International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

25

Increased Cover Capacity using Advanced Multiple LSBAlgorithms

H B KekreSr. Professor. Department of

Computer ScienceMPSTME, NMIMS University

Mumbai, India+91 - 9323557897

hbkekre@yahoo.com

A A AthawaleAsst. Prof. Computer Engg.

DepartmentTSEC, Bandra (W) Mumbai, India

Research Scholar, NMIMS+91 - 9226977842

athawalearchana@gmail.com

U A AthawaleLecturer, Department of MCA

Bharati Vidyapeeth’s Institute ofManagement & Information Tech.

Navi Mumbai, India+91 - 9221289613

uttara.athawale@gmail.com

ABSTRACT

This paper describes the methods to increase the capacity ofthe cover image for information hiding. In this paper we haveproposed two novel multiple least significant bit replacementtechniques KIMLA and KAMLA to increase the informationhiding capacity of the cover image. Comparative results ofKIMLA, KAMLA and their previous version KMLA [1] arealso shown. These novel techniques use multiple LSB’S bitreplacement technique to hide the secret information. Theproposed methods KAMLA and KIMLA have shown betterresults for AFCPV, PBC ratios and they have been tested forsecret information in image as well as text form.

Categories and Subject DescriptorsI.4.m Miscellaneous

General TermsPerformance, Experimentation, Algorithm

Keywords:Steganography, Steganalysis, LSB matching, LSBsubstitution, AFCPV, PBC, KMLA, KIMLA, KAMLA

1. INTRODUCTIONSteganography is the art and science of writing hiddenmessages in such a way that no one, apart from the senderand intended recipient, suspects the existence of the message,a form of security through obscurity [1, 6]. The wordsteganography is of Greek origin and means "concealedwriting" from the Greek words steganos meaning "covered orprotected", and graphein meaning "to write" [2, 3, 5].

Secret data can be an image file, text file, audio file, etc.Since, in all steganographic methods, secret data usually isconverted into bit stream, free of its content, all types of dataare the same in this respect. Images that are used for insertingand hiding secure data are called ‘cover image’ and the image

after the secret bits are inserted is called ‘stego image’.

2. RELATED WORKThere are many different steganographic algorithms. Some ofthem are in spatial domain [7-9, 11-14] and others are intransform domain.[15-19] LSB (Least Significant Bit)replacement steganography is a popular and simple techniquethat can hide message bits in LSB planes of image pixels.LSB based methods can be divided into two main groups:LSB replacement, which simply replaces LSB bits of coverimage with secret bits, and LSB matching where pixels arerandomly incremented or decremented [2, 4, 10]

In spatial domain, lowest bit plane of a bitmap image is usedto hide the secret data, because it cannot be detected by humaneyes. The capacity of the cover image to hide the data is 1/8th

of the total size of the image if only the least significant bit isused. To increase the capacity the 2nd and the 3rd lowest bitplanes can be used to hide the data. Capacity of the cover isincreased but the quality of the stego image deteriorates. If thedata is hidden in 2nd, 3rd and 4th LSB bits only for selectedimage bytes the cover capacity can be increased withoutdeteriorating the overall quality of the stego image. Thistechnique of using Multiple LSB’s was proposed in the KMLAalgorithm [1].

The rest of this paper is organized as follows. Section 3describes the new proposed methods. Section 4 shows theexperimental results. Section 5 shows the comparison of thethree algorithms. The paper is concluded in Section 6.

3. THE PROPOSED ALGORITHMSHere, two new Algorithms are introduced namely KIMLA andKAMLA to increasing the hiding capacity of the cover image.

KIMLA ALGORITHM:

Kekre’s Improved Multiple LSB Algorithm (KIMLA) is aMultiple LSB replacement technique; a spatial domainmethod. In this method up to four LSB’S of an image byte canbe changed. In the Table 1, the algorithm is given, wheremultiple LSB‘s are changed based on the first Four MSB’s.The column 2 shows the decimal equivalent and the nextcolumn shows the binary version of the decimal number plusone except 15 is retained as binary 15. The next column showsnumber of bits to be replaced and the last column shows thelocation of the bits to be replaced. This algorithm is based onmaintaining the upper bound of the error at 6.25%. This

Permission to make digital or hard copies of all or part of this workfor personal or classroom use is granted without fee provided thatcopies are not made or distributed for profit or commercialadvantage and that copies bear this notice and the full citation on thefirst page. To copy otherwise, or republish, to post on servers or toredistribute to lists, requires prior specific permission and/or a fee.ICWET’11, February 25–26, 2011, Mumbai, Maharashtra, India.Copyright © 2011 ACM 978-1-4503-0449-8/11/02…$10.00.

