Self-Funded PhD Proposals
Designing Adaptive Persona-Based Conversational Agents for Personalised Interaction
Supervisors: Abdel-Karim Al-Tamimi, Dean Petters, and Laurynas Rutkauskas
Recent research indicates that conversational agents designed to reflect individual personality characteristics can create more engaging, trustworthy, and effective interactions. However, most existing studies rely on static personas or focus on limited application areas, leaving unexplored whether a general-purpose, transparent, and adaptive chatbot that embodies distinct personality styles and infers users’ traits from dialogue can meaningfully enhance user experience and interaction quality. This proposed PhD project would develop an open and inspectable conversational system with varied personas, systematically evaluate the impact of personality alignment across different contexts, and contribute to advancing knowledge and practice in personalised conversational artificial intelligence.
Dynamic Authorization Based on Self-Adaptation for Dealing with Insider Threats
Supervisor: Dr Carlos Da Silva
This line of research looks at the use of self-adaptation techniques (monitor-analyse-plan-execute feedback control loop) for dynamically modifying access control rules in response to incidents caused by insider threats. It involves the use of digital identity management and access control techniques (role-based access control and attribute-based access control) in different domains. It builds on top of our experience with the definition of reconfiguration policies for distributed firewalls in response to software vulnerabilities (https://doi.org/10.1186/s13174-018-0083-6), dynamic reconfiguration or role-based access control rules in the openstack cloud platform (https://doi.org/10.1186/s13174-018-0090-7), analysis of business processes logs based on role-based access control rules for identification of insider threats (https://doi.org/10.1109/SEAMS.2017.13), the externalisation of access control decisions in Python DJango (https://github.com/welkson/PEP4Django) and the definition of an API for supporting the self-adaptation of authorization infrastructures (https://github.com/welkson/SecAuthAPI).
Smart Ecosystem Monitoring Platform for Multifunctional Landscape Climate Adaptation
Supervisor: Dr Carlos Da Silva
This project aims to research, design and develop a “smart multifunctional landscape” to support sustainable and resilient landscape management in the face of environmental change. Based on the EU Data Space initiative (https://dssc.eu) we are interested in investigating how a coordinated data infrastructure, could support the execution and posterior sharing of data produced by diverse ecological, social, cultural and economic/policy organisations working in the landscape system.
This project is set in the context of a multi-stakeholder partnership of leading environment-sector organisations (the Sheffield and Rotherham Wildlife Trust, Environment Agency, Natural England, Yorkshire Water and Sheffield City Council) with a 10-year strategic plan for landscape management in the ‘Sheffield Lakeland’ to achieve climate resilience, nature recovery and other benefits to society and the economy. The project involves the research and development of an IoT-based solution which integrates geographical information systems, existing datasets and available national data sources (via API) with an expanding network of sensors deployed across the landscape. These will be used to enhance real-time monitoring, predictive modelling and post-hoc evaluation of environmental interventions informing practical environmental management actions towards goals of climate adaptation, resilience and nature recovery in the Sheffield region.
Software Engineering for Development of Service-Oriented Systems
Supervisor: Dr Carlos Da Silva
This line of research looks at software development processes for (micro-)service-oriented systems, including aspects of DevOps, Technical Debt and the use of Business Processes for guiding the services definition. In particular we are looking to expand the work started with the SPReaD approach (https://doi.org/10.3390/software1040018), exploring our technique for the capture of business processes (https://doi.org/10.1109/RE54965.2022.00029) and the use of BPMN models to identify bounded contexts and their relationships (https://doi.org/10.3390/software1040018) as guidance for identifying (micro-)services.
Enhancing Medical Supply Chain Security through Zero Trust Architecture (ZTA): A Framework for Implementation
Supervisor: Dr Steve Mvalo
The medical supply chain is a critical lifeline, ensuring that medicines and medical products reach patients safely and effectively. However, it faces increasing threats, including cyber-attacks, data breaches, and the infiltration of counterfeit or substandard products. These challenges endanger public health, erode trust in healthcare systems, and result in significant financial losses worldwide.
The PhD proposal project focuses on addressing these challenges by designing and implementing a Zero Trust Architecture (ZTA) tailored to the medical supply chain. Unlike traditional security models, ZTA assumes no entity can be inherently trusted, ensuring rigorous verification and continuous monitoring throughout the supply chain. By integrating blockchain technology, Internet of Things (IoT) sensors, and Artificial Intelligence (AI), the proposed framework will provide unprecedented levels of transparency, traceability, and security.
Enhancement of an Augmentative Alternative Communication Aid Prescription
Supervisors: Dr Peter O'Neill & Dr Abdel-Karim Al Tamimi
Within Augmentative Alternative Communication (AAC) there are three basic methods used to represent language these are: single meaning pictures, alphabet-based systems and semantic compaction (Romich, Vanderheiden and Hill, 2000). Apart from alphabet-based systems the others are non-literacy systems, which means they can be taught to pre-school aged children. The semantic compaction system requires the user to combined symbols (cells), normally no more than two, to create a word / phrase.
The focus of this project will be on users with single input capabilities but will not be exclusive, which inherently means these single switch users are the most non-productive, thus the neediest user group to enhance their user output. Historically, the enhancement of these systems has been twofold: first generating data logs while the system is in use (O’Neill P, 2006 & Horstmann H M, Levine S P, 1990) and analysing these logs to try and predict the user’s future selection/s: secondly, with the selection of multiple cells to generate a single word, or phrase, reducing the number of subsets of cells.
