An Improved Image Steganalysis Using a Novel Feature Selection Algorithm Based on Artificial Bee Colony

One of the most important phases of pre-processing is Feature selection, which can improve the predictive accuracy of steganalysis. In this study, we have presented a novel feature selection-based method to image steganalysis for detecting stego images from cover images based on artificial bee colony (ISBC). The experiments show that the proposed method is easy to be employed for steganalysis purposes. Moreover, its overall performance is better than recent ABC-based feature selection methods. 2 PROPOSED METHOD (ISBC) This section presents a new method to resolve feature selection problems in steganalysis, which the paper proposes two subsections. Section 2.1 refers to a schematic method to find the best subset features. Section 2.2 briefly explains the structure of presenting improved artificial bee colony based feature selection. The General structure of presenting a method contains three important steps. These steps are phases of ISBC respectively feature extractor, artificial bee colony as a feature selection, support vector machine .This structure illustrates how to employ ISBC, and first of all, it is needed to provide a feature vectors‟ dataset extracted from a lot of images. For this issue we used a feature extractor (SPAM) to solve that issue, we will discuss it in the following section. After the feature extractor step, next we need to provide a method to select relevant features using ABC and compute fitness using SVM to evaluate the selected subset features and the outcome comes to ABC to validate the result, next if the condition is met or pre-determined number of cycles is reached, the process will finish. 2.1 Structure of proposed feature selection approach Generally, a particular feature selection algorithm surrounds four factors: a subset generation, a fitness function, a stopping condition in ABC, and a validation process. 2.2 A novel artificial bee colony algorithm for feature selection (ISBC) In the presented ABC-based feature selection method, ABC algorithm improves the procedure of feature selection and generates the ideal feature subset that improves the performance of the classifier. Figure 1 reveals the overall pseudo code of implementing feature selection using ABC. ABC is used as a feature selector and produces the feature subsets and a classifier is employed to evaluate every feature subset built by the onlookers; so, the presented approach is a kind of wrapper based systems that we are going to explain more as follows: 2.2.1 Initial population In this paper, the ISBC is used to explore the new search space. Initial swarm is sometimes created randomly. We are going to set parameters, ABC parameters contain the number of food source, the number of colony size, lower bound, upper bound, limit , max cycle. The population of employed bees and onlooker bees are equal to the dimension of features. After that, we use one of the best classifier Support Vector Machine (SVM) to evaluate the predictive accuracy of selected feature to obtain the discriminating ability of every single feature in the dataset. The accuracy (xr) of per feature I is computed by using 10-fold cross validation. After that, the objective (Fr) is computed for per feature from its unknown and implicit relation.