Abstract: There is more to an image than what just meets the naked eye. Images are no longer just memories or records of the past. Image steganography has made sure of this. With the growing need to secure our valuable information and data, the concept of information hiding was born. And from the day of its invention, it has evolved significantly. It has evolved from cryptography to watermarking for the copyright protection to Steganography. This study describes how the covert communication takes place effectively by means of Pixel Indicator (PI) and Pixel Value Differencing (PVD) for multi-user. The former is employed in the color cover image to separate it into three planes and any one plane indicating the data channel through its last two bits of its intensity values to entrench and the later introduces the variable bit embedding in cover image. To increase the complexity, scrambling of secret data is introduced before embedding itself. Steganalysis results proved that this method is more resistive to chi-square attack. The proposed methods performance is appraised by manipulating MSE and PSNR and the results are tabulated.
INTRODUCTION
The dramatic increase in the multimedia and communications over the Internet protocol, the secretive transmission of data has become the want of the hour nowadays. A quick method for this purpose is the use of steganography (Amirtharajan et al., 2010, 2012; Wang and Wang, 2004; Hmood et al., 2010a, b; Zaidan et al., 2010) but the direct modification of the Least Significant Bit (Amirtharajan and Rayappan, 2012a-c; Chan and Cheng, 2004; Thanikaiselvan et al., 2011b) has resulted in quite some distortions in the image (Stefan and Fabin, 2000; Zanganeh and Ibrahim, 2011).
Apart from the general services offered by the internet, generation is getting more innovative in terms of transmission and is presently at remote desktop connections giving the users a complete Home-like experience when not in home. But the security of the Home control is lost! Hence the image or data hiding principles (Bender et al., 1996; Cheddad et al., 2010) appeared and these principles are classified based on the operation as Cryptography (Salem et al., 2011; Schneier, 2007) and Steganography (Janakiraman et al., 2012a, b; Provos and Honeyman, 2003; Rajagopalan et al., 2012). Cryptography being widely used in digital communications while the latter is best used for image, audio transmission purposes. One more classification is watermarking especially for authentication purposes (Abdulfetah et al., 2010; Stefan and Fabin, 2000; Zeki et al., 2011).
Steganography's significance was nothing other than the sky after the internet evolution (Stefan and Fabin, 2000; Zanganeh and Ibrahim, 2011). The process uses a carrier image for the blinding the message and sending it over any carrier like image (Cheddad et al., 2010; Luo et al., 2011; Mohammad et al., 2011; Wu et al., 2005; Zhao and Luo, 2012), audio (Zhu et al., 2011), video (Al-Frajat et al., 2010) and text (Al-Azawi and Fadhil, 2010; Shirali-Shahreza and Shirali-Shahreza, 2008; Xiang et al., 2011). Importantly the steganographic embedding process must have the three important characteristics to provide a platform with high capacity for heavy and mass data to be hidden and send across (Thanikaiselvan et al., 2011a), aesthetic look to ensure security against visual findings and resisting any other steganalysis (Qin et al., 2010; Amirtharajan and Rayappan, 2012d) or random randomizations of steganography being applied on the stego image (Thenmozhi et al., 2012; Luo et al., 2011).
Looking deeper into the area of interest, image steganography is sub-classified into two domains-Spatial and Transform (Amirtharajan and Rayappan, 2012d; Thanikaiselvan et al., 2011a). The former is the direct manipulation on the intensities (Padmaa et al., 2011) and the latter involves various transforms (Amirtharajan and Rayappan, 2012d; EI-Safy et al., 2009; Provos and Honeyman, 2003; Thanikaiselvan et al., 2011a) like DCT, DWT, IWT etc.
A combination of the two domains is also possible, while the hot cake now being the adaptive technique (Gutub, 2010) which is analogous to an adaptive array of antennas where they steer based on the direction of the signal but here the method is decided based on the image type and the other parameters (Padmaa et al., 2011). Hence the adaptive technique can also be called smart technique leads the way to a clueless means of embedding which have proven to be more efficient in all terms, from the appeal of the cover to the maximum amount of hiding of the message (Amirtharajan and Rayappan, 2012a, b, d).
The interesting part here is that these schemes do not in any form spoil the natural appearance of the image since the area of embedding and capacity is changed dynamically depending on the cover image and payload parameters (Padmaa et al., 2011; Zhao and Luo, 2012). This approach is the first in the category.
The central aim of this study is to share the secret in the image confidentially and securely. For this, the secret from four different users are encrypted by four different methods using four different keys. Then, the encrypted results undergo conventional but with twist, embedding procedure to form a stego image. This indeed is a complex routine, because, to get the secret back, one should have the knowledge about encryption methods employed and their corresponding keys. Also, unless and until, one knows the plot of embedding variable bits, the retrieval process becomes a nightmare. Thus, this model assures all the parameters of a good steganographic scheme are well convinced with likelihood to put into action.
