Steganalysis in High-Dimensional Feature Space Using Selective Ensemble Classifiers

Abstract

Today, modern steganalysis needs to start using high-dimensional feature spaces, which makes the complexity of traditional classifiers such as support vector machine (SVM) increase rapidly. This paper proposes a frame of selective ensemble classifiers as an alternative to SVM for steganalysis by applying the selective theory that ensemble some instead of all the available base learners. A family of weak classifiers is built on random subspaces of the high-dimensional feature spaces. Then, assign a random weight to each classifier and employ genetic algorithm to evolve those weights based on a validation set. The Final classifier is constructed by fusing the decisions of individual classifiers whose weight is bigger than a pre-set threshold λ. Experiments with the steganographic algorithms nsF5 and MBS demonstrate the usefulness of the approach over current popular methods.