Data processing method and device and method for identifying infringement of neural network model

The invention discloses a data processing method. The method comprises the steps of obtaining target watermark information; selecting a target convolution kernel from the first neural network model, and obtaining a first operation result of elements in a target convolution kernel matrix; and mapping the target watermark information to the target convolution kernel according to a mapping relationship between the watermark information and the first operation result of the elements in the convolution kernel matrix to obtain a target neural network model. The problem that in the prior art, a watermark embedding algorithm of a neural network model is large in storage overhead due to matrix storage is solved. 本申请公开了一种数据处理方法，包括：获得目标水印信息；从第一神经网络模型中选择出目标卷积核，获得目标卷积核矩阵中的元素的第一运算结果；根据水印信息与卷积核矩阵中的元素的第一运算结果的映射关系，将所述目标水印信息映射到所述目标卷积核中，获得目标神经网络模型。以解决现有技术下神经网络模型的水印嵌入算法存在的保存矩阵导致的存储开销较大的问题。