Verified security remediation
A fix needs evidence.
How can an AI-assisted fix be checked for both security improvement and correct software behaviour?
A proposed patch needs more than a plausible explanation. This exploratory area investigates how AI can connect a vulnerability finding to a candidate repair, a targeted security test and checks for unintended changes. It also asks how developers can judge remaining uncertainty before accepting a fix.
Research areas
Connect the cause and conditions of a vulnerability to the smallest appropriate repair.
Generate targeted tests that distinguish the vulnerable behaviour from the intended secure behaviour.
Check compatibility, related code paths and regression risk when a repair changes shared functionality.
Compare candidate repairs using security evidence and functional tests, and explain when automatic verification is insufficient.
Publications & related resources
Projects, publications and research materials will be added here.
Contribute to this research.
Explore student research, internships and collaboration with the lab.