Security of Neural Networks
Defending deep models against adversarial examples, backdoor and data-poisoning attacks — and understanding when and why they fail.
Research studies, software and prototypes from across the NPerseus team.
Research studies, prototypes and software described in public profiles. Status, roles and funding follow each source record; a profile entry does not by itself establish a released product.
74 matching profile project records · Page 3 of 7
research on reversible data-hiding and deep-learning methods for protecting medical-image content.
contributed to a React and Express paper-submission and review-workflow application; its repository explicitly marks the project abandoned and no longer maintained.
contributed Kotlin application and layout changes to a hosting-services app; its repository explicitly marks the project abandoned.
Ph.D. research on adversarial robustness and trustworthy AI, including reversible adversarial perturbations.
Developing Esperal, an offline single-player game, using third-party tools and libraries. The game is in development.
Implemented forex-pair administration features, including backend APIs, a database model and an admin interface, within a Bicrypto-based exchange platform.
developed a MobileNetV2-based presentation-attack detection pipeline on the LCC FASD dataset, with augmentation and preprocessing for varied spoofing conditions.
Learning/prototype experiment using the Flexi-Store Kotlin e-commerce client, administration and backend components.
circular timer with notification-area play/pause controls.
prioritized notes, Sqflite storage and persistent dark theme.
Java and Spring Boot developer tool for retrieving public repositories and inspecting source structure with configurable heuristic complexity scores.
Built developer tooling for grid-based code-golf tasks, including a task browser, solution comparison, example-based validation and code-size analysis, using third-party tools and libraries.
Defending deep models against adversarial examples, backdoor and data-poisoning attacks — and understanding when and why they fail.
Reversible data hiding in encrypted images, image and text steganography, and robust watermarking for ownership and integrity.
Detecting manipulated and GAN-generated media — deepfakes, splicing and median-filtering traces.
Federated and privacy-preserving learning, explainable and trustworthy models, and methods for limiting the disclosure of training data.