Spatial steganography changes smooth characteristics between adjoining pixels of the raw image. We present a novel steganalysis method for steganography based on feature vectors derived from the co-occurrence matrix in the spatial domain. We investigate how steganography affects the bit planes of an image and show that it changes more least significant bit (LSB) planes of it. The co-occurrence matrix is derived from an image in which some of its most significant bit planes are clipped. By this preprocessing, in addition to reducing the dimensions of the feature vector, the effects of embedding were also preserved. We compute the co-occurrence matrix in different directions and with different dependency and use the elements of the resulting co-occurrence matrix as features. This method is sensitive to the data embedding process. We use a Fisher linear discrimination (FLD) classifier and test our algorithm on different databases and embedding rates. We compare our scheme with the current steganalysis methods. It is shown that the proposed scheme outperforms the state-of-the-art methods in detecting the steganographic method for grayscale images.