Internet created a revolution in the cycle of life and has made the world a very small place with its thrilling speeds. As technology evolves, Man gets the information to his pocket on the GO. But acts of rivalry are affecting man greatly, even significantly in the case of technology. Data can be brought to the human beings pocket but wondering about the security, boggles the mind to square ONE. As a matter of fact, Man can't guarantee his own security then how could DATA! So the Data that reaches to our feet has a threat from intruders which the internet has failed drastically despite its up-gradations. Privacy of data is prioritized and a personnel or technology maintaining the privacy information is now on the deck today.
One such Upper Deck technology is Information Hiding/Steganography (Stefan
and Fabin, 2000; Petitcolas et al., 1999;
Rabah, 2004; Cheddad et al.,
2010), where in all the needs of man are more than fulfilled with no intruders
poaching on the data. The encrypted data by information hiding is secure, faster,
accurate and zooming ahead into the peoples lives, empowering them with
unmatched speeds and latest generations.
Cryptography (Schneier, 2007) is the ancient art of writing
messages by scrambling the secret data into meaningless junks and presenting
them in a manner unrecognizable to eavesdroppers and not for cryptography participants.
Steganography involves hiding the data to be transmitted in cover object which
could be audio, video, image, text (Al-Frajat et al.,
2010; Al-Azawi and Fadhil, 2010; Amirtharajan
et al., 2010d; Bender et al., 1996,
2000; Hmood et al., 2010a;
Shirali-Shahreza and Shirali-Shahreza, 2008) and sometimes
may be the transmission itself is not noticeable to the eavesdroppers. How to
combine Cryptography and Steganography effectively to offer better data security
and their differences are detailed in (Zaidan et al.,
2010) the dexterity with which data can be hidden by this technique is unparalleled
and is the most efficacious one present.
In addition to the requirement of security, equally pressing demands of the
steganography system are imperceptibility and high data capacity known as payload
(Cheddad et al., 2010; Hmood
et al., 2010b). In general, a system is considered to be suitable
for information hiding when it satisfies the three requirements of information
hiding magic triangle, i.e., imperceptibility, robustness and high capacity
(Stefan and Fabin, 2000; Petitcolas
et al., 1999; Rabah, 2004; Cheddad
et al., 2010). The other side of steganography is called Steganalysis
which is an art of revealing the presence of the hidden data (Fridrich
et al., 2001; Wang and Wang, 2004; Qin
et al., 2010). Furthermore, steganography can be classified into
spatial domain and transform domain steganography.
In the spatial domain, data is hidden in the system by direct manipulation
of the pixels, the most common technique being normal Least Significant Bit
(LSB) substitution method through raster scan (Chan and
Cheng, 2004; Thien and Lin, 2003; Wang
et al., 2001; Zanganeh and Ibrahim, 2011)
or random traversing path (Amirtharajan and Balaguru, 2009,
2010; Provos and Honeyman, 2003;
Luo et al., 2008). In the transform domain technique,
the image is first transformed using techniques like Discrete Cosine Transform
(DCT) (Provos and Honeyman, 2003)or discrete wavelet
transform (Thanikaiselvan et al., 2011) and then
the coefficients are exploited for the purpose of data concealment.
Gutub et al. (2008) and Gutub
(2010) pioneered the Pixel Indicator Technology (PIT) (Upreti
et al., 2010) based on random color image steganography where the
last two bits of the indicator plane would decide whether the remaining data
channels are useful for embedment.
Amirtharajan et al. (2010a, b,
c) have also proposed few variants using this technology.
In all their aforementioned algorithms, the authors have used pixel indicator
as a means to suitably find planes to embed the secret data, whose length is
determined by excess 3 values of indicating pixel.
In another technique, from the same author pixel value differencing (Amirtharajan
et al., 2010c; Park et al., 2005;
Wu et al., 2005) decides the number of bits for
embedment, along with pixel indicator. Another variant of the aforesaid method
was employed by Amirtharajan et al. (2010a)
where a new factor E is introduced which provides the user a flexibility to
embed the secret data from a particular position in a pixel, thereby boosting
the robustness of the system itself.
