This chapter shows how to perform an analysis of a digital image to obtain information about the data that may have been hidden there. It also shows that steganalysis has been widely studied as a tool for evaluating steganography methods. Steganalysis can be subdivided into two distinct problems: the first relates to the extraction of relevant characteristics, and the second relates to learning to automatically classify, based on a vast set of examples, characteristics from Cover and Stego images. The detection methods are often less reliable than signature detection methods, but have the advantage of being much more general, in the sense that they aim to detect changes related to hiding information in the very content of a medium. The chapter describes how the two phases, features extraction and supervised learning, are usually implemented in steganography. It presents the latest developments in learning-based steganalysis methods, namely, the use of deep neural networks.