Study of steganographic techniques for digital images

Abstract

Steganography, meaning hidden writing, is the technique of concealing secret information into another cover-media, such as audio, video, image and text, in such a manner that the very existence of the information is camouflaged. Unlike the goal of cryptography, which prevents the data from illicit access and modification by transforming the secret data into messy and meaningless bit-streams, the aim of steganography is to conceal their very existence such that the embedded data is inaudible or invisible to the grabbers. Using cryptographic techniques, the transformed secret bit-streams may draw the eavesdroppers’ attention, while using steganographic techniques, the interceptors may not notice the existence of the hidden data. In this research work, steganographic techniques for digital images, which are used as a cover-media, are investigated and analyzed. A rectified steganographic technique and two data hiding techniques are proposed so that the hiding capacity can be maximized without incurring perceptible degradation in the image quality and the embedded secret data can be recovered correctly without referencing the original cover-image. In the proposed rectified steganographic technique, the side information of neighboring pixels is used to estimate the number of bits which can be carried in the pixel of the cover-image. The pixel value is then replaced by the new value which includes the secret data. More secret bits are embedded in the edge area than those in the smooth area as the change in smooth area is more easily discovered by human eyes. Besides, the embedded secret data can be extracted completely from the resulting stego-image without any errors. Using the proposed technique, the absolute value of the greatest difference between the original pixel value and the new pixel value is 2n −1 when n-bit secret bits are embedded into a pixel. It is the same as using a simple Least-Significant-Bit (LSB) substitution method. Therefore, a data hiding technique using dynamic-length LSB substitution is proposed to reduce the difference value so that the distortions occurred in the smooth areas can be lessened. In the first proposed technique, two neighboring pixels are used to determine the number of bits to be embedded into the pixel. The pixels in the edge areas can embed more bits than those in the smooth areas. Using the simple LSB substitution method with an Optimal Pixel Adjustment Process (OPAP), the proposed technique provides a relatively high hiding capacity and relatively good quality stego-image. The secret data, embedded into the cover-image using the above rectified steganographic technique or data hiding technique, can be in any type of format. However, when the embedded data is a digital image, a data hiding technique is proposed to enhance the image quality of the resulting stego-image. In the second proposed technique, the secret-image is embedded into the cover-image by exploring the correlation between the cover-image block and the secret-image block. The best match image block of the secret-image block is selected based on the block difference from both of the cover-image and the secret-image. Using this data hiding technique, the image quality of the resulting stego-image can be greatly improved when embedding a smooth secret-image into a noisy cover-image.

Date of Award	15 Feb 2006
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Lee Ming CHENG (Supervisor)