Duc-Son Pham
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Duc-Son Pham has a total of 72 co-authors. Here are the top co-authors by number of shared publications:
- Svetha Venkatesh
Applied Artificial Intelligence Institute, Deakin University, Waurn Ponds, VIC, Australia
8 shared publications - Mihai Lazarescu
Curtin University
6 shared publications - Sieteng Soh
School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Perth, Australia
6 shared publications - Kit Yan Chan
School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Perth, WA, Australia
4 shared publications - Kwan-Wu Chin
University of Wollongong
4 shared publications - Tanmay Singha
The University of Notre Dame, Fremantle, WA 6160, Australia
4 shared publications
Latest Publications
- Open Access
- 2 July 2026
- Published by Elsevier in SSRN Electronic Journal
Precision-Aware Multi-Stage Edge Server Upgrade with Service Caching
The rapid growth of AI-enabled applications is making MEC infrastructure planning more challenging. These workloads often require configurable output precision, where higher-precision output requires greater computational resources, increases processing delay, and ultimately affects long-term edge resource planning, provisioning, and allocation. This paper presents a novel network planning problem for upgrading a multi-access edge computing (MEC) network across multiple stages, e.g., over several years. The problem, called Precision-Aware Multi-Stage Edge Server Upgrade with Service Caching (M-ESCA), decides in each stage (i) whether to deploy new edge servers and/or increase the resource capacity of existing servers, and (ii) which service components should be cached and which subtasks should be offloaded to edge servers or the cloud. The goal is to maximize the average number of tasks that meet their delay requirements. We present a Mixed-Integer Linear Programming (MILP) model and an efficient heuristic algorithm, called M-ESCA/H, to solve M-ESCA. M-ESCA considers budget constraints, deployment and upgrade costs with depreciation, server computing and storage capacities, growing task demand, increasingly strict delay requirements across stages, and precision-dependent subtask processing times. The MILP model yields optimal solutions for small networks, whereas M-ESCA/H is designed for large-scale networks. Simulation results show that M-ESCA/H achieves near-optimal performance on small networks, with an optimality gap generally below five percentage points and a runtime of less than one second. For large networks, M-ESCA/H consistently outperforms M-ESCA/DF and M-ESCA/UF, which prioritize server deployment and server upgrade, respectively, under varying budgets, task demands, planning stages, precision settings, and service-size distributions.
- 2 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPrecision-Aware Multi-Stage Edge Server Upgrade with Service Caching
The rapid growth of AI-enabled applications is making MEC infrastructure planning more challenging. These workloads often require configurable output precision, where higher-precision output requires greater computational resources, increases processing delay, and ultimately affects long-term edge resource planning, provisioning, and allocation. This paper presents a novel network planning problem for upgrading a multi-access edge computing (MEC) network across multiple stages, e.g., over several years. The problem, called Precision-Aware Multi-Stage Edge Server Upgrade with Service Caching (M-ESCA), decides in each stage (i) whether to deploy new edge servers and/or increase the resource capacity of existing servers, and (ii) which service components should be cached and which subtasks should be offloaded to edge servers or the cloud. The goal is to maximize the average number of tasks that meet their delay requirements. We present a Mixed-Integer Linear Programming (MILP) model and an efficient heuristic algorithm, called M-ESCA/H, to solve M-ESCA. M-ESCA considers budget constraints, deployment and upgrade costs with depreciation, server computing and storage capacities, growing task demand, increasingly strict delay requirements across stages, and precision-dependent subtask processing times. The MILP model yields optimal solutions for small networks, whereas M-ESCA/H is designed for large-scale networks. Simulation results show that M-ESCA/H achieves near-optimal performance on small networks, with an optimality gap generally below five percentage points and a runtime of less than one second. For large networks, M-ESCA/H consistently outperforms M-ESCA/DF and M-ESCA/UF, which prioritize server deployment and server upgrade, respectively, under varying budgets, task demands, planning stages, precision settings, and service-size distributions.
