Research
My research explores how artificial intelligence and distributed systems can be designed and governed to remain secure, privacy-preserving, trustworthy, and meaningfully accountable to people.
Research interests
Trustworthy AI Governance
Human authority, accountability, transparency, auditability, and organisational oversight of intelligent systems.
Privacy-Preserving Federated Learning
Adaptive privacy methods that protect participant data while maintaining useful collaborative learning.
Secure Peer-to-Peer Systems
Decentralised learning and trust architectures that do not depend entirely on a central coordinator.
AI-Driven Incident Response
Governance, human oversight, and accountability when AI is used to support cybersecurity incident response.
Public-Sector Digital Trust
Responsible adoption of AI and digital technologies in government and public-sector organisations.
Featured research
Adaptive Trust-Aware Privacy Framework for Secure Peer-to-Peer Federated Learning
- Research problem
- Peer-to-peer federated learning removes the central coordinator that classical federated learning depends on — and with it, the single point where privacy budgets, participant trust, and update integrity are managed. Existing approaches typically assume static privacy parameters and fixed trust relationships, which fail in dynamic networks where participants join, leave, misbehave, or become compromised.
- Aim
- To design and evaluate a framework in which privacy protection and trust assessment adapt continuously to observed peer behaviour, allowing decentralised collaborative learning that remains secure, private, and useful without any central authority.
- Proposed contribution
- An integrated model combining adaptive differential privacy (protection that tightens as trust signals weaken), behaviour-based decentralised trust evaluation, and secure peer-to-peer exchange protocols — evaluated against poisoning, inference, and free-riding adversaries in realistic network conditions.
- Core research themes
- Adaptive privacy mechanisms · decentralised trust models · secure aggregation without a coordinator · robustness under adversarial participation · the privacy–utility–trust trade-off.
- Current status
- Research proposal in development
Publications and working papers
- Concept Note
Adaptive Trust-Aware Privacy Framework for Secure Peer-to-Peer Federated Learning
Adaptive privacy mechanisms and decentralised trust models for collaborative learning without a central authority.
- Working Paper
Governing AI-Driven Incident Response in Public-Sector Organisations
Human oversight, escalation design, and accountability when AI supports cybersecurity incident response in government.
- Manuscript in Development
Symbolic versus Substantive AI Governance
Distinguishing governance that changes decisions from governance that documents them, with implications for non-Western regulatory contexts.
- Research Framework
Governance and Accountability Framework for Responsible AI Adoption
A structured framework for keeping AI-supported decisions accountable, explainable, and auditable across their lifecycle.