A Fast and Accurate Steganography Technique

Abstract

Communication technology has been advanced in the last few years, which increases the requirements of secure data communication and information management. For this reason many researchers have exerted much of their time and efforts in an attempt to find suitable ways for hiding important information. The proposed technique presents an efficient storage security mechanism for the protection of digital medical images by improving the previous Discrete Wavelet Transform (DWT) techniques. The quality of the stego image and the recovered image showed acceptable visual quality. The effectiveness of the proposed scheme will be proven through the well-known of imperceptibility measure of Weighted Peak Signal-to-Noise Ratio (WPSNR), Mean Square Error (MSE) and Normalized Cross Correlation Coefficient (NCCC). Experimental results have been compared to the previous techniques. Working environment for proposed system is MATLAB.

Introduction

The copyright protection of the digital multimedia like image, audio and video is very required with the network and communication development. In today’s world the cloud environment became very popular and important for many people as a platform for storing and retrieving huge data. So as to use it must be secured. Steganography is a way to maintain the confidentiality of the information being transmitted. Many methods like cryptography, watermarking, fingerprinting and encryption and decryption techniques were advanced in order to make the information secured during communication. Cryptography computation and key management limit its employment and therefore the recovered image has a poor quality. Watermarking and fingerprinting are mainly concentrating on preserving the copyright property and have different algorithms. In both watermarking and fingerprinting the fact that the secret message hidden inside the carrier may be visible. But in steganography the fact that there is a secret message hidden inside a file itself will be a secret.

Most commonly used file format for communication is Digital Image due its high frequency on the Internet. The image resulted after hiding process can be called as stego image.

Steganography is categorized into two types: Spatial domain and Transform domain. In spatial domain techniques the secret data is hidden in the intensity of the pixels directly. The most popular widely used and simplest steganographic technique in the data hiding is least-significant-bit (LSB) substitution. The process of embedding data in the transform domain (frequency domain) is much stronger than embedding techniques that operate in the time domain. Today most of the strong steganographic systems operate in the transform domain. Various applications have different requirements of the steganography method used. For example, some applications may require absolute invisibility of the secret information, while other applications require a larger secret message to be hidden. Discrete Wavelet Transformation (DWT) will be presented in details and improvements will be applied to give a proposed algorithm to solve the problem of the long processing time required for the existing DWT techniques.

The paper presented in five sections. Section I gives the introduction of this paper. Section II discusses the problem analysis. Section III explains the proposed method. Section IV shows the implementation and results. Finally section V gives the conclusion of this paper.

Problem Analysis

In image based steganography, it is required for the steganography technique to be able to hide more secret message bits as possible in an image in a manner that will not affect the two important requirements that are essential for the success of the hiding process:

Security/Imperceptibility: which means that human eye cannot distinguish between the original image and the stego image.

Capacity: which means the amount of secret data that can be embedded in a cover media.

The relationship between the two requirements should be balanced, in other words if we increase the capacity more than a certain limit then the imperceptibility will be affected and so on, therefore the digital steganography parameters should be chosen very carefully.

Both the DCT and the DWT methods come under transform domain analysis and are the most common nowadays. Both the methods have good imperceptibility and also Robustness against statistical attacks. But as we know the major aim of the steganography is to increase the robustness against attacks and also to increase the payload capacity. In the case of capacity and processing time, DWT is good compared to DCT.

The DWT is very suitable to identify the two important regions in the images, Region of Interest (ROI) and Region of Non-Interest (RONI). The region that contains most important information/data, essentially the diagnostic part in case of radiographic images, is called the ROI which has most of the image energy and called lower frequency sub-band (LL). So embedding the secret message in (LL) sub- bands may degrade the image quality significantly. While, the high frequency sub-bands (HH), (LH) and (HL) includes the edges and textures of the image that classified as (RONI) at which a secret message can be embedded effectively and the human eye is less sensitive to changes in such sub-bands. This allows the secret message to be embedded without being perceived by the human eye so most of the steganographic methods use RONI for data embedding.

