The most important steganographic schemes are usually based on slight modifications in cover media. Two basic criteria in designing each embedding algorithm, which are quantitatively and simultaneously evaluated by the value of embedding efficiency, are security and capacity. In this dissertation, after thorough investigation of the main concepts of steganography and steganalysis, two novel families of steganographic schemes based on cover modification, are proposed. The first one, so called Bn, is a binary group of embedding methods which have better embedding efficiencies than the custom LSB replacement and LSB matching algorithms. In addition, a group of ternary embedding algorithms, so called Tn, with high value of embedding efficiency is proposed. All of the innovative steganographic schemes in this thesis are included in the kind of ±K embedding methods which can be modeled by additive noise information hiding. During the process of data embedding, the histogram of cover signal is softened by a low pass filter to generate the histogram of stego signal. The impulse response of this filter is the probability density function (PDF) of stegonoise signal. Using this fact, we suggest some methods whose equivalent filters have less effect on the histogram of cover signals, yielding more security and therefore less detectability against steganalysis attacks. Keywords Steganography, Embedding efficiency, Steganalysis, Power spectrum density of stegonoise, Receiver Operating Characteristic (ROC).
