This thesis explores DeepSteg, a novel approach that combines deepfake technology with steganography to enhance the security of embedded information within digital media. By using deepfake technology as a cover for steganographic content, the study aims to improve the imperceptibility and capacity of hidden data.

The research involves encoding data with steganography, generating deepfakes with encoded data, and decoding the information, utilizing the "Human Face Object Detection" dataset from Kaggle. Evaluation metrics like MSE, PSNR, and SSIM indicate that DeepSteg outperforms traditional steganography in maintaining image quality and fidelity, making it a robust method for secure data embedding.