International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

26

algorithm is explained by an example given below. The 4-bitdecimal value of the MSB of the image byte is determined.The number of LSB’S that can be replaced by the message bitsis the number of 1’s in the 4-bit binary representation of thenext value of the decimal number. If the Image byte is00111100, the 4 MSB’s are 0011 equivalent to decimal 3. Thenext decimal number is 4 and the 4 - bit binary representationfor 4 is 0100. It contains one 1’s so BRC (Bit replacementcount) is 1. Therefore 1 LSB in image byte will be replaced by1 message bit, so that byte in the image will be changed to0011110c. If the 4MSB’s are all 1 i.e. 1111 then all the fourLSB’s are replaced by message bits. i.e. the image byte1111xxxx will change to 1111cccc as shown in Table 1.

Table 1: KIMLA Bit Replacement Pattern

C: bit to be replaced, X: unchanged bit.

KAMLA ALGORITHM: This method is advanced versionof KIMLA algorithm. In Kekre’s Advanced Multiple LSBAlgorithm (KAMLA) up to four LSB’S of an image byte canbe changed. BRC is determined as in KIMLA. Sum of 4MSB’s is determined in the 4-bit binary representation of theimage byte. The maximum of the two values i.e. BRC or sumof MSB is taken as the MAXBRC. If the Image byte is00111100, the decimal equivalent of 4 MSB’s is 3 and its nextdecimal number is 4. According to Table 1, the BRC is 1.Therefore only 1 bit can be replaced according to KIMLA. Inthe new algorithm, first the number of 1’s in the 4 MSB’s iscalculated.

Accordingly, for this example this works out to be 2. Now, themaximum of these two numbers is selected and that is 2.Thus 2 LSB’s are changed by message bits. The image byteis therefore changed to 001111cc. The Bit replacementpattern is for KAMLA is as shown in the Table2

Table 2: KAMLA Bit Replacement Pattern

C: bit to be replaced, X: unchanged bit.

4. EXPERIMENTAL RESULTSTo test the performance of the proposed method sixteen 24bit BMP color images of size 512 X 512 are used as cover

(a) (b)

(c) (d)

(e) (f)

(g) (h)Cover images Stego images

Figure 1: Cover and Stego for KIMLA algorithm

Results for KIMLA algorithm, Message Type: Image

4 MSB Dec NextDec BRC LSB bits

changed0000 0 0001 1 XXXC0001 1 0010 1 XXXC0010 2 0011 2 XXCC0011 3 0100 1 XXXC0100 4 0101 2 XXCC0101 5 0110 2 XXCC0110 6 0111 3 XCCC0111 7 1000 1 XXXC1000 8 1001 2 XXCC1001 9 1010 2 XXCC1010 10 1011 3 XCCC1011 11 1100 2 XXCC1100 12 1101 3 XCCC1101 13 1110 3 XCCC1110 14 1111 4 CCCC1111 15 1111 4 CCCC

4 MSB BRC No. ofOne’s

MaxBRC

LSB bitschanged

0000 1 0 1 XXXC0001 1 1 1 XXXC0010 2 1 2 XXCC0011 1 2 2 XXCC0100 2 1 2 XXCC0101 2 2 2 XXCC0110 3 2 3 XCCC0111 1 3 3 XCCC1000 2 1 2 XXCC1001 2 2 2 XXCC1010 3 2 3 XCCC1011 2 3 3 XCCC1100 3 2 3 XCCC1101 3 3 3 XCCC1110 4 3 4 CCCC1111 4 4 4 CCCC

International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

27

images. The Stego images obtained for KIMLA, KAMLAfor message type- image are shown in Figure 1 and Figure 2

(a) (b)

(c) (d)

(e) (f)

(g) (h)

Cover images Stego images

Figure 2: Cover and Stego for KAMLA algorithm

Results for KAMLA algorithm, Message Type: Image

The secret message consists of varying size bitmap files,plain text, word documents and pdf files.The messages selected were Human eye, Driver license,finger print etc. They are shown in Figure 3

128 X128 180 X 115 120 X

198 218 X 138 237 X153

MSG 1 MSG 2 MSG 3 MSG 4 MSG 5

Eye Id.-Card1 FingerPrint Id.-Card2 Id.-Card3

Figure 3: Message files Image embedded in stego files.

