Abstract:
Steganography, the art of hiding information within harmless media, has seen significant progress from 2000 to 2025. This thesis presents a detailed comparison of steganography techniques used for both text and images. It reviews foundational methods such as Least Significant Bit (LSB) substitution and transform-domain embedding, along with recent innovations that involve deep learning and natural language processing (NLP). The analysis assesses these techniques based on factors like invisibility, data capacity, strength, and processing demand. Image steganography shows better data capacity and strength, especially with the use of convolutional neural networks (CNNs) and generative adversarial networks (GANs), which improve concealment against detection. On the other hand, text steganography provides subtle and context-sensitive data hiding through linguistic and AI-driven methods, but it has limited capacity and is vulnerable to formatting changes. The thesis also explores hybrid approaches that combine both media as promising future options. Lastly, the thesis addresses emerging challenges such as detection by adversaries, processing requirements, and the need for techniques that are safe against quantum threats. It offers a pathway for future research in secure and discreet communication
