SLSB: Improving the Steganographic Algorithm LSB

Juan José Roque

Universidad Nacional de Educación a Distancia, Spain

Abstract. This paper presents a novel steganographic algorithm based on the

spatial domain: Selected Least Significant Bits (SLSB). It works with the least

significant bits of one of the pixel color components in the image and changes

them according to the message’s bits to hide. The rest of bits in the pixel color

component selected are also changed in order get the nearest color to the

original one. This new method has been compared with others that work in the

spatial domain and the great difference is the fact that the LSBs bits of every

pixel color component are not used to embed the message, just those from pixel

color component selected.

1 Introduction

The steganography can be considered as a branch of cryptography that tries to hide

messages within others, avoiding the perception that there is some kind of message.

To apply steganographic techniques cover files of any kind can be used, although

archives of image, sound or video files are the most used today.

There are two trends at the time to implement steganographic algorithms: the

methods that work in the spatial domain (altering the desired characteristics on the

file itself) and the methods that work in the transform domain (performing a series of

changes to the cover image before hiding information. To select the best areas the

Discrete Cosine Transform DCT, Wavelet Transform, etc. are used).

While the algorithms that work in the transform domain are more robust, that is,

more resistant to attacks, the algorithms that work in the spatial domain are simpler

and faster.

The best known steganographic method that works in the spatial domain is the

LSB [1] (Least Significant Bit), which replaces the least significant bits of pixels

selected to hide the information. This method has several implementation versions

that improve the algorithm in certain aspects [2][3][4][5][6][7].

This paper proposes a new method, SLSB (Selected Least Significant Bit), that

improves the performance of the method LSB hiding information in only one of the

three colors at each pixel of the cover image. To select the color it uses a Sample

Pairs analysis, given that this analysis is more effective to detect hidden information.

Finally, applies a LSB Match [8] method so that the final color is as close as possible

to the original one.

The paper is organized as follows. Section 2 gives a brief classification of the

steganographic methods that works in spatial domain. Section 3 describes the

Roque J. (2009).

SLSB: Improving the Steganographic Algorithm LSB.

In Proceedings of the 7th International Workshop on Security in Information Systems, pages 57-66

DOI: 10.5220/0002169700570066

 SciTePress

proposed method. Section 4 is on the experimental results, followed by conclusions at

Section 5.

2 Methods in Spatial Domain

A basic classification of steganographic algorithms operating in the spatial domain as

the method for selecting the pixels distinguishes three main types: non-filtering

algorithms, randomized algorithms and filtering algorithms.

2.1 Non-filtering Algorithm

This is the simplest steganographic method based in the use of LSB, and therefore the

most vulnerable. The embedding process consists of the sequential substitution of

each least significant bit of the image pixel for each bit of the message. For its

simplicity, this method can camouflage a great volume of information [9].

This technique is quite simple. It is necessary only a sequential LSB reading,

starting from the first image pixel, to extract the secret message. This method also

generates an unbalanced distribution of the changed pixels, because the message is

embedded at the top of the image.

2.2 Randomized Algorithm

This technique was born as a solution for the problems of the previous method. Each

one, the sender and the receiver of the image has a password denominated stego-key

that is employed as the seed for a pseudo-random number generator. This creates a

sequence which is used as the index to have access to the image pixel. The message

bit is embedded in the pixel of the cover image as the index given by the pseudo-

random number generator. All the methods based on the pseudo-random number

generator must use an array to control the collisions [9].

The two main features of the pseudo-random permutation methods are the use of

password to have access to the message, and the well-spread message bits over the

image.

2.3 Filtering Algorithm

This algorithm filters the cover image by using a default filtert and hides information

in those areas that get a better rate. The filter is applied to the most significant bits of

every pixel, leaving the less significant to hide information. The filter ensures the

choice of areas of the image in the least impact with the inclusion of information,

which affects a greater difficulty of detecting the presence of hidden messages [10].

The retrieval of information is ensured because the bits used for filtering are not

changed, implying that the reapply the filter will selecte the same bits in the process

of concealment. It is the most efficient method to hide information.

The algorithm SLSB belongs to this type.

3 Description of the Algorithm SLSB

Figure 1 shows the structure of the algorithm SLSB:

Fig. 1. Structure of the algorithm SLSB.

3.1 Hiding Information in Only One Color

Most of the algorithms that work in the spatial domain using a LSB method (or any of

its derivatives) as the algorithm for information hiding, that is, hide one bit of

information in the least significant bit of each color of a pixel.

But these methods can’t stand a type of statistical analysis (such as RS [11] or

Sample Pairs [12]), even if partly camouflaged in the amount of information hidden.