Padmaa et al. (2011) proposed a multiuser Pixel
indicator method, where the 2nd number of users shares the same image
to transfer the secret information in a single cover image. Kumar
et al. (2011) proposed multiuser steganography in Orthogonal Frequency
Division Multiplexing (OFDM) symbols which would be useful to embed additional
information during high data rate transmission in the presence of Additive White
Gaussian Noise (AWGN) channel. Furthermore Code Division Multiple Access (CDMA)
based multiuser Steganographic embedding elucidate by Amirtharajan
et al. (2011). In all the aforesaid multiuser stego methods there
is a possibility of sharing a common cover object to transfer the secret.
In this study, three methods have been suggested with an aim of achieving high capacity, imperceptibility or both. These proposed methods are implemented in SCIlab for random Image steganography and the results are encouraging.
LSB substitution method revisited: The most well-known steganographic
technique in the data hiding is Least Significant Bits (LSBs) substitution.
This method embeds the fixed-length secret bits in the same fixed-length LSBs
of pixels (Chan and Cheng, 2004; Thien
and Lin, 2003).
Let C be the original cover object (Image) of MxN pixels and m be the secret data (message or image).
The stego object (Image) S could be computed as follows:
Even though this LSB method is simple and easy to implement but it generally
causes perceptible alteration in the stego Image, if the number of embedded
bits for each pixel exceeds three. The other methods like adaptive LSB vary
the number of embedded bits in each pixel. These methods possess better image
quality in comparison with simple LSBs substitution with reduction in the embedding
capacity or payload.
Optimum pixel adjustment procedure (OPAP): Chan
and Cheng, (2004) elucidate Optimal Pixel Adjustment Procedure (OPAP) which
reduces the deviation caused by the LSB substitution method. In this process,
the payload is embedded in a way such that the overall quality of the stego
output is good and it would not affect the hidden data
Procedure for data hiding through OPAP: Initially few least significant bits are embedded with the data.
Next step is to reduce the error, it would be carried out by properly adjusting all the remaining k+1th bits.
Let n be the number of LSB substituted in each pixel and d be the decimal equivalent of the pixel after embedding.
d1 is the decimal value of last n bits of the pixel and d2 be the decimal value of n bits hidden in that pixel. Then compute the following:
This adjustment would considerably reduce the error to 2k-1 instead
PIXEL INDICATOR TECHNIQUE
Pixel indicator technique, as the name implies uses the pixel value as an indicator to see whether a secret data bit is embedded or not in that pixel. As the pixel value determines the embedding process it is not sequential, thus nonlinear nature of embedding bits will improve security. In the gray scale the pixel value ranges from 0 to 255. The binary represented bits in locations specified will determine the bits to be embedded. PIT technique is followed in color image with great effect.
PIT considering same pixel as indicator and channel: This pixel indicator technique considers a single pixel of cover image as indicator and the secret information is embedded in that pixel itself. Many consideration could be possible for this pixel indicator technique, the methodology adapted in this PIT by considering the 3rd, 2nd bit of the pixel as indicator and 1st, 0th bit as channel for secret information hiding. The tabulation will depict the methodology. In this PIT methodology the maximum number of bits that could be embedded per pixel is two and the minimum is 0 bit per pixel.
Let Pi = b7 b6 b5 b4 b3 b2 b1 b0
Here, b3 and b2 acts as indicator pixels and b1 and b0 are the channels for secret messages.
The adapted methodology is given in Table 1 and the embedding and extraction flowchart is given in Fig. 1 and 2.
PIT considering three pixels as a block (Method 2): In PIT technique considering 3 pixels as a block, the entire image is divided into blocks of three pixels each. The first pixel value acts as indicator while the successive second and third pixel acts as channel for secret information to be hidden. The corresponding embedding and extraction flowchart are detailed in Fig. 3 and 4.
Let Pi, Pi+1 and Pi+2 forms a block of three pixel values. Let us name those bits of these 3 pixel in a block as follows:
||b23 b22 b21 b20
b19 b18 b17 b16
||b15 b14 b13 b12 b11
b10 b9 b8
||b7 b6 b5 b4 b3
b2 b1 b0
Here, in this PIT of gray image steganography, the b17 and b16
bits acts as indicator while the following bits b10 , b9
, b8 , b2 , b1 and b0 acts
as a data channel. The adapted methodology is given in Table 2.