- Open Access
- 31 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 3168-3188
Delay-Aware Multi-Stage Edge Server Upgrade With Budget Constraint
In this paper, the Multi-stage Edge Server Upgrade (M-ESU) is proposed as a new network planning problem, involving the upgrading of an existing multi-access edge computing (MEC) system through multiple stages (e.g., over several years). More precisely, the problem considers two key decisions: (i) whether to deploy additional edge servers or upgrade those already installed, and (ii) how tasks should be offloaded so that the average number of tasks that meet their delay requirement is maximized. The framework specifically involves: (i) deployment of new servers combined with capacity upgrades for existing servers, and (ii) the optimal task offloading to maximize the average number of tasks with a delay requirement. It also considers the following constraints: (i) budget per stage, (ii) server deployment and upgrade cost (in $) and cost depreciation rate, (iii) computation resource of servers, (iv) number of tasks and their growth rate (in %), and (v) increase in task sizes and stricter delay requirements over time. We present two solutions: a Mixed Integer Linear Programming (MILP) model and an efficient heuristic algorithm (M-ESU/H). MILP yields optimal solutions for small networks, whereas M-ESU/H is used for large-scale networks. For small networks, the simulation results show that the solution computed by M-ESU/H is within 1.25% of the optimal solution while running several orders of magnitude faster. For large networks, M-ESU/H is compared against three alternative heuristic solutions that consider only server deployment, or give priority to server deployment or upgrade. Our experiments show that M-ESU/H yields up to 21.57% improvement in task satisfaction under identical budget and demand growth conditions, confirming its scalability and practical value for long-term MEC systems.
- 31 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 3168-3188
Open AccessDelay-Aware Multi-Stage Edge Server Upgrade With Budget Constraint
In this paper, the Multi-stage Edge Server Upgrade (M-ESU) is proposed as a new network planning problem, involving the upgrading of an existing multi-access edge computing (MEC) system through multiple stages (e.g., over several years). More precisely, the problem considers two key decisions: (i) whether to deploy additional edge servers or upgrade those already installed, and (ii) how tasks should be offloaded so that the average number of tasks that meet their delay requirement is maximized. The framework specifically involves: (i) deployment of new servers combined with capacity upgrades for existing servers, and (ii) the optimal task offloading to maximize the average number of tasks with a delay requirement. It also considers the following constraints: (i) budget per stage, (ii) server deployment and upgrade cost (in $) and cost depreciation rate, (iii) computation resource of servers, (iv) number of tasks and their growth rate (in %), and (v) increase in task sizes and stricter delay requirements over time. We present two solutions: a Mixed Integer Linear Programming (MILP) model and an efficient heuristic algorithm (M-ESU/H). MILP yields optimal solutions for small networks, whereas M-ESU/H is used for large-scale networks. For small networks, the simulation results show that the solution computed by M-ESU/H is within 1.25% of the optimal solution while running several orders of magnitude faster. For large networks, M-ESU/H is compared against three alternative heuristic solutions that consider only server deployment, or give priority to server deployment or upgrade. Our experiments show that M-ESU/H yields up to 21.57% improvement in task satisfaction under identical budget and demand growth conditions, confirming its scalability and practical value for long-term MEC systems.
- Open Access
- 1 October 2025
- Published by MDPI AG in Microplastics
- Vol. 4  (4) , 69
A Domain-Adaptive Deep Learning Approach for Microplastic Classification
Microplastics pose a growing environmental concern, necessitating accurate and scalable methods for their detection and classification. This study presents a novel deep learning framework that integrates a transformer-based architecture with domain adaptation techniques to classify microplastics using reflectance micro-FTIR spectroscopy. A key challenge addressed in this work is the domain shift between laboratory-prepared reference spectra and environmentally sourced spectra, which can significantly degrade model performance. To overcome this, three domain-adaptation strategies—Domain Adversarial Neural Networks (DANN), Deep Subdomain-Adaptation Networks (DSAN), and Deep CORAL—were evaluated for their ability to enhance cross-domain generalization. Experimental results show that while DANN was unstable, DSAN and Deep CORAL improved target domain accuracy. Deep CORAL achieved 99% accuracy on the source and 94% on the target, offering balanced performance. DSAN reached 95% on the target but reduced source accuracy. Overall, statistical alignment methods outperformed adversarial approaches in transformer-based spectral adaptation. The proposed model was integrated into a reflectance micro-FTIR workflow, accurately identifying PE and PP microplastics from unlabelled spectra. Predictions closely matched expert-validated results, demonstrating practical applicability. This first use of a domain-adaptive transformer in microplastics spectroscopy sets a benchmark for high-throughput, cross-domain analysis. Future work will extend to more polymers and enhance model efficiency for field use.