Another issue is that during the embedding process of the secret image into the cover image the embedding and extraction process time differs with the secret image size. The time increases with the size. This problem has been solved through the proposed algorithm.

Proposed Method

The proposed technique depends on applying the DWT on four stages instead of implementing it at once. This economizes the processing time whereas the total embedding time is less than the embedding time of the one stage embedding process.

The proposed system of the embedding algorithm has the following five modules.

• Splitting
• Applying DWT
• Embedding process
• Applying IDWT
• Merging

At module 1 the splitting process splits both of the cover and the secret images into four equal size sub-images each of them has a quarter of the original image size. At module 2 DWT is applied on each of the cover and secret sub-images (eight sub-images). At module 3 each of the transformed secret sub-images is embedded into the cover transformed sub-images and it results the stego sub-images (four sub-images). At module 4 IDWT is applied on each of the stego sub-images. At module 5 the final stego image is constructed by merging the four sub-images into one image.

By the same manner the extraction algorithm has the same sequence of the embedding algorithm except at module 3 the extraction process is applied. At module 1 the splitting process splits the stego image into four equal size sub-images each of them has a quarter of the stego image size. At module 2 DWT is applied on each of the stego image sub-images (four sub-images). At module 3 the extraction process is applied to the stego image sub-images to extract the secret sub-images. At module 4 IDWT is applied on each of the secret sub-images. At module 5 the final secret image is constructed by merging the four sub-images into one image.

The performance of the proposed algorithm was evaluated by three benchmark techniques. The Mean Square Error (MSE) and the Weighted Peak Signal to Noise Ratio (WPSNR) which are used to measure the distortion between the original cover image and the stego image after embedding the secret information in the cover. And also by the normalized cross correlation coefficient measure (NCCC) which is used to measure the similarity between the cover and the stego image.

Mean Square Error (MSE)

MSE= 1/(M*N) ∑_(J=1)^(M-1)▒∑_(K=1)^(N-1)▒〖(X_(J,K)-X_(J,K)^')〗^2 (1)

Whereas X_(J,K) is the cover image that contains M*N pixels and X_(J,K)^' is the stego image.

Normalized Cross Correlation Coefficient (NCCC) is given by (2).

NCCC= ∑_(J=1)^M▒〖∑_(K=1)^N▒〖(X_(J,K ) . X_(J,K)^' 〗)1/(∑_(J=1)^M▒∑_(K=1)^N▒〖〖(X〗_(J,K))〗^2 )〗 (2)

NCCC can measure similarity up to some amount. The larger NCCC value between the cover and the stego image indicates the higher performance of the technique.

Weighted Peak Signal to Noise Ratio (WPSNR) is given by (3).

WPSNR=10log_10 〖(L_max/(RMSE ×NVF))〗^2 (3)

, NVF(i,j)=1/(1+θσ_x^2 (i,j))

The WPSNR uses an additional parameter called the Noise Visibility Function (NVF) which is a texture masking function. The WPSNR uses the value of NVF as a penalization factor. Where σ_x^2 (i,j) denotes the local variance of the image in a window centered on the pixel with coordinates (i, j) and θ is a tuning parameter corresponding to the particular image. For flat regions, the NVF is close to 1. While for edge or textured regions NVF is more close to 0. This means that for smooth image, WPSNR approximately equals to PSNR. But for textured image, WPSNR is a little bit higher than PSNR.

Conclusion

A novel steganography algorithm has been proposed and a comparative study presented with the existing techniques. The proposed method produces no statistical or visual changes in the images, fast, and secure system for the telemedicine applications. The algorithm is vigorous against to any attacks. Implementation results showed that the proposed method gives high quality images and requires much less processing time than previous techniques. The algorithm is appropriate for every sort of images. Further the system can be applied to different types of files like Audio, Video etc.