A Scheme Based on Convolutional Neural Network and Fuzzy Logic to Identify the Location of Possible Secret Data in a Digital Image

Recently, there have been various proposals for improving the precision of steganalysis, which is the art of detecting the presence of a steganographic payload. In addition, a few existing research works focus on identifying the specific location of concealed data by a contemporary adaptive steganographic algorithm. This work presents a new algorithm that employs fuzzy logic and a Convolutional Neural Network (CNN) to reveal any hidden information within the content of a digital image. The proposed model comprises two primary components: a Mamdani-based inference module to generate the stego image’s fuzzy correlations and a CNN module that classifies the image's features to locate the locations of the steganographic payload. The method uses recall rate, precision rate, and accuracy for evaluation metrics, demonstrating superior performance compared to the existing models. The experimental results identify the proposed approach's outperformance over the existing approaches. Notably, locating the payload hidden under WOW achieves an accuracy superior to 90% with a payload of 0.5 bpp, which indicates that it can accurately identify almost all the modified pixels.