- 1 October 2025
- Published by MDPI AG in Microplastics
- Vol. 4  (4) , 69
Open AccessA Domain-Adaptive Deep Learning Approach for Microplastic Classification
Microplastics pose a growing environmental concern, necessitating accurate and scalable methods for their detection and classification. This study presents a novel deep learning framework that integrates a transformer-based architecture with domain adaptation techniques to classify microplastics using reflectance micro-FTIR spectroscopy. A key challenge addressed in this work is the domain shift between laboratory-prepared reference spectra and environmentally sourced spectra, which can significantly degrade model performance. To overcome this, three domain-adaptation strategies—Domain Adversarial Neural Networks (DANN), Deep Subdomain-Adaptation Networks (DSAN), and Deep CORAL—were evaluated for their ability to enhance cross-domain generalization. Experimental results show that while DANN was unstable, DSAN and Deep CORAL improved target domain accuracy. Deep CORAL achieved 99% accuracy on the source and 94% on the target, offering balanced performance. DSAN reached 95% on the target but reduced source accuracy. Overall, statistical alignment methods outperformed adversarial approaches in transformer-based spectral adaptation. The proposed model was integrated into a reflectance micro-FTIR workflow, accurately identifying PE and PP microplastics from unlabelled spectra. Predictions closely matched expert-validated results, demonstrating practical applicability. This first use of a domain-adaptive transformer in microplastics spectroscopy sets a benchmark for high-throughput, cross-domain analysis. Future work will extend to more polymers and enhance model efficiency for field use.
- Open Access
ICS-SimLab: A Containerized Approach for Simulating Industrial Control Systems for Cyber Security Research
Industrial Control Systems (ICSs) are complex interconnected systems used to manage process control within industrial environments, such as chemical processing plants and water treatment facilities. As the modern industrial environment moves towards Internet-facing services, ICSs face an increased risk of attacks that necessitates ICS-specific Intrusion Detection Systems (IDS). The development of such IDS relies significantly on a simulated testbed as it is unrealistic and sometimes hazardous to utilize an operational control system. Whilst some testbeds have been proposed, they often use a limited selection of virtual ICS simulations to test and verify cyber security solutions. There is a lack of investigation done on developing systems that can efficiently simulate multiple ICS architectures. Currently, the trend within research involves developing security solutions on just one ICS simulation, which can result in bias to its specific architecture. We present ICS-SimLab, an end-to-end software suite that utilizes Docker containerization technology to create a highly configurable ICS simulation environment. This software framework enables researchers to rapidly build and customize different ICS environments, facilitating the development of security solutions across different systems that adhere to the Purdue Enterprise Reference Architecture. To demonstrate its capability, we present three virtual ICS simulations: a solar panel smart grid, a water bottle filling facility, and a system of intelligent electronic devices. Furthermore, we run cyber-attacks on these simulations and construct a dataset of recorded malicious and benign network traffic to be used for IDS development.
Open AccessICS-SimLab: A Containerized Approach for Simulating Industrial Control Systems for Cyber Security Research
Industrial Control Systems (ICSs) are complex interconnected systems used to manage process control within industrial environments, such as chemical processing plants and water treatment facilities. As the modern industrial environment moves towards Internet-facing services, ICSs face an increased risk of attacks that necessitates ICS-specific Intrusion Detection Systems (IDS). The development of such IDS relies significantly on a simulated testbed as it is unrealistic and sometimes hazardous to utilize an operational control system. Whilst some testbeds have been proposed, they often use a limited selection of virtual ICS simulations to test and verify cyber security solutions. There is a lack of investigation done on developing systems that can efficiently simulate multiple ICS architectures. Currently, the trend within research involves developing security solutions on just one ICS simulation, which can result in bias to its specific architecture. We present ICS-SimLab, an end-to-end software suite that utilizes Docker containerization technology to create a highly configurable ICS simulation environment. This software framework enables researchers to rapidly build and customize different ICS environments, facilitating the development of security solutions across different systems that adhere to the Purdue Enterprise Reference Architecture. To demonstrate its capability, we present three virtual ICS simulations: a solar panel smart grid, a water bottle filling facility, and a system of intelligent electronic devices. Furthermore, we run cyber-attacks on these simulations and construct a dataset of recorded malicious and benign network traffic to be used for IDS development.
