Research theme
AI for Cybersecurity
AI that detects intrusions, malware and zero-day attacks before they reach hospitals, banks and homes.
-
TransCall: Catching Brand-New Malware by Watching How Programs Behave
TransCall uses a Transformer model to analyse system call sequences and detect zero-day malware, reaching an F1-score of 0.93 on malware families it had never seen.
-
Using Large Language Models to Make Sense of Conflicting Cyber Threat Data
An LLM-powered framework fuses threat intelligence from five sources into one Threat Confidence Index, matching expert consensus 90.8% of the time and cutting investigation time by…
-
IntruSafe: Guarding Hospital Devices Against Text and Image Attacks
IntruSafe combines FCNN and LSTM networks in a “sandwich” design to protect medical IoT networks, detecting intrusions with 97.66% accuracy and only 0.18% service interruption.