Security of Neural Networks
Defending deep models against adversarial examples, backdoor and data-poisoning attacks — and understanding when and why they fail.
Adversarial MLBackdoor DefenseRobustness
Research programme
Projects & research.
Released project records from NPerseus, alongside the lab's current evidence-backed research directions.
No individual project records have passed public release review yet. Current research directions are shown below without exposing internal work.
Data Hiding & Watermarking
Reversible data hiding in encrypted images, image and text steganography, and robust watermarking for ownership and integrity.
Reversible Data HidingSteganographyWatermarking
Multimedia Forensics
Detecting manipulated and GAN-generated media — deepfakes, splicing and median-filtering traces — to keep digital content verifiable.
Deepfake DetectionImage ForensicsAuthenticity
Trustworthy & Private AI
Federated and privacy-preserving learning, explainable and trustworthy models, and AI that respects the data it learns from.
Federated LearningExplainable AIPrivacy