- Open Access
Towards AI-Based Strep Throat Detection and Interpretation for Remote Australian Indigenous Communities
Streptococcus pharyngitis (strep throat) poses a significant health challenge in rural and remote Indigenous communities in Australia, where access to medical resources is limited. Delays in diagnosis and treatment increase the risk of serious complications, including acute rheumatic fever and rheumatic heart disease. This paper presents a proof-of-concept AI-based diagnostic model designed to support clinicians in underserved communities. The model combines a lightweight Swin Transformer–based image classifier with a BLIP-2-based explainable image annotation system. The classifier predicts strep throat from throat images, while the explainable model enhances transparency by identifying key clinical features such as tonsillar swelling, erythema, and exudate, with synthetic labels generated using GPT-4o-mini. The classifier achieves 97.1% accuracy and an ROC-AUC of 0.993, with an inference time of 13.8 ms and a model size of 28 million parameters; these results demonstrate suitability for deployment in resource-constrained settings. As a proof-of-concept, this work illustrates the potential of AI-assisted diagnostics to improve healthcare access and could benefit similar research efforts that support clinical decision-making in remote and underserved regions.
Open AccessTowards AI-Based Strep Throat Detection and Interpretation for Remote Australian Indigenous Communities
Streptococcus pharyngitis (strep throat) poses a significant health challenge in rural and remote Indigenous communities in Australia, where access to medical resources is limited. Delays in diagnosis and treatment increase the risk of serious complications, including acute rheumatic fever and rheumatic heart disease. This paper presents a proof-of-concept AI-based diagnostic model designed to support clinicians in underserved communities. The model combines a lightweight Swin Transformer–based image classifier with a BLIP-2-based explainable image annotation system. The classifier predicts strep throat from throat images, while the explainable model enhances transparency by identifying key clinical features such as tonsillar swelling, erythema, and exudate, with synthetic labels generated using GPT-4o-mini. The classifier achieves 97.1% accuracy and an ROC-AUC of 0.993, with an inference time of 13.8 ms and a model size of 28 million parameters; these results demonstrate suitability for deployment in resource-constrained settings. As a proof-of-concept, this work illustrates the potential of AI-assisted diagnostics to improve healthcare access and could benefit similar research efforts that support clinical decision-making in remote and underserved regions.
- 31 August 2025
- Published by Springer Nature
- p. 323-334
Lightweight Scene Parsing for Real-Time Structural Crack Detection
- 31 August 2025
- Published by Springer Nature
- p. 323-334
Lightweight Scene Parsing for Real-Time Structural Crack Detection
- Open Access
- 11 June 2025
- Published by Elsevier in Computer Networks
- Vol. 269
Delay-aware multi-stage edge server placement and task offloading with budget constraint
- 11 June 2025
- Published by Elsevier in Computer Networks
- Vol. 269
Open AccessDelay-aware multi-stage edge server placement and task offloading with budget constraint
- Open Access
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1616-1625
Leveraging Prompt Engineering with Lightweight Large Language Models to Label and Extract Clinical Information from Radiology Reports
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1616-1625
Open AccessLeveraging Prompt Engineering with Lightweight Large Language Models to Label and Extract Clinical Information from Radiology Reports
- 18 November 2024
- Published by Springer Nature
- p. 124-135
Some Considerations for the Preservation of Endangered Languages Using Low-Resource Machine Translation
- 18 November 2024
- Published by Springer Nature
- p. 124-135
Some Considerations for the Preservation of Endangered Languages Using Low-Resource Machine Translation