File compression

Pixel filtering

Color selection

Steganographic

image

Message to hide

Bit replacement

LSB matching

Cover image

The problem stems from the fact that modifying the three colors of a pixel produces a

major distortion in the resulting color. This distortion is not visible to the human eye,

but detectable by statistical analysis.

For example, if a pixel of the cover image with the RGB (Red-Green-Blue code)

color A8A8A8 #

is used, binary 10101000-10101000-10101000, and 1 bit with value

1 is set on each LSB bit of each color component, to hide the message 111, the result

would be 10101001-10101001-10101001:

Table 1. Results obtained hiding the message 111 in the pixel 10101000-10101000-10101000

with the LSB method.

Hexadecimal Decimal Red Green Blue

Original pixel A8A8A8 11053224 168 168 168

Modified pixel A9A9A9 11119017 169 169 169

In theory the three least significant bits of the pixel have changed, introducing a

small distortion, but the difference between the old and new color represents a leap of

65793 colors in the scale of colors.

One method that would introduce more efficiency and less distortion would store

the 3 bits of information to hide in the same color. Using the same example, the 3 bits

of information will be introduced in the 3 LSB bits of green color (10101000-

10101111-10101000):

Table 2. Results obtained hiding the message 111 in the pixel 10101000-10101000-10101000

with the SLSB method.

Hexadecimal Decimal Red Green Blue

Original pixel A8A8A8 11053224 168 168 168

Modified pixel A8AFA8 11055016 168 175 168

In this case the leap in the scale of colors is 1792 colors (in the case of changing

the color green, if modify the blue color difference would be only 7 colors), that

being the extreme case because it has been replaced last 3 bits with 0 value for 3 bits

with a 1 value, that is, in most cases the distortion will be much lower.

In order to choose the color for the concealment, the SLSB algorithm performs a

preliminary Sample Pairs analysis and select the color with higher ratio because it

represents more diversity, leading to less noticeable changes. The choice of Sample

Pairs analysis over other stegoanalitics methods is due to the results provided by the

work of Ker [13], where this analysis shows that it is offering better results in terms

of detecting hidden information. Thus, the chosen color will be the one that provides

greater distortion and, therefore, the result of the withholding of information will be

less detectable.

3.2 LSB Match Adaptation

Following the work of Van Dijk [14] and Goljan [15], the method LSB Match

(designed to work with a single LSB bit) has been adapted to allow an LSB Match

with any number of LSB bits.

This method calculates the distance between the original color and the

steganographic color. Should the distance is greater than a certain threshold

(determined by the number of bits to hide) the color is decremented to get a final

color closest to the original, implying a further reduction in the distortion caused by

the hidden information.

For example, using a cover byte 11001000 to hide 3 bit of information (111), with

a simple LSB results in 11001111, which has a difference of 7 values with respect to

the original.

Applying the method proposed here to the above example results in 11000111,

with a distance of 1 from the original byte but with the same hidden information.

4 Results

To be able to compare the performance of this improvement on the LSB method, the

image on Fig. 1 will be used as cover with BMP (Bit Mapped Picture) format and

512x512 pixels in size (24 bits/pixel).

Fig. 2. Cover image.

4.1 Histogram Analysis

The purpose of the histogram analysis is to detect significant changes in frequency of

appearance of the colors by comparing the cover image with the steganographic

image.

To better align this analysis it has been carried out a detailed examination of the 4

components of any image: brightness and red, green and blue colors.

Histograms in Fig. 3 shows a frequency histogram of the image on Fig. 2 for the

four components mentioned above.

Fig. 3. Histograms of brightness, green, blue and red colors in the image on Fig. 2.

Histograms on Fig. 4 presents a frequency histogram of the image on Fig. 2 with a hidden

message of 141.744 bits and a method of 1 bit/pixel, producing a hiding rate of 54%, for the

four components above cited.

Fig. 4. Histograms of brightness, green, blue and red colors in the image on Fig. 2 with hidden

information.

There are only changes in the histograms of brightness and green color (the one

chosen by the algorithm as the optimal color of concealment).

Despite having a hiding rate of 54%, the changes are negligible (0.01 in the

standard deviation of brightness and an average of 0.01 in the green color).

According to the results, it can be said that the new proposed algorithm is immune

to attacks based on a comparison of histograms of the original image and the

steganographic image.

4.2 Another Steganographic Tools Comparison

To conclude the analysis of the results of the new proposed algorithm its performance

is compared with that from the best known and more used today steganographic tools.

This comparison focuses on two aspects: the results of the RS and Sample Pairs

analysis of steganographic images and the analysis of the results of the metrics of

distortion.