PROPOSED METHOD
Steganography is used to protect the information from illegal parties. The block diagram for this proposed method is represented in Fig. 1. In embedding block, first encrypt the secret data from four users using keys preferably Advanced Encryption Standard (AES) or of users choice may be any public key cryptography (Schneier, 2007).
| Fig. 1: | Block diagram for multiuser tri-color random image steganography |
Then embed it in cover image using Pixel Indicator (Gutub, 2010; Janakiraman et al., 2012a) and PVD (Padmaa et al., 2011; Wu et al., 2005). In recovery block, first recover the data and then decrypt it using keys which are used in encryption.
Three methods are discussed in this study. In method 1, red plane is the indicator channel forever and remaining two planes (i.e., blue and green planes) are data channels. In method 2, users have to select their indicator channel. In method 3, all the planes are considered as indicator channel in cyclic manner. In first two methods, data should be embedded in data channels but not in indicator channel. In third method, data should be embedded in data channels as well as in indicator channel. So Method3 gives good embedding capacity compared to the previous two methods. Flow chart for embedding and recovery are represented in Fig. 2 and 3.
| Method 1: Multi user PVD with tri-color random image steganography |
| Recovery algorithm: |
Method 2: Multi user PVD with custom-indicator-plane tri-color random image steganography: Method 2 is same as that of Method1, except that user defined Indicator plane is defined here.
| Algorithm |
| Fig. 2: | Flowchart for Embedding |
| Recovery algorithm |
RESULTS AND DISCUSSION
In this presentation, Lena, Baboon, Mahatma Gandhi and Temple are chosen as cover images of size 256x256x3 and its stego images and histograms are shown in Fig. 4-7 for Method 1 and in Fig. 8-11 for Method 3.This implementation has been simulated in MATLAB 7.1 and its MSE and PSNR values are tabulated in Table 1-3 for all the three methods.
Method 1: The proposed stego method effectiveness can be estimated by calculating MSE and PSNR for stego image Si,j and cover image Oi,j . The mathematical expressions are given here:
For method1 and method3, the stego images and its histograms are shown in Fig. 4-11. The result shows that, there is no visual distortion in those images. The histograms of cover and Stego images are nearly identical, so it is difficult to hit the data. This grants better image quality even after embedding data of all the users.
| Fig. 3: | Flowchart for Extraction |
| Fig. 4(a-c): | Method 1, Lena (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 5(a-c): | Method 1, Baboon (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 6(a-c): | Method 1, Mahatma Gandhi (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 7(a-c): | Method 1, Temple (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 8(a-c): | Method 3, Lena (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 9(a-c): | Method 3, Baboon (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 10(a-c): | Method 3, Mahatma Gandhi (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Fig. 11(a-c): | Method 3, Temple (a) Cover images (b) Stego images and (c) Corresponding histograms of cover and stego images before and after hidding data |
| Table 1: | MSE, PSNR values for method 1 |
| Table 2: | MSE, PSNR values for method 2 |
| Table 3: | Comparative results of MSE, PSNR values for method 3 with Padmaa et al. (2011) |
From the Table 1-3, its observed that, baboon color image of size 256x256x3 gives better embedding capacity compared to other three images in all the three methods and the proposed method offers better imperceptibility and also it varies from cover to cover. Mahatma Gandhi image provides good PSNR in all the three methods, which gives better imperceptibility along with good image quality.
Security analysis: To randomize each user secret data, Advanced Encryption Standard (AES) is introduced here, it offers 2^128 complexity in the system. Assuming 25% probability on each cases say 00,01,10,11, which decides the embedding capacity. The total complexity for four users will be 2^128*0.5*2*4.
Chi-square attack is a Steganalysis technique. This is based on probability of embedding data in an image. Figure 12 presents the graphical view of chi-square attack in Mahatma Gandhi image.
| Fig. 12: | Graphical representation of Mahatma Gandhi against chi-square attack |
The graph is plotted between probability of embedding and number of rows in an image. It clearly proves that, the cover and stego are nearly alike, so this proposed method survives against chi-square attack.
CONCLUSION
Today innovative thinking sounds good, so this presentation gives the creative idea by using PVD, PI and OPAP. Pixel value differencing process gives smart embedding in all the planes provides visually undistorted image. Pixel Indicator for color cover image offers increase in embedding capacity by the way of embedding secret data in indicator plane. To make the stego image indistinguishable with cover image, of course OPAP afford this. The power of the algorithm can be tested against chi-square attack, which gives the awesome result. So this proposed method stand unique compared with all other existing methods.