In this study, a secure mode of steganography is projected in which three main entities are exercised. They are Pseudo random generator, Least Significant Bit (LSB) substitution and Optimal Pixel Adjustment Process (OPAP). A popular cliché All roads lead to Rome says that there endure countless ways to attain a target. Likewise, this goes very well with Steganography than any other things. Digital epoch of steganography embraces numerous trials to surmount the universal hitch of concealment and refuge in every form of computer communication. This study reveals one more methodology for the same but with unique characteristics. Here, cyclic pixel indicator technique is used wherein two channels are used as data channels and the remaining channel is used as an indicator channel. Red plane is taken as the indicator channel for the first pixel for the subsequent pixels the indicator channels follows a periodic cycle of R, G, B. barring the indicator channel, the other two channels act as the data channels for the corresponding pixels. The pixel intensity determines the bits to be embedded. i.e., LSBs of the indicator channel. If the LSBs of indicator channel say R channel are 00 Embed 1 bit in G and 2 bits in B; in case of 01, embed 2 bits in G and B each. If LSB(R) is 10, 2 are inlayed in G and 3 bits in B. Finally id LSB(R) equals 11 and then 3 bits are ingrained in G and 3 bits in B. A novel 2-key based pseudo random generator is employed which is used to embed data completely in a unique random fashion based on users choice. Thus, it introduces obscurity, while composite embedding of secret bits will result in additional intricacy. Therefore, this study is yet another practical cum unassailable means for secret sharing. Justification for this study is provided by the analytical results.
PDF Abstract XML References Citation
How to cite this article
Information communication through internet is in diverse forms and rivets variety of applications. This, indeed, requires various levels of concealment based on the applications demand. The electronic transmission faces numerous problems like sham, copyright control, data security, etc. So, the desired information needs to be transmitted safely, first and should endure the attacks of intruders (Salem et al., 2011; Schneier, 2007; Stefan and Fabin, 2000). The refuge of such data transfer has been a chief apprehension and has become the study in limelight, probably, forever. The most universal means of doing so is to send the desired data in disguise so that the beneficiary only can avail it (Amirtharajan and Rayappan, 2012a-d). True, this led to the development of information hiding techniques and secret communication methods (Bender et al., 1996). The prime bases for the so called methods are cryptography and steganography (Hmood et al., 2010a, b; Rajagopalan et al., 2012) and its counter attack called cryptanalysis and steganalysis (Qin et al., 2010). While the former (cryptography) deals with text or number alone, the latter (steganography) brings all digital media into play (Cheddad et al., 2010).
The place where mathematics and engineering meets is be called cryptography. It is one of the means by which a readable data or text is made unreadable (for the eavesdropper) by encrypting the readable data. A sender transforms an original text (plaintext) into a modified text (cipher text) by means of a cryptographic key using encryption (Schneier, 2007). The receiver performs the reverse operation to retrieve back the original message by decryption. So, a interloper cannot tamper the concealed information. The remarkable cryptosystem services are Confidentiality, legitimacy, Access Control, veracity and Non-repudiation (Schneier, 2007). A good encryption method should thrive for two fundamental attributes, viz., confusion and diffusion (Salem et al., 2011). Cryptanalysis is nothing but the attack performed on cipher text. Putting it simple it is the way to crack a cryptosystem. The attacks aim at acquiring key and they may be active or passive. Some of the known cryptographic attacks are Ciphertext Alone attack, Known Plaintext attack, Chosen Plaintext attack (Schneier, 2007).
Steganography exists from long past; some means in history were micro dots (Provos and Honeyman, 2003), invisible inks, shaved head tattoos, wax tablets etc (Kahn, 1983; Stefan and Fabin, 2000). In this day and age, digital images are brought into play as cover files devoid of any suspicion for the reason that their charisma is ubiquitously on the Internet (Cheddad et al., 2010; Hmood et al., 2010a, b; Gutub, 2010; Amirtharajan et al., 2012; Janakiraman et al., 2012a, b; Padmaa et al., 2011; Zaidan et al., 2010). Furthermore, designing steganography algorithms that are statistically unnoticeable and acquiescing a large capacity is the core aspiration of Steganography (Thenmozhi et al., 2012). Also, a steganography system is ideally safe if the statistics of the cover image and the stego image are one and the same (Thanikaiselvan et al., 2011a, b). For that reason, the embedding capacity is liable to be superior to the robustness (Amirtharajan et al., 2012). The former is nothing but the size pertaining to both cover files size and one which contains secreted message.
