A novel steganalysis algorithm based on paired sampling and selective ensemble is proposed to improve detection accuracy of steganalysis.Considering the characteristics of ense mble classification and the particularity of steganalysis,we analyze the relation between two classes of s amples and the link among samples of the same class,put forward four different sampling strategies,and build selective ensemble classifier based on paired sampling strategy for steganalysis．Experimental results showed that different s ampling strategies could influence the performance of detection in steganalysis to some extent.Compared with the exist ing methods in steganalysis,the proposed method can significantly reduce the error rate in detection performance of steganalysis system．