Abstract: In his research Bashmakov analyzes accuracy of background area selection in static digital images by using the histogram method as part of steganalysis performed by Weighted Stego Image and WSPAM methods. He examines the dependence of practical accuracy of steganalysis of static digital images by using Weighted Stego Image and  WSPAM methods on the kind of prediction model in gradient regions of an image as part of resistance to data transmission channels that use the method of embedding the least significant bit of spatial domain in static digital images with a significant part of homogeneous background. The author analyzes the Weighted Stego steganalysis algorithm and WSPAM modification thereof. To evaluate the analysis efficiency, the author has used the BOWS2 collection. To evaluate efficiency of homogenous background selection, the author has used images selected from a wide range of sources. The information is built in by changing the least significant bits of images in spatial domain with an actual load from 3-5%. Efficiency of methods is defined based on true-positive, true-negative, false-positive and false-negative values of image classification. The author demonstrates the low accuracy of homogenous background selection using the histogram method. The author suggests to select homogenous background using the segmentation neural net and proves its efficiency. He also offers an improved model of pixel prediction in image gradient areas, this model allowing to achieve the highest accuracy of steganalysis. The results of the research can be used to create systems of passive resistance to steganographic data transmission channels that are based on the Weighted Stego algorithm.