엔트로피 분석과 서포트 벡터 머신을 이용한 LSB 스테가노그래피 판별

Abstract

This paper proposes a discrimination method of spatial steganography by analyzing LSBs characteristics. The image pixels are usually very similar to neighboring pixels, and the LSBs of neighboring pixels have high correlation. However, the MSBs and LSBs of a pixel do not show high correlation. The spatial steganography distributes the bits of original image pixels to the LSBs in the cover image pixels. Thus, the correlation between LSBs in the neighboring pixels are relatively low in the case of stego-images. This paper exploits this characteristics in the stegano-encoded images to discriminate the stego-images.