Steganalysis based on feature reducts of rough set by using genetic algorithm

The supervised learning based statistical detection is generally used in steganalysis. Compared to the specific detecting method, this method has the advantages of flexibility and ability to be quickly adjusted to new or completely unknown steganalytic method. Otherwise, it has the disadvantages in large-scale data, low calculate speed. Knowledge reduction can delete the non-important knowledge while not alter the classification ability of knowledge. While, it is difficult for the original rough set to find out the minimal reduct when dealing with the large-scale and high-attribute data. The GA can solve this matter. It is proved that the speed of the detection system is improved by GA reduction while the ability of classification can be preserved as the formerly level.