As a result, countless novel embedding practices have been recommended so as to boost the imperceptibility and swell the capacity of steganography methods. Steganography carrier archives may be of the extension txt, mp3, gif, bmp, wav, jpeg etc. (Al-Azawi and Fadhil, 2010; Al-Frajat et al., 2010; Xiang et al., 2011; Zanganeh and Ibrahim, 2011; Zhu et al., 2011; Zhao and Luo, 2012). Of these jpeg is the universal image format for local convention and internet given that it affords large compression ratio and retain high image quality. Therefore, JPEG compressed images are the mainly apposite cover images used for steganography. Steganography in spatial domain (Luo et al., 2011; Mohammad et al., 2011) includes Least Significant Bit (LSB) (Chan and Cheng, 2004), Pixel Value Differencing (PVD) and Pixel Indicator (PI) (Janakiraman et al., 2012a; Padmaa et al., 2011) whereas frequency domain includes Discrete Cosine Transform DCT (Abdulfetah et al., 2010), Discrete Wavelet Transform DWT (Abdulfetah et al., 2010) and Integer Wavelet Transform IWT (Thanikaiselvan et al., 2011a; Amirtharajan and Rayappan, 2012d).
If we roll back, watermarking concept was used around 700 years ago in Italy in the Fabriano town. Watermarking is all about establishing distinctiveness of information to avert illicit use (Stefan and Fabin, 2000). The noteworthy aspects of watermark are that they are supposed to be undividable from the carrier, indiscernible and should stay entrenched even through transformation (Abdulfetah et al., 2010; Zeki et al., 2011; Zhang et al., 2006). Unlike Steganography, Watermarking has the additional perception of buoyancy against endeavours to confiscate the veiled data (Cheddad et al., 2010).
The common terminologies employed in watermarking are visible and fragile watermarks, labelling and fingerprinting, embedded signatures, bit stream watermarking. Based on the keys, the watermarking techniques are categorized as secret and public. The general properties of robust watermarking schemes are redundancy, keys and imperceptibility as they are applicable to formatted text, image, audio, video 3D models etc. Popular applications of watermarking are to give ownership proof, data tracking, monitoring, fingerprinting etc. (Stefan and Fabin, 2000). Reviewing the available literature suggest to implement a random image steganography method to offer good payload without compromising imperceptibility.
The main goal of this study is to offer a substantive steganographic plot for communication of surreptitious messages through images. It entices most common traits of digital image steganography but constructs the plot very in a different and complex way so as to make steganalysis outrageous. Thus, this is one such method which is easy to realize but hard to attack.
This study asserts a diverse pixel indicator technique for both variable capacity and randomization in the cover. Here, Cyclic pixel indicator technique is used wherein two channels were used as data channels and the remaining channel is used as an indicator channel. Red plane was taken as the indicator channel for the first pixel for the subsequent pixels the indicator channels follows a periodic cycle of R,G,B. barring the indicator channel, the other two channels act as the data channels for the corresponding pixels.
The bits needed for embedding is decided by the pixel intensity i.e., LSBs of the indicator channel. If the LSBs of indicator channel, say R channel were 00 Embed k bit in G and k+1 bits in B. Similarly, if LSB(R) is 01, then k+1 bits are embedded in G and k+1 bit in B. If LSB(R) is 10, k+1 bit were embedded in G and k+2 bits in B. Finally id LSB(R) equals 11, then k+2 bits were embedded in G and k+2 bits in B. A novel 2-key based pseudo random generator is employed which was used to embed data completely in a unique random fashion based on users choice. The proposed methodologys block diagram is given in Fig. 1.
The conditions for embedding in data channels as per the two LSBS of indicator channel are given in Table 1.
