CONTENT ADAPTIVE STEGANOGRAPHY - DESIGN AND DETECTION
Vojtöch Holub · 2014
iii Currently, the most successful approach to steganography in empirical objects, such as digital images, is to embed the payload while minimizing a suitably defined distortion function. This powerful concept allows the steganographer to evaluate distortion caused by embedding changes based on local image content, hence the name content adaptive steganography. The design of the distortion is essentially the only task left to the steganographer since efficient practical codes exist that embed near the payload–distortion bound. One of the contributions of this dissertation is a novel approach to steganography of JPEG images. Instead of attempting to preserve an inherently incomplete heuristic model of DCT coefficients, we design a simple distortion in the better-understood spatial domain and use it for computing the distortion caused by modifying JPEG coefficients. Besides avoiding the difficult task of modeling dependencies among JPEG coefficients, JPEG steganography using spatial domain distortion pro-vides a superior security with respect to other current state-of-the-art methods. Virtually all current steganographic methods are implemented with additive distortion functions. However, embedding