Privacy-Preserving AI & Homomorphic Encryption
Secure artificial intelligence and machine learning over sensitive data using Homomorphic Encryption, encrypted inference and privacy-enhancing technologies.
RESEARCH
Secure, Privacy-Preserving & Trustworthy Computing
Research at the intersection of artificial intelligence, homomorphic encryption, cybersecurity, secure systems and emerging cryptographic technologies.
4
Core Research Themes
RESEARCH OVERVIEW
Designing Intelligent Systems That Preserve Trust
My research focuses on the design, evaluation and practical deployment of secure, privacy-preserving and trustworthy computing systems, particularly where artificial intelligence interacts with sensitive data.
A central theme of my work is the use of Homomorphic Encryption and related privacy-enhancing technologies to enable useful computation without exposing confidential information.
This has led to research in encrypted clinical prediction, privacy-preserving machine learning, federated learning, secure agentic AI, cybersecurity and emerging cryptographic systems.
My broader objective is to bridge the gap between theoretical security guarantees and deployable real-world systems, with emphasis on reproducibility, measurable performance, privacy protection and practical applicability.
Research Goal
To develop intelligent systems that can operate on sensitive information while preserving privacy, security, trust and practical usability.
RESEARCH THEMES
Core Areas of Investigation
Cybersecurity & Secure Systems
Insider-threat mitigation, information leakage, secure workflows, access control, traceability and privacy-aware system design.
Agentic & Trustworthy AI
Security, privacy, policy control, auditing and governance for increasingly autonomous and agentic artificial intelligence systems.
Emerging Cryptographic Security
Post-quantum cryptography, cryptographic migration, secure computation and next-generation security readiness.
FEATURED RESEARCH PROJECTS
Selected Research Initiatives
FLAGSHIP RESEARCH
Privacy-Preserving Agentic AI for Secure Clinical Decision Intelligence
A privacy-preserving agentic AI framework integrating Homomorphic Encryption with policy-controlled intelligent workflows for secure clinical prediction over sensitive health information.
The framework investigates whether AI agents can coordinate authorised access, encrypted inference, controlled decryption, policy enforcement and audit logging while maintaining confidentiality of patient data.
SustainHE-Med
A sustainable Homomorphic Encryption framework investigating secure clinical prediction in resource-constrained computing environments.
Research emphasis: encrypted prediction, computational efficiency, privacy protection and practical deployment.
View Repository →Federated Learning vs Homomorphic Encryption
A comparative investigation of centralised learning, Federated Learning and CKKS-encrypted inference for privacy-preserving clinical prediction.
Research emphasis: predictive performance, privacy, computational runtime and communication overhead.
View Repository →ECT-AMF
An Exposure-Control and Traceability framework addressing insider information leakage during sensitive academic assessment workflows.
Research emphasis: controlled exposure, access control, watermarking, traceability and insider-risk mitigation.
Post-Quantum Cryptography Readiness
Research examining institutional and technological preparedness for migration toward quantum-resistant cryptographic systems.
Research emphasis: security readiness, migration planning, cryptographic resilience and infrastructure preparedness.
RESEARCH FRAMEWORKS & SYSTEMS
From Research Questions to Technical Architectures
Encrypted Clinical Inference Pipeline
Sensitive clinical data remain encrypted while privacy-preserving prediction is performed using Homomorphic Encryption.
DATA → ENCRYPT → HE INFERENCE → DECRYPT → RESULT
Secure Agentic AI Workflow
Policy-controlled agents coordinate authentication, authorisation, encrypted computation, controlled decryption and audit logging.
REQUEST → POLICY → AGENT → HE → AUDIT
FL–HE Comparative Architecture
A common experimental architecture for evaluating centralised learning, federated training and encrypted inference.
CENTRAL → FL → CKKS → COMPARE
ECT-AMF Security Framework
A secure workflow combining controlled access, limited exposure, traceability and incident-response mechanisms.
ACCESS → EXPOSURE → TRACE → RESPOND
REPRODUCIBILITY & OPEN RESEARCH
Transparent, Reproducible Experimental Research
Reproducibility is an important component of my research practice. Where appropriate, experimental code, workflows, implementation artefacts and supporting materials are organised into public repositories to support transparency, validation and future extension.
RESEARCH COLLABORATION
Interested in Collaborating?
I welcome collaboration with academics, research groups, students and institutions working in privacy-preserving AI, Homomorphic Encryption, cybersecurity, secure computing, agentic AI and emerging cryptographic technologies.
Areas of interest include joint research projects, co-authored publications, comparative experimental studies, student research supervision, research visits and interdisciplinary partnerships.