High-dimensional statistical supervised learning for extracting information in steganography
Sunjun Hwang, Kyeong Seon Kim, Yejin Kim, Junhui Kim, Miji Park, Soohyun Park, Joongheon Kim · 2018
One of major research topics in digital steganography is the methods which can avoid information loss while the image files experience distortion. The image distortion may lead to the key value distortion, and eventually, it becomes very hard to get correct information from the distorted images. Therefore, this paper proposes a scheme which encodes and decodes the given images with high-dimensional regression-based supervised learning. With this proposed method, several original messages are embedded in a file; and high-dimensional regression-based supervised learning conducts for extracting exact one final message from the several original messages where every single original message is not same with the final message. In this case, if attackers read some messages from the image, it is hard to extract our final message.