is feasible, the proposed framework can reliably ensure secure communications for all users under FBR constraints. In conclusion, this thesis develops robust and scalable optimization frameworks for data rate maximization and secrecy provisioning in IRS-aided 6G networks under both LBR and FBR regimes. The proposed methods provide scalable solutions to practical settings with large IRSs, enabling secure, reliable, and high-rate communication in future 6G systems. Description: University of Technology Sydney. Faculty of Engineering and Information Technology. 2025-01-01T00:00:00Z http://hdl.handle.net/10453/195172 Title: Towards Robust and Privacy-Preserving Federated Learning: Reinforcement-Guided Unlearning and Multi-Attack Defense Mechanisms Authors: Gao, Kun Abstract: Federated Learning (FL) has become an attractive new trend for distributed model training where the data is decentralized and is not required to be uploaded to a central server itself, alleviating the direct privacy issues. However, this decentralized paradigm also gives rise to new security risks such as gradient inversion, poisoning, and inference threats. Meanwhile, with rigorous regulatory requirements and the changing nature of privacy, federated unlearning is a crucial building block, bringing about new issues related to security and efficiency. This thesis explores secure and private defenses of FL against various threat models by combining reinforcement learning-informed defense strategies with statistical machine unlearning. The contributions of this paper are as follows: It proposes a statistical-unlearning-based defense through gradient inversion attacks while also achieving the balance between high model utility and communication efficiency. It reveals potential vulnerabilities in federated unlearning by revealing invisible attacks (i.e., camouflaged poisoning attacks that can still be effective after unlearning operations) and theoretically analyzes the effects of the attacks for the long term. It introduces the first reinforcement learning-based federated unlearning mechanism that can dynamically balance the client contribution, the privacy cost, and the computational efficiency, leading to enhanced robustness to both inference and poisoning. It also introduces a data importance-aware reinforcement learning defense that adaptively victims protection strategies at a sample level and achieves multi-attack robustness against backdoor, model stealing, and membership inference attacks. Both theoretically and empirically, we verify that the proposed approaches achieve better trade-offs between robustness, accuracy, and efficiency over different datasets and adversarial settings than state-of-the-art counterparts. Taken together, this thesis contributes to the understanding of attack defense dynamics on FL and proposes reinforcement-guided unlearning as a principled basis for adaptive, secure, and privacy-compliant decentralized learning. Description: University of Technology Sydney. Faculty of Engineering and Information Technology. 2025-01-01T00:00:00Z http://hdl.handle.net/10453/195165 Title: Designing up-converting nanomaterials for intracellular cargo dynamics Authors: Tang, Wilson Z. Abstract: The advancement of nanotechnology has enabled the development of upconversion nanoparticles (UCNPs), a class of luminescent materials capable of converting low-energy near-infrared light into higher-energy visible or ultraviolet emissions. This unique optical property makes UCNPs highly promising for applications in biomedical imaging, diagnostics, and sensing. However, a deeper understanding of their optical behavior at the single-particle level and their interactions within biological environments remains limited. This thesis investigates both the fundamental and applied aspects of UCNPs through a combination of synthesis, optical characterization, and computational modeling. It begins by exploring the synthesis of UCNPs with varied dopants, crystal structures, and surface modifications to tune their luminescent and chemical properties. A key part of the project involved building a custom, multi-purpose optical system for high-resolution imaging and single-particle characterization, allowing for precise measurements of luminescence, lifetime, and energy transfer efficiency. The optical properties of UCNPs were evaluated using techniques such as scanning confocal microscopy, widefield imaging, and photon-counting spectroscopy. These results attempt to predict emission behavior based on dopant combinations and particle environments. In biological contexts, the uptake and movement of UCNPs within living cells were tracked over time. Using statistical tools such as mean square displacement and Hidden Markov Models, the study classified intracellular transport behaviors and explored the mechanisms of vesicle trafficking. The findings contribute to a better understanding of how UCNPs behave optically and biologically, enabling their optimized use in single-particle imaging, photodynamic therapy, and diagnostic assays. This research lays a strong foundation for the future design of intelligent, responsive nanomaterials for biomedical applications, combining theoretical insight with practical instrumentation and cellular investigation. Description: University of Technology Sydney. Faculty of Science. 2025-01-01T00:00:00Z http://hdl.handle.net/10453/195160 Title: 3D Scene Reconstruction with Explicit Geometric Modeling Authors: Lin, Ancheng Abstract: 3D scene reconstruction aims to recover the geometric and visual structure of real-world environments from sensor data. Existing methods generally fall into two categories: explicit approaches, which represent geometry using discrete primitives like meshes, and implicit approaches, which encode scenes as continuous functions. While implicit methods have achieved impressive visual fidelity, they often lack the direct geometric access required for physically grounded applications. This thesis addresses this limitation by improving explicit geometric modeling through a systematic investigation targeting three complementary levels of reconstruction complexity. First, at the data level, we propose the Hybrid Geometric Transformer (HGT). This network fuses visual semantics with sparse LiDAR points to estimate accurate surface normals, providing robust geometric priors. Second, for static scene representation, we introduce a hybrid model that integrates 3D Gaussian Splatting with a triangle mesh. By explicitly binding Gaussians to mesh faces, we enforce structural constraints while enabling high-fidelity rendering. Third, extending to dynamic environments, we propose the Dynamic Appearance Particle Neural Radiance Field (DAP-NeRF), which utilizes Lagrangian particles to explicitly model motion and appearance evolution. Experimental results demonstrate that these contributions significantly enhance both reconstruction accuracy and physical interpretability. Description: University of Technology Sydney. Faculty of Engineering and Information Technology. 2025-01-01T00:00:00Z