Table 3 shows a comparison of the results for the steganographic images obtained

with the various tools in front of the RS analysis and the Sample Pairs analysis.

Table 3. Results obtained using a cover image of 786.486 bytes (Fig. 2) and a hidden message

of 31.071 bytes (TXT file).

Tool RS analysis

Sample

Pairs

analysis

Hermetic Stego [20] 75,46911

73,39835

Invisible Secrets [23] 70,06539

69,32617

Hide4PGP [21] 30,60135

30,19531

Contraband [16] 14,78796

11,83324

wbStego [27] 14,33760

13,42652

White Noise Storm [28] 11,91106

10,12546

Digital Invisible Ink Toolkit [18] 9,61806

7,84342

JPHS [24] 2,62679

2,68511

S-tools [26] 2,30629

2,10435

EikonaMark [19] 1,86631

1,31909

Data Privacy Tools [17] 1,43103

0,96443

Hide In Picture [22] 1,03530

1,06373

SLSB algorithm (1 bit/pixel) 0,89172

0,61744

SLSB algorithm (2 bits/pixel) 0,80084

0,60556

Original image 0,67766

0,51907

SLSB algorithm (3 bits/pixel) 0,64431

0,47867

Steghide [25] 0,63543

0,37671

The results show that the new algorithm, in its 3 versions, offers among the best ever

results.

Table 4 shows a comparison of the results of the metrics of distortion [9] (Average

Absolute Difference, Mean Squared Error, Lp-Norm, Laplacian Mean Squared Error,

Signal to Noise Ratio, Peak Signal to Noise Ratio, Normalised Cross-Correlation and

Correlation Quality) applied to steganographic images obtained by different tools.

Table 4. Results obtained using a cover image of 786.486 bytes (Fig. 2) and a hidden message

of 31.071 bytes (TXT file).

Tool AAD MSE LP LMSE

SNR PSNR

Original

image

0,0 0,000

0,0

0,000

0,0

1,000 1,26643

SLSB

algorithm (1

bit/pixel)

5,3 0,020

36,9

1,113

6,7

2,0

0,999 1,26643

SLSB

algorithm (2

bits/pixel)

6,1 0,043

53,2

2,279

3,2

9,9

1,000 1,26643

SLSB

algorithm (3

bits/pixel)

9,1 0,137

94,9

7,288

1,0

3,1

1,000 1,26643

Contraband 10368,4 0,872

26038,6

0,002

136124,5

415146,6

0,999 1,26597

Data Privacy

Tools

25155,1 13,760

110709,9

0,045

7530,1

22965,0

1,001 1,26798

Digital

Invisible Ink

Toolkit

10377,1 0,947

26053,1

0,002

135973,3

414685,4

1,000 1,26662

EikonaMark 150528,2 90,996

232052,0

0,279

1713,9

5227,1

0,997 1,26307

Hermetic

Stego

42354,0 4,500

52662,9

0,009

33278,5

101491,3

1,000 1,26643

Hide4PGP 10523,7 0,477

26211,1

0,002

134339,1

409701,5

1,000 1,26643

Hide In

Picture

10493,4 0,665

26194,6

0,002

134508,4

410218,1

1,000 1,26747

Invisible

Secrets

32934,0 3,007

46420,3

0,007

42830,9

130623,8

1,000 1,26649

JPHS 54609,9 9,141

89871,2

0,011

11427,0

34849,6

1,000 1,26643

Steghide 1465,3 0,320

16270,8

0,001

348621,5

1063210,6

0,999 1,26643

S-tools 552,4 0,025

6005,0

1,202

2559393,7

7805527,2

1,000 1,26643

wbStego 10408,3 0,947

26092,5

0,002

135563,4

413435,6

1,000 1,26668

White Noise

Storm

13828,4 1,101

30071,6

0,003

102060,9

311260,8

0,999 1,26640

This table can verify that the new algorithm (in any of its three versions) offers the

best results in the metrics AAD, MSE, LP, SNR, PSNR, NCC and CQ, and provide

the same result as the original image in the last two columns.

5 Conclusions

This paper proposes a new method, SLSB (Selected Least Significant Bit), that

improves the performance of the LSB method hiding information in only one of the

three colors at each pixel of the cover image. For the selection of color it uses a

Sample Pairs analysis, given that this analysis is more effective to detect hidden

information. Finally, applies a LSB Match [8] method so that the final color is as

close as possible to the original one.

A summary of its features could be:

- It is based on the LSB method, but can hide the same information much more

effectively using bits of just one color.

- Implement the LSB Match method to reduce the difference between the original

pixel and the steganographic pixel.

