Large Language Models (LLMs) are rapidly transforming business operations through intelligent assistants, autonomous agents, decision support systems, and process automation. However, the integration of LLM agents into enterprise environments introduces significant cybersecurity, privacy, governance, and trust challenges. Businesses require intelligent systems that can autonomously assist employees while remaining secure, explainable, compliant, and resilient against cyber threats.
This PhD project investigates the development of trustworthy LLM-based cybersecurity agents capable of monitoring, analysing, and responding to security threats across complex business environments. The research will explore how autonomous AI agents can support Security Operations Centres (SOCs), identify emerging cyber risks, detect anomalous activities, assess vulnerabilities, and provide explainable recommendations to decision-makers.
Key research challenges include:
- Design of multi-agent LLM systems for enterprise cybersecurity.
- Detection of phishing, social engineering, insider threats, and advanced persistent threats.- Secure retrieval-augmented generation (RAG) architectures for organisational knowledge bases.
- Explainable AI techniques enabling transparency and auditability of agent decisions.
- Adversarial robustness against prompt injection, data poisoning, model manipulation, and jailbreak attacks.
- Human-AI collaboration frameworks supporting effective cybersecurity decision making.
- Integration of cybersecurity regulations, governance policies, and compliance requirements into AI-driven security operations.
- Development of risk-aware autonomous agents capable of balancing operational efficiency and security assurance.
- Privacy-preserving learning and secure management of sensitive organisational data.
The successful candidate will develop novel AI methodologies combining large language models, machine learning, cybersecurity analytics, knowledge graphs, agentic AI architectures, and explainable artificial intelligence (XAI). The project is expected to contribute both theoretical advances and practical solutions applicable to finance, healthcare, manufacturing, government, and critical infrastructure sectors.
Potential research outcomes include intelligent cyber-defence platforms, explainable security decision-support tools, autonomous threat-hunting agents, and frameworks for secure deployment of AI agents in business environments.
Applicants should have a strong background in Computer Science, Artificial Intelligence, Cybersecurity, Data Science, Machine Learning, Software Engineering, or a related discipline.