Steganalysis in Overlapping Images

Steganalysis aims to detect the presence of hidden information, known as steganography, in digital images. Modern detectors rely on machine learning techniques, using features from digital images to classify them as covers or stegos. Often the ability to detect steganography can be improved through a process called calibration. Calibration attempts to establish a reference for cover features independent of an image’s content. This reference can then be used to aid in the detection of steganography, which violates the feature characteristics of a cover through the presence its stego signal. This paper tests the novel idea that using independent images containing overlapping content for calibration can improve the detectability of steganographic images. To this end, a dataset including cover and stego objects of overlapping, uncompressed images was created and tested with state of the art steganalysis methods. In laboratory conditions I show that existing steganalysis can be increased by a factor between 2 and 10 when using cover images with overlapping content for calibration. The discovery of this new method of calibration suggests that state of the art steganalysis can be improved when overlapping images are present, however the extent is subject to further research.