Watermark adding method and device, electronic equipment and storage medium

Abstract

The invention discloses a watermark adding method and device, electronic equipment and a storage medium, and the method comprises the steps: determining N first values included in all digits of a first secret key; n is an integer greater than or equal to 1; positioning N first weights in the neural network model; each first weight in the N first weights corresponds to one first numerical value in the N numerical values; embedding a second key as a watermark into the neural network model; wherein the second secret key is obtained by replacing a first numerical value on each digit in the first secret key with a second numerical value of the corresponding first weight on a first set digit.

Description

technical field [0001] The present application relates to the field of encryption technology, and in particular to a watermark adding method, device, electronic equipment and storage medium. Background technique [0002] Neural network technology plays a vital role in the field of machine vision. Training a neural network model with excellent performance often needs to be based on a large number of high-quality image samples and a high-performance graphics processor (GPU, Graphics Processing Unit) computing cluster. , therefore, in order to ensure that the trained neural network model is not copied, redistributed or illegally used, a watermark can be added to the neural network model. In related technologies, adding a watermark to a neural network will have a negative impact on the performance of the neural network. Contents of the invention [0003] In order to solve related technical problems, embodiments of the present application provide a watermark adding method, dev...

AI Technical Summary

Problems solved by technology

In related technologies, adding a watermark to the neural network will have a negative impact on the performance of the neural network

Images