Image steganography method and system based on random texture

Abstract

The embodiment of the invention discloses an image steganography method and system based on random textures, and the method comprises the steps: inputting a random hidden variable and a face image into a GAN network, and obtaining face data with consistent contents and different random textures, wherein the GAN network is trained based on a large-scale face data set, and Gaussian noise b is introduced into AdaIn operation by the GAN network so as to control pixel-level textures; recording noise data B as a batch identifier ID of the generated face data in the process that the GAN generates the face data each time; carrying out the texture classification by training a classification network D, wherein a classification label is a batch identification ID, so that source tracing is realized for different batches of generated contents; wherein the training data is generated by adding set noise to a plurality of random implicit vectors. Source tracing can be quickly and accurately realized for different distribution batches of generated contents.

Description

technical field [0001] The embodiment of the present application relates to the field of deep learning technology, and in particular to an image steganography method and system based on random texture. Background technique [0002] Image steganography refers to adding specific information to an image without attracting the attention of the viewer, and using corresponding technical means to extract the hidden information. This technology has numerous application scenarios, including digital content version control. Due to the reproducibility of digital content, it is difficult to identify its source from the content itself, and it is impossible to trace the source of the content. Through the image steganography technology, the ID information of the copy holder can be added to the digital content copy, so that when a digital content dispute occurs, the ID information can be extracted for traceability and evidence collection. [0003] Digital watermarking is a kind of image s...

AI Technical Summary

Problems solved by technology

However, watermark-based methods are not robust against some tampering attacks, such as image compression, noise addition, multi-watermark mixing, etc. After such operations, it is difficult to extract the original watermark from the image, and the current network communication is characterized by image It may be edited multiple times, which makes it impossible to maintain the watermark information carried by the image during transmission, resulting in a significant reduction in the probability of successfully extracting the watermark

Images