Hence, in this case of PIT with three pixels as a block the capacity of this proposed method increased considerably. The minimum number of bits embedded in that block is four and a maximum of six bits per block is achieved. This is comparatively higher than method 1 which considers one pixel at a time. As this case of PIT considers embedding of bits more than 1, there is a chance of applying OPAP in order to reduce the error metrics values.
Pit and channel selection criteria (Method 3): In this PIT technique,
the indicator and channel selections are decided during the time of embedding.
||Function of the proposed pixel indicator choices in gray cover
object (method 1)
|| Function of the proposed 3 pixel block embedding (method
|| Flow chart for Embedding (PIT considering same pixel as indicator
|| Flow chart for Retrieval (PIT considering same pixel as indicator
|| Flowchart for embedding (PIT considering 3pixels as a block)
|| Flowchart for retrieval (PIT considering 3pixels as a block)
The secret information size or length decides the indicators and channels for
itself. Thus the indicator and channels are not decided by the authorized person
but by the nature of secret data. Hence, the idea about the channel and indicators
is not given to anyone. As the secret information size decides the indicators
and channels, this should also be hidden and not revealed to others. The hiding
of the secret information size happens to be the LSB 2 bit insertion at the
starting of the cover image pixel.
The length of the secret message is embedded in the 2 LSB bit positions of Green and Blue channels of first four pixels. In this present method, It is assumed that 16 bits space is enough to specify the size of the secret information. Then leaving those first four pixels, embedding the secret message is carried on the data channel based on the selected indicators bit values. The adapted methodology is given in Table 3 and the work flow has been detailed in Fig. 5 and 6.
Thus, first level of selection considers the numerical nature of the length of secret information (N). Based on N to be even, prime or other, the indicator is selected from the three color combinations. The binary nature of the secret information decides the second level selection of channel order from the remaining pair.
The selection of indicator and channel alone varies here, while the embedding methodology remains the same as the first technique of PIT described earlier as of Table 1. If the indicator LSB bit value is one, then corresponding channel will carry 2 bits of secret information. Thus, in this case also, maximum of 4 bits per pixel and a minimum of 0 bit can be achieved. Capacity for this case is not increased but the security level is better compared to fixed indicator PIT.
RESULTS AND DISCUSSION
For all the three methods, Baboon or Parrot 256x256 pixel gray and color image has been taken as cover images and city top view of size 91x91, 117x117 and 129x129 pixels are considered as secret images. To evaluate the performance of the proposed system MSE, PSNR and BER have been computed for all the three methods.
Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE): The PSNR is calculated using the equation:
where, Imax is the intensity value of each pixel which is equal to 255 for 8 bit gray scale images.
The MSE is calculated by using the Eq. 2 given below:
where, M and N denote the total number of pixels in the horizontal and the vertical dimensions of the image Xi,j represents the pixels in the original cover image and Yi,j represents the pixels of the stego-image.
Bit Error Rate (BER) and bit error: BER evaluates the actual number of bit positions which are replaced in the stego image in comparison with cover image. It has to be computed to estimate exactly how many bits of the original cover image (Ic) are being affected by stego process. The BER for the Stego image (Is) is the percentage of bits that have errors relative to the total number of bits considered in Ic.
Let Icbin and Isbin are the binary representations of the cover image and stego cover then, the total number of bit errors:
and the bit error rate:
Tn is the total number of bits considered for the gray image of size MxN pixels. Tn will be MxNx8.
|| Functions of PIT and channel selection criteria (method 3)
|| Flowchart for embedding (PIT with indicator and channel selection
|| Flowchart for retrieval (PIT with indicator and channel selection
||Baboon cover image and city secret Image
||Baboon stego Image and recovered city secret Image
||Baboon cover Image and city secret Image
||Baboon Stego Image and Extracted city secret Image through
||Baboon stego image with OPAP and extracted secret image through
|| Error metrics of the proposed method 1
|| Error metrics of the proposed method 2 with payload
Results of PIT considering 1 pixel at a time: To evaluate the performance
of the proposed method 1, Baboon cover image of size 256x256 pixels gray image
and City secret gray image of size 91x91 pixels has been considered and the
resultant stego image and extracted secret image are shown in Fig.
