Steganalysis in Spatial Domain Images for Securing Data Transmission in IoT Environments

Ensuring strong security and privacy in IoT applications is vital as the number of IoT devices grows, introducing the possibility of illegal data transmission. Unlawful data sharing uses several approaches, such as steganography, which consists of concealing confidential information in the cover medium, which may be audio, video, image, etc. Based on the ubiquitousness of images in IoT environments, this research intends to mitigate the possible use of IoT environments to transmit harmful data using images by proposing a steganalysis method to detect hidden data in them. The method proposed in this work leverages the convolutional neural network (CNN) paradigm with enhanced feature learning based on stochastic feature selection (SFS) and multinomial distribution. The stochastic components contribute to avoiding the overfitting issue, and the use of multinomial distribution enhances the learning ability of the CNN model to improve detection accuracy. The experimental outcomes, attaining an accuracy of 95.2 % in detecting the existence of confidential information, show a noteworthy efficacy of the proposed method to prevent an unnoticed transmission of secret data using IoT environments.