Signal noise and its relevance in adaptive steganography and steganalysis

Steganography may serve as a viable solution for the government and military to maintain the integrity of classified and sensitive material or allow strategic communication or coordination between our internal entities or between the U.S. and our allies. However, some would use steganography with malicious intent, such as terrorist trying to secretly communicate and coordinate their horrendous attacks. There is an obvious need for more secure methods of communication and better means of detecting steganography-bearing media. Signal noise exists in all aspects of today's world. Digital imagery, for example, will always have a natural level of noise caused by quantization error during capture, unstable camera electronics, and from noisy transmission channels. Since inserting steganography into a digital image is very similar to inserting a small level of noise, by understanding noise models and the principles of noise detection, we can improve methods of inserting secret information into a digital image. A fundamental understanding of the characteristics of noise will also allow us to distinguish signal noise from 'steganography noise' and thus improve our detection accuracy. This thesis will focus on accomplishing the following 3 objectives: (1) Developing a filtering method that will accurately identify and classify image pixels corrupted by noise and appropriately filter them, (2) Develop new steganography with more embedding capacity and higher security, and (3) Develop new steganalysis techniques that can detect hidden information with a high degree of accuracy and be useful for detecting a large range of embedding methods.