- Perform a Sample Pairs analysis prior to steganography, which allows you to

select the best color of the three possible to hide information.

- Use a pixel selection filter to obtain the best areas to hide information.

- It is immune to visuals attacks. Changes are undetectable with the naked eye, and

a filter of LSB bits doesn’t present areas of random information that could

indicate the presence of hidden information.

- It is immune to statistical attacks, as two colors for each pixel are equal to those of

the original image, and the final ratio of analysis is very close to the original

image, which doesn’t raise suspicion it contains hidden information. Even in some

cases get better rates than those of the original image, creating confusion over

which of two images would be the original.

- It is immune to attacks by comparing histograms, as the frequency of appearance

of colors in the steganographic image is very similar to that of the cover image.

- It yields well above that of most steganographic tools used today, both in RS and

Sample Pairs analysis and in metric of distortion.

Future works will aim to achieve better performance and be undetectable by the most

famous steganographic analysis, for example, changing bits undisturbed by the

concealment of the message.

References

1 Kurak, C. and McHugh, J.: A Cautionary Note on Image Downgrading. Computer Security

Applications Conference. San Antonio, USA (1992)

2 Moskowitz, I., Longdon G. and Chang, L.: A New Paradigm Hidden in Steganography.

Center for High Assurance Computer Systems. Naval Research Laboratory (2000)

3 Sharp, T.: An implementation of key-based digital signal steganography. Proc. of 4th

International Workshop on Information Hiding. Springer-Verlag (2001)

4 Kawaguchi, E. and Eason, R.: Principle and applications of BPCS-Steganography. Kyushu

Institute of Technology. Japan (1998)

5 Bender, W., Gruhl, D., Morimoto, N. and Lu, A.: Techniques for data hiding. IBM Systems

Journal, Vol. 35, Nos. 3&4 (1996)

6 Moskowitz, I., Johnson, N. and Jacobs, M.: A detection study of an NRL steganographic

method. Naval Research Laboratory. Washington (2002)

7 Noto, M.: MP3Stego: Hiding text in MP3 files. Sans Institute (2003)

8 Sharp, T.: An implementation of key-based digital signal steganography. Proc. 4

International Workshop on Information Hiding. Lecture Notes on Computer Science, vol.

2137, Springer-Verlag (2001)

9 Katzenbeisser, S. and Petitcolas, F.: Information hiding techniques for steganography and

digital watermarking. Artech House Books (1999)

10 Hempstalk, K.: Hiding behind corners: using edges in images for better steganography.

Computing Womens Congress conference in Hamilton. New Zealand (2006)

11 Fridrich, J., Goljan, M. and Du, R.: Reliable detection of LSB steganography in color and

grayscale images. ACM Workshop on Multimedia and Security (2001)

12 Dumitrescu, S., Wu, X. and Wang, Z.: Detection of LSB steganography via sample pairs

analysis. Information Hiding Workshop. Lecture Notes on Computer Science, vol. 2578,

Springer-Verlag (2003)

13 Ker, A.: Improved detection of LSB steganography in grayscale images. Information

Hiding. Lecture Notes on Computer Science, vol. 3200, Springer-Verlag (2005)

14 Van Dijk, M. and Willems, F.: Embedding information in grayscale images. Proc. 22

Symposium on Information and Communication Theory in the Benelux. The Netherlands

(2001)

15 Goljan, M. and Holotyak, T.: New blind steganalysis and its implications. Proc. SPIE

Electronic Imaging, Photonics West (2006)

16 Contraband 9g, http://www.GalaxyCorp.com/009.

17 Data Privacy Tools 3.5, http://www.xs4all.nl/~bernard/home_e.html

18 Digital Invisible Ink Toolkit 1.5, http://diit.sourceforge.net

19 EikonaMark 4.7, http://www.alphatecltd.com/watermarking/eikonamark/eikonamark.html

20 Hermetic Stego 7.41t, http://www.hermetic.ch/hst/hst.htm

21 Hide4PGP 2.0, http://www.heinz-repp.onlinehome.de/Hide4PGP.htm

22 Hide In Picture 2.1, http://sourceforge.net/projects/hide-in-picture

23 Invisible Secrets 4.6.3, http://www.invisiblesecrets.com

24 JPHS 0.5, http://linux01.gwdg.de/~alatham/stego.html

25 StegHide 0.5.1, http://steghide.sourceforge.net

26 S-tools 4, http://www.spychecker.com/program/stools.html

27 wbStego 4.3, http://www.8ung.at/wbailer/wbstego/

28 White Noise Storm 2.10, http://www.nic.funet.fi/pub/crypt/steganography/