Suppose the information hider has a set of information hiding schemes to be employed in the presence of a set of steganalysis detectors. Let the information hider have only an incomplete information about the steganalysis detectors. That is, suppose the probability of the jth steganalysis detector will be used is pj, j = 1, 2 . . . n, then the information hider only knows the ordering of these probabilities, say, p1 p2 . . . pn. Under this circumstance how can the information hider choose the optimal embedding strategy? This is the question addressed in this paper. Theoretical analysis and some numerical results are presented.