Excerpts from the Abstract:

Steganography is a relatively new and exciting field in the world of computer science. It involves embedding data into a medium in such a manner that it cannot be easily detected. This paper provides the reader with a basic overview of steganography. This overview includes the purpose, goals, and pre-digital history of steganography. The emphasis is on hiding data within images, with a focus on the robustness of the hidden message. Current spatial embedding techniques are not robust to basic compression algorithms. It examines the 'least-significant bit' paradigm which drives the spatial technique, and formulates a 'more-significant bit' paradigm in order to produce a morerobust method. Following this paradigm leads to the idea of transform embedding, hiding data in the frequency domain.

This paper proposes a transform embedding method (STEM) which uses existing ideas from digital watermarking, improving upon them with empirical data to make them suitable for steganography. This method is more robust than existing spatial embedding methods. Experimental results are presented which support this claim.