Securing SME Networks through SASE: Integrating SD-WAN and Zero Trust for Scalable, Cost-Effective Security
Supervisor: Dr. Abdussalam Salama
Small and Medium Enterprises (SMEs) are increasingly embracing digital transformation, driven by the rise in cloud services, remote work, and BYOD (Bring Your Own Device) practices. These trends demand modernised network architectures that offer both agility and robust security. Traditional perimeter-based security models are proving insufficient, especially for SMEs operating with limited IT resources and budgets. Secure Access Service Edge (SASE) offers an innovative, cloud-native framework that integrates Software-Defined Wide Area Networking (SD-WAN) with key security functions such as Zero Trust Network Access (ZTNA), Firewall-as-a-Service (FWaaS), and Multi-Factor Authentication (MFA).
This project aims to develop a lightweight and modular SASE framework tailored to SME environments. It will be designed to deliver scalable, cost-effective, and user-friendly security without requiring significant infrastructure investment. The research will follow an applied methodology combining simulations, physical labs, and cloud environments to test and validate the solution under real-world conditions. Key metrics such as security, performance, usability, and user experience will be evaluated against traditional network setups to assess the practical benefits for SME adoption.
Investigating Facial Dynamics of Micro Expression
Supervisor: Dr Pratikshya Sharma
This project aims to investigate the temporal facial dynamics of micro expressions with an intention to build an end-to-end micro expression recognition system. It shall explore in detail the changing features of the face that occur during emotional expression. It will also examine the underlying motion patterns associated with facial micro expressions.
To investigate facial dynamics of micro expressions (ME) and build an end-to-end automatic ME recognition system, in this project computer vision and AI based novel methods shall be developed. For measuring facial movements objectively in ME Facial action coding system (FACS) can be experimented with. There is low inter class differences in ME due to the low intensity, therefore concepts of action unit (AU) analysis shall also be examined. This could aid in reducing ambiguity in representing distinct expressions and enhance the performance of facial ME recognition systems. Advanced pre-processing methodologies will be examined to assess their influence on the accuracy and effectiveness of ME recognition. This project shall in particular utilize spontaneous micro-expressions to realise its aim.
Advanced Link Prediction for Dynamic Healthcare Systems
Supervisor: Dr Jaya Lakshmi Tangirala
This PhD research aims to explore the application of link prediction techniques in graph-based healthcare systems to uncover hidden relationships, identify emerging risks, and improve decision-making in clinical and patient-centered contexts. Link prediction, a fundamental problem in network science, seeks to infer missing, future, or potential associations between entities in a graph. In healthcare, this can translate to discovering unknown disease-disease associations, patient-drug interactions, comorbidity risks, or referral pathways. Leveraging machine learning, graph neural networks, and domain-specific knowledge graphs, this project seeks to develop interpretable and efficient algorithms that enhance the predictive and diagnostic capabilities of intelligent healthcare systems.
Strong programming skills in Python (preferred) or R; Familiarity with data science libraries (e.g., Pandas, NumPy, Scikit-learn) are required.
Graph-Based Machine Learning Approaches for Mental Health Analysis and Early Intervention
Supervisor: Dr Jaya Lakshmi Tangirala
Mental health disorders such as depression, anxiety, and bipolar disorder affect hundreds of millions globally. However, the diagnosis and treatment of mental illnesses remain complex due to the multifactorial nature of these conditions such as biological, psychological, social, and environmental factors. Recent advances in data science have opened up possibilities to model these complex interdependencies using graph theory, where individuals, symptoms, risk factors, and treatments can be represented as nodes and edges. Graph-based approaches can reveal hidden patterns, predict mental health deterioration, and personalize interventions. While data on mental health is growing, it is often sparse, unstructured, and disconnected across platforms (e.g., clinical records, wearable sensors, social media, therapy logs). There is a lack of unified frameworks that integrate multi-source mental health data using graph models, predicts mental health outcomes with high accuracy and interpretability, and detects early signs of deterioration or crisis using dynamic graph models.
Strong programming skills in Python (preferred) or R; Familiarity with data science libraries (e.g., Pandas, NumPy, Scikit-learn) are required.
Intelligent Recommender Systems for Healthcare: Enhancing Clinical Decision Support, Patient Engagement, and Treatment Personalization
Supervisor: Dr Jaya Lakshmi Tangirala
This PhD proposal advances AI-powered recommender systems for healthcare, bridging gaps in personalized medicine and clinical decision-making. By combining recommender systems, knowledge graphs, and interpretable AI, this work aims to improve patient outcomes while ensuring safety and fairness. Healthcare decision-making involves navigating vast amounts of patient data, medical literature, and treatment options. Traditional clinical decision support systems (CDSS) often rely on rule-based approaches, which may not adapt to individual patient needs or evolving medical knowledge. Recommender systems (RS), widely used in e-commerce and entertainment, can be adapted to healthcare to provide personalized treatment suggestions, medication recommendations, and patient engagement strategies.
Strong programming skills in Python (preferred) or R; Familiarity with data science libraries (e.g., Pandas, NumPy, Scikit-learn) are required.
Scalable Distributed Algorithms for Link Prediction in Large-Scale Graphs
Supervisor: Dr Jaya Lakshmi Tangirala
Link prediction is a key problem in network science and machine learning, aiming to identify missing or future connections in a graph based on existing structure and node features. It has a wide range of applications including: Social network growth modeling; Recommender systems (e.g., "People You May Know"); Knowledge graph completion and Drug-target interaction prediction in bioinformatics etc. However, most link prediction methods are designed for centralized settings and do not scale well to massive graphs with billions of nodes and edges. Distributed algorithms offer a promising approach to scale link prediction in a fault-tolerant and efficient way. The core motivation of this research is to develop distributed, scalable, and accurate link prediction algorithms, capable of running on massive, real-world graphs under realistic network constraints.
Strong programming skills in Python (preferred) or R; Familiarity with data science libraries (e.g., Pandas, NumPy, Scikit-learn) are required