Hybrid CNN and Ensemble Learning-based Steganalysis Framework for Detecting Hidden Data in Social Media Image

The widespread use of social media platforms such as WhatsApp, Instagram, and Twitter has led to an unprecedented volume of image sharing, making them potential carriers of hidden information through steganography. Steganography enables malicious actors to embed secret messages or sensitive data within seemingly harmless images, posing serious threats to cybersecurity, digital forensics, and information assurance. Traditional detection methods often struggle to cope with modern embedding techniques, particularly those enhanced by deep learning, which can conceal data more effectively. In this paper, we propose an AIpowered steganography detection framework that utilizes deep learning models to accurately distinguish between clean and stego images. A custom dataset is created by embedding hidden information using classical methods such as Least Significant Bit (LSB) and Discrete Cosine Transform (DCT), as well as advanced deep steganography approaches. Convolutional Neural Networks (CNNs) are trained to capture subtle pixel-level anomalies and high-frequency noise patterns that are invisible to the human eye but indicative of hidden data. The system is further tested for robustness against distortions common in social media environments, including image compression, resizing, and filtering. Experimental results demonstrate that the proposed approach achieves high accuracy, offering a practical solution for enhancing social media security and preventing covert communication.