The process of steganalysis comprises of two major steps: extraction of features and classification based on the features that have been extracted. The classification step of steganalysis process is tortuous and time-consuming as the feature sets are of high dimensions. This paper proposes a swarm optimization technique- Artificial Bee Colony (ABC) algorithm for feature selection. The proposed ABC feature selection algorithm selects reduced feature set. The Artificial Bee Colony, is a swarm optimization technique based on swarm intelligence nature of honeybees, that has been adopted and modified successfully for steganalytic feature selection in this work. ELM is used as the classifier and Fisher score is used as measure of separability. The experiment is conducted on two different set of JPEG images of 256 x 256 size each. To show the adequacy and potency of the algorithm, the experiments is conducted on four different types of steganalytic feature sets. The results of the experiment shows that the dimensionality of the features has reduced efficaciously and the detection accuracy of the steganalysis process has enhanced when the proposed feature selection method is used. The code is implemented in Matlab.