Toolkit includes watermarking to protect machine learning IP
Machine learning IP is protected by the new eIQ Model Watermarking tool which has been added the eIQ toolkit by NXP.
The new tool embeds a watermark in a machine learning model, adding copyright ownership to the model. It also allows developers to prove that a machine learning model is a copy or clone of their model IP without requiring access to that model’s code.
A machine learning model’s quality depends on the quality of the data with which it is trained. Domain expertise is also critical in creating efficient models, advised NXP. “Developers invest huge amounts of time and resources in creating optimal models and it is important that they are able to protect that investment,” said Ali Osman Ors, director, global AI and machine learning strategy and technologies at NXP Semiconductors. Machine learning models typically lack the copyright protection which protects from unauthorised copying or cloning. eIQ Model Watermarking, said NXP, allows developers to copyright machine learning models, while also enabling developers to detect unauthorised use.
The eIQ Model Watermarking tool is designed to work with any vision-based machine learning model creation and is included as part of the eIQ Toolkit for machine learning development.
NXP Semiconductors specialises in secure connectivity for embedded applications in the automotive, industrial and IoT, mobile and communication infrastructure markets.
