Depth adaptive image hiding method based on adversarial sample generation

Abstract

The invention relates to a depth adaptive image hiding method based on adversarial sample generation, which comprises the following steps: firstly, designing an experimental sample, and determining asecret graph and a cover graph for receiving hidden information; secondly, establishing an SSD network selected by a local hidden patch, and finding out an area most suitable for hiding the secret graph in the cover graph; then, establishing an encode network, generating disturbance through the encode network by using an adversarial sample method for the secret graph, and directly adding the disturbance to a selected area of the cover graph to generate a container graph with hidden information; then, establishing a code network, and solving a real graph which is highly similar to the set graph; and finally, training and testing an encode network and a code network, and updating a coding network for each group of inputs to realize personalized hiding of each group of images. Through the above three main steps, the method can achieve the adaptive local hiding of the image while maintaining the good hiding and restoring effect of coding, improves the reconstruction quality of image hiding, and enlarges the actual application range of image hiding.

Description

technical field [0001] The present invention realizes more flexible, larger capacity and safer information transmission through image hiding. Aiming at the problems of chromatic aberration in image hiding and the fact that image hiding can only achieve the effect of global hiding, a deep Adaptive image hiding methods. Background technique [0002] The most popular image hiding algorithm is based on the least significant bit (LSB) method. The image to be hidden is defined as a secret image, the image that accepts hidden information is called a cover image, and the result image that adds hidden information through the encode network is a container image. The image parsed from the container image is a revealed image. The main idea of ​​these LSB-based algorithms is to change at least 4 significant bits of the cover image to place hidden information. In this way, the color change of the cover image during concealment can be minimized, and the resulting distortions are usually ...

AI Technical Summary

Problems solved by technology

The main reason for this problem is that once the model is trained, the parameters of the encode network and the decode network are all fixed, and the whole is a generalized model. Each group of test pictures uses the same weight, so the rules are easy be found

Images