Analysis of Compliance Issues of Machine Learning Technology in Intellectual Property Protection and Its Legal Regulation

Abstract

The deepening application of machine learning technology in intellectual property protection makes its compliance gradually become a research hotspot in related fields. This paper uses deep neural network technology to design a digital watermark generation method for intellectual property. And based on the deep neural network watermarking method, the overall system framework of serial number watermarking is constructed. The framework maintains structural features consistent with the neural network output layer through specific matrix operations, thus effectively integrating serial number and deep learning technology. Meanwhile, in order to strengthen the compliance of the deep learning technique, in the generation process of key sample watermarking, the encoder is used to receive normal samples and exclusive signs and output key samples, and the discriminator is utilized to receive normal samples and key samples. As a result, the construction of the digital watermark generation model based on deep neural network is completed. The F1 value of the model is always maintained at 0.7 and above at different Hamming distances, and the maximum is up to 0.973. It shows extremely superior operational performance, which provides a good foundation for its adaptation in intellectual property protection.