The processes can be laid down to three block process, apart from the Optimal Pixel Adjustment process incorporated while embedding the data.
|Table 1:||No. of bits embedded|
|Fig. 1:||Block diagram of proposed method|
|Fig. 2:||Flow chart for pseudo random number generator|
PSEUDO RANDOM GENERATOR
The sequence employs generation of random sequences which act as the index numbers creating a space filling curve. The curves routing path is determined by a combination of user defined keys. Fig. 2 represents the flowchart for Pseudo Random Number Generator (PRNG) and Fig. 3 and 4 represent the flowchart for embedding and extraction.
|Algorithm for PRNG:|
|Fig. 3:||Flowchart for embedding|
|Fig. 4:||Flowchart for extraction|
|This process aims at embedding the secret data as follows.|
|This algorithm retrieves the secret data from the image efficiently.|
RESULTS AND DISCUSSION
The results had excelled as far imperceptibility and randomness was concerned. We considered images Temple, Baboon and Mentor of 256x256x3 pixels color image as cover image and its stego image. To prove the efficiency of the algorithm under any cover, 2 practical images had been used. Cover image of Temple and its stego output with histograms of the image are given in Fig. 5. Graphical results for comparison of MSE and embedding capacity of Temple image is given in Fig. 6. Similarly, the cover and stego images of Baboon and Mentor along with the histograms are shown in Fig. 7 and 8. Graphical results for comparison of MSE and embedding capacity of Mentor image is given in Fig. 9. An example of randomized image is depicted in Fig. 10. MSE and PSNR values of the images Temple, Baboon and Mentor are given in Table 2, 3 and 4, respectively.
|Fig. 5(a-b):||(a) Cover and Stego images of temple (b) Histogram of temple image|
|Table 2:||MSE and PSNR values for temple|
|Fig. 6:||MSE vs. number of bits for the temple cover image|
|Fig. 7(a-b):||(a) Cover and Stego images of Baboon (b) Histogram of Baboon image before and after hiding data|
|Table 3:||MSE and PSNR values for baboon|
|Fig. 8(a-b):||(a) Cover and Stego images of Mentor (b) Histogram of Mentor image before and after hiding data|
|Fig. 9:||MSE vs. Number of bits for the Mentor cover image|
Also, the pseudo random method with the two key systems has shown very good strength with high randomness visible in Fig. 10. Since this routine has PSNR of 48dB and above it exhibits more than enough imperceptibility and capacity.
|Fig. 10:||Sample of randomized image for team.jpeg|
|Table 4:||MSE and PSNR values for mentor.jpeg|
Inclusion of the randomization algorithms in no way have raised suspicion or compromised the imperceptibility of the image, setting up new benchmarks for stego system that aim to supersede the present algorithm. The results have also affirmed our conclusions. The one way pseudo random algorithm makes it difficult for any system to back track the entire algorithm even if routing pattern for any key is discovered.
LIMITATIONS AND FUTURE ENHANCEMENTS
The work has presented itself to be adept not just in imperceptibility but also at randomness. However, this is due to customization of the algorithm that is fixed to a particular dimension. The application of the algorithm to a generic image dimension is a daunting task. This is owing to the fact that the Strength of the pseudo random algorithm is dependent on prime numbers which make it foolproof. Building a generic algorithm requires higher computing specifications and takes up a lot of executing time.
The Future benchmarks that could outperform the algorithm could aim to accomplish a certain properties. Constructing an algorithm that could route data to suppress the Mean Square Error would not only do the job of rerouting the data in a random fashion but also act as complex stego tool. However, proposing such an algorithm could compromise over the security of the one way function against backtracking, making it a reversible function. Discovery of newer high capacity techniques could replace the model and when merged with the present algorithm could give even better results.
- Provos, N. and P. Honeyman, 2003. Hide and seek: An introduction to steganography. IEEE Secur. Privacy, 1: 32-44.
- Thanikaiselvan, V., S. Kumar, N. Neelima and R. Amirtharajan, 2011. Data battle on the digital field between horse cavalry and interlopers. J. Theor. Applied Inform. Technol., 29: 85-91.
- Zeki, A.M., A.A. Manaf and S.S. Mahmod, 2011. High watermarking capacity based on spatial domain technique. Inform. Technol. J., 10: 1367-1373.