Copyright Protection of Deep Neural Network Models Based on Watermarking. This project aims to develop robust watermarking techniques for copyright protection of deep neural network models. Existing t

Description

Copyright Protection of Deep Neural Network Models Based on Watermarking. This project aims to develop robust watermarking techniques for copyright protection of deep neural network models. Existing techniques to protect this copyright are preliminary, difficult to apply and susceptible to abuse. This project employs novel approaches, methods and techniques to solve these shortfalls. Expected outcomes include a set of innovative mechanisms and algorithms that provide solutions to the key challenges in deep neural network watermarking. Expected benefits include significant advance in deep neural network watermarking research, greater commercial opportunities, expansion of deep learning for real-world applications and economic gain through greater protection of intellectual investment.. Scheme: Discovery Projects. Field: 4604 - Cybersecurity and Privacy. Lead: Prof Yong Xiang