Based on wavelet package decomposition（WPD）

a new blind image steganalysis method was presented

which can detect the stego images with comparatively high accuracy.Firstly

by three scales of WPD

image was decomposed into some coefficient subbands

and the multi-order absolute characteristic function moments of histogram were extracted as features from these subbands and image itself.And then

these features were processed and a back-propagation（BP） neural network was designed to classify original and stego images.A series of experiments were made to validate the performance of proposed method for five kinds of typical steganography methods

including LSB

SS

Jsteg

F5 and MB.Results show the method can detect stego and original images reliably

and the average detection accuracy of this method exceeds those of its closest competitors by at least 7.5% and up to 17.2%.Moreover

the influence of integral and non-integral WPD for the detection accuracy was discussed.