2026-09-20T12:06:17Zhttps://digital.library.adelaide.edu.au/server/oai/request

oai:digital.library.adelaide.edu.au:2440/1396192026-06-14T23:47:55Zcom_2440_14759col_2440_14760

Guo, Mingyu Shen, Hong (Sun Yat-sen University, China) Tian, Hui (Griffith University, Australia) Goel, Diksha School of Computer and Mathematical Sciences 2023 https://hdl.handle.net/2440/139619 With the burgeoning advancements of computing and network communication technologies, network infrastructures and their application environments have become increasingly complex. Due to the increased complexity, networks are more prone to hardware faults and highly susceptible to cyber-attacks. Therefore, for rapidly growing network-centric applications, network resilience is essential to minimize the impact of attacks and to ensure that the network provides an acceptable level of services during attacks, faults or disruptions. In this regard, this thesis focuses on developing effective approaches for enhancing network resilience. Existing approaches for enhancing network resilience emphasize on determining bottleneck nodes and edges in the network and designing proactive responses to safeguard the network against attacks. However, existing solutions generally consider broader application domains and possess limited applicability when applied to specific application areas such as cyber defense and information diffusion, which are highly popular application domains among cyber attackers. These solutions often prioritize general security measures and may not be able to address the complex targeted cyberattacks [147, 149]. Cyber defense and information diffusion application domains usually consist of sensitive networks that attackers target to gain unauthorized access, potentially causing significant financial and reputational loss. This thesis aims to design effective, efficient and scalable techniques for discovering bottleneck nodes and edges in the network to enhance network resilience in cyber defense and information diffusion application domains. We first investigate a cyber defense graph optimization problem, i.e., hardening active directory systems by discovering bottleneck edges in the network. We then study the problem of identifying bottleneck structural hole spanner nodes, which are crucial for information diffusion in the network. We transform both problems into graph-combinatorial optimization problems and design machine learning based approaches for discovering bottleneck points vital for enhancing network resilience. This thesis makes the following four contributions. We first study defending active directories by discovering bottleneck edges in the network and make the following two contributions. (1) To defend active directories by discovering and blocking bottleneck edges in the graphs, we first prove that deriving an optimal defensive policy is #P-hard. We design a kernelization technique that reduces the active directory graph to a much smaller condensed graph. We propose an effective edge-blocking defensive policy by combining neural network-based dynamic program and evolutionary diversity optimization to defend active directory graphs. The key idea is to accurately train the attacking policy to obtain an effective defensive policy. The experimental evaluations on synthetic AD attack graphs demonstrate that our defensive policy generates effective defense. (2) To harden large-scale active directory graphs, we propose reinforcement learning based policy that uses evolutionary diversity optimization to generate edge-blocking defensive plans. The main idea is to train the attacker’s policy on multiple independent defensive plan environments simultaneously so as to obtain effective defensive policy. The experimental results on synthetic AD graphs show that the proposed defensive policy is highly effective, scales better and generates better defensive plans than our previously proposed neural network-based dynamic program and evolutionary diversity optimization approach. We then investigate discovering bottleneck structural hole spanner nodes in the network and make the following two contributions. (3) To discover bottleneck structural hole spanner nodes in large-scale and diverse networks, we propose two graph neural network models, GraphSHS and Meta-GraphSHS. The main idea is to transform the SHS identification problem into a learning problem and use the graph neural network models to learn the bottleneck nodes. Besides, the Meta-GraphSHS model learns generalizable knowledge from diverse training graphs to create a customized model that can be fine-tuned to discover SHSs in new unseen diverse graphs. Our experimental results show that the proposed models are highly effective and efficient. (4) To identify bottleneck structural hole spanner nodes in dynamic networks, we propose a decremental algorithm and graph neural network model. The key idea of our proposed algorithm is to reduce the re-computations by identifying affected nodes due to updates in the network and performing re-computations for affected nodes only. Our graph neural network model considers the dynamic network as a series of snapshots and learns to discover SHS nodes in these snapshots. Our experiments demonstrate that the proposed approaches achieve significant speedup over re-computations for dynamic graphs. Thesis (Ph.D.) -- University of Adelaide, School of Computer and Mathematical Sciences, 2023 en Network Resilience Machine Learning Cyber Defense Information Diffusion Enhancing Network Resilience through Machine Learning-powered Graph Combinatorial Optimization: Applications in Cyber Defense and Information Diffusion Thesis This electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at: http://www.adelaide.edu.au/legalsopen.access