Research area
Security of and with machine learning
How models and AI applications are attacked, and how machine learning detects threats.
The question
How do machine-learning systems fail under attack, and how can machine learning help defend other systems?
Our angle
How we approach it.
We study attacks to build defences that hold up in deployed systems, not only in the lab. The work goes straight into how we test AI applications for clients.
What it covers
- Adversarial attacks on models and AI applications
- Prompt injection and LLM application security
- Machine learning for threat detection
Outcomes
- TBC · outcome or result that can be evidenced
- TBC · outcome or result that can be evidenced