Disclosed is a spatial-domain characteristic-based digital image steganalysis method. The method comprises the following steps: for category information-marked digital images in a training set, calculating pixel grayscale difference value information on multiple directions in symmetrically adjacent domains of the digital images; processing and rotational invariance coding the pixel grayscale difference value information; then calculating a histogram of the coded information, normalizing the histogram to serve as characteristics; training and classifying the characteristics extracted, acquiring a classifier model; for an arbitrary digital image inputted, calculating adjacent domain information thereof to extract the characteristics; inputting the extracted characteristics into the classifier model, and acquiring classification information of the image inputted. The present method obviates the need for analysis of the characteristics of any specific steganography and is a universal method, the method therefore can achieve broad application in common (blind) steganalysis of digital images.
展开▼