7a and b, respectively. The proposed method tested for
full embedding capacity (Expected capacity is 12.5%, since k = 1 bit embedding)
and the results are tabulated in Table 4. The obtained PSNR
value is 47 dB with BER 0.087. It is also observed that there is slight improvement
in the capacity with compromised reduction in PSNR due to random embedding.
Results of PIT considering 3 pixels as a block: To evaluate the performance
of the proposed method 2, Same Baboon cover image of size 256x256 pixels gray
image and City secret gray image of size 91x91 pixels has been considered and
the resultant stego image, modified stego image with OPAP and extracted secret
image are shown in Fig. 8a-c, respectively.
The proposed method tested for fully embedding capacity and the results are
tabulated in Table 5. The obtained capacity is 20.8%, PSNR
value is 41.3 dB without OPAP and there is a huge improvement in PSNR 44.46
PIT with indicator and channel selection criteria: Parrots color cover image
of size 256x256x3 pixels and secret gray image of size 129x129 pixels has been
considered to evaluate the performance of the proposed method 3.
||Parrots cover image and secret image
||Parrots stego image and extracted secret image through simple
||Parrot Stego Image with OPAP and Extracted Secret Image through
|| Error metrics with and without OPAP
Resultant stego image and extracted secret image along with cover are shown
in Fig. 9 a and b, respectively. The proposed
method tested for fully embedding capacity and the results are tabulated in
Table 6. The obtained PSNR value is 48.72 dB with BER 0.057.
It is also observed that there is a significant improvement in the PSNR with
OPAP of 51.1 dB.
||First the indicator can be selected in three ways (Prime,
Even or Else)
||First the indicator can be selected in three ways (Prime, Even or Else)
||Then the channel is allotted in two ways (Odd or Even parity)
||The probability that two channels are selected is 2/3 (RG, GB, RB)
||Bits are embedded in a channel based on the indicator value
||The probability that at least two bits are embedded is 3/4 (00-no bits,
01, 10, 11-at least two bits)
||Then the embedding channel is selected in two ways
||So total complexity is 3*2*1/(2/3)*1/(3/4)*2=24
||If secret data scrambled with DES, the total complexity increases to 264x24
The MSE is around 0.5 and the corresponding PSNR is 51.1 dB for method 3, PSNR
of 44.46 dB for method 2 and PSNR of 47 dB for method 1 which is comparable
with Chan and Cheng (2004) and Amirtharajan
et al. (2010 a- c). the other LSB based substitution
are not considered (Amirtharajan and Balaguru, 2009,
2010) because they are uniform k bit embedding and their
complexity are less in comparison with this proposed method. In Comparison with
(Amirtharajan et al., 2010a, b,
c) these methods has better complexity of 264x24
CONCLUSION AND FUTURE WORK
This study is for enhancing the information security using the hiding technique (Steganography). The degradation of the cover images with improvement in capacity is common almost in all techniques. But this variation is handled with great care such that it is not visually perceptible to human eye. This variation could be further reduced using OPAP methodology which was discussed and this benefit was utilized at most in all implemented techniques. These three techniques provides considerably better security were discussed and the two optimal techniques, one for gray scale and another for color image steganography was recommended based on several aspects of level of security, capacity and error metrics values.
The common limitation that was faced in these techniques was loss of hidden
bits which could be recovered. The information security was mainly put forth
as many data communication happens through Internet. In this global village
of networks, there is a high possibility of cracking the sent information. If
sent stego- image got damaged during transmission or scrambling of the secret
information by attackers will cause permanent loss of the secret information.
Simple compression of the stego-image is also of no use, since the secret messages
are hidden in the LSB positions of cover image. This compression will modify
the LSB bits and once a stego-image is compressed then there is no chance of
The future scope of this study lies in eliminating the limitation of loss of
secret message in stego-image. This could be achieved by check methodologies
like parity bit checking, Integrity checking of the secret message etc. For
improving the security of each technique, cryptography could be combined with