Protecting machine learning IP with watermarking tool

NXP Semiconductors now offers a new eIQ Model Watermarking tool to its eIQ Toolkit for machine learning development. The new tool is the industry’s first effective tool to assist developers in protecting their machine learning investments. The tool embeds a watermark in a machine learning model, adding copyright ownership to the model and enabling developers to verify that a machine learning model is a copy or clone of their model IP without needing access to that model’s code.

It’s been communicated that data is the new gold. Nowhere is that truer than in machine learning, where a model’s quality depends heavily on the quality of the data it is trained with. Domain expertise is crucial in creating highly efficient models. Though they are a valuable and differentiating asset to a company, machine learning models generally lack the copyright protection that shields ordinary software from unauthorised copying or cloning. The new tool enables developers to copyright their machine learning models while also allowing developers to detect unauthorised usage and protect their unique IP.