Similarly word, txt and pdf files were used as message toobtain the stego files for all the cover files. The details ofmessage files for image, word document, plain text and pdfare given in Table 3 to Table 6

Table 3: Details of Message Image Files,File type: 24 windows bitmap image (3BPP)

MessageImage

Sizein kb

Size inpixels

(WXH)

Total bytes usedfor image

(W x H x 3)1. Eye 49 128 X 128 491522. Id.-Card1 61 180 X 115 621003. Finger Print 70 120 X 198 712804. Id.-Card2 89 218 X 138 902525. Id.-Card3 107 237 X 153 108783

Table 4: Details of Message Word document filesFile type: Word document (.docx)

Worddocument

Sizein kb

No. ofwords

Containsimage / table

No. ofpages

MSG1 15 1262 N 3MSG2 23 4171 N 7MSG3 27 663 Y (1 image) 3MSG4 31 5374 Y (2 tables) 12

Table 5: Details of plain text filesFile type: plain text file (.txt)

Text Files Size inkb

No. of words No. of pages(A4 size)

MSG1 7 1262 3MSG2 24 4171 7MSG3 30 5209 10MSG4 50 8976 26

Table 6: Details of pdf (acrobat) fileFile type: pdf file (.pdf)

pdf file Size(kb)

No. ofwords

Contains image /table (Y/N)

No. ofpages

MSG1 17 352 N 1

MSG2 20 1440 Y ( 1image, 1 table ) 3

MSG3 27 1802 Y (1 image, 1 table) 6

5. COMPARISION OF KMLA, KIMLAand KAMLA.The Capacity of Cover in bits for different methods can becalculated as given below

LSB replacement method: For every image byte we canhide one bit of message, so it is equal to the number of imagebytes.

International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

28

KMLA algorithm: n1*4 + n2*3 + n3* 2 + n4 bits, wheren1 = number of bytes that have MSB as 1111n2 = number of bytes that have MSB as 1110n3 = number of bytes that have MSB as 1100 or 1101n4 = number of remaining bytes.

KIMLA: n0 * BRC (0000) + n1 * BRC (0001) + n2 x BRC(0010) + …. + n15 * BRC (1111), where ni is the No. of byteswith first 4 MSB’s being the binary representation of i.

KAMLA: n0 x MAXBRC (0000) + n1 x MAXBRC (0001)+…. + n15 x MAXBRC (1111), where ni is the No. of byteswith first 4 MSB’s being the binary representation of i.

The percentage increase in capacity over the LSBreplacement technique is calculated for KMLA; KIMLA &KAMLA The average cover capacity in bits is calculated forall covers for the different methods and is shown in theTable7

Table 7: Cover Capacity over LSB for KMLA KIMLAand KAMLA

Algorithm Cover CapacityKMLA 1117199KIMLA 1751349KAMLA 1937323

Figure 4: Cover Capacity

Table 8: % Increased Cover Capacity over LSB for

KMLA KIMLA and KAMLA

Algorithm % of Increase of Cover Capacity overLSB Replacement Method

KMLA 142KIMLA 223KAMLA 246

Figure 5: Cover Capacity

It is observed that the cover capacity for KMLA is increasedby 42%, for KIMLA by 123% and for KAMLA by 146 %.The quality of the stego images is good and not degraded.

To compare the quality of stego image, original cover imageand stego images are compared for all covers for everymessage. Mean Square Error for Cover MSEC, Peak Signalto Noise Ratio for Cover PSNRC, Average FractionalChange in Pixel Value AFCPV and PBC Percentage ByteChange were calculated and their average was taken for eachmessage image for the three algorithms KMLA, KIMLA,KAMLA. The results are shown in Table 9 to Table 20

MESSAGE TYPE: IMAGETable 9: Algorithm- KMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 1.36 48.33 0.51 20.68MSG2 1.53 47.41 0.63 26.23MSG3 2.14 45.94 0.73 29.95MSG4 2.34 45.51 0.96 38.43MSG5 3.11 44.39 1.19 46.03

Average 2.10 46.32 0.80 32.26

Table 10: Algorithm- KIMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 1.96 45.88 0.49 16.71MSG2 2.34 44.94 0.61 21.14MSG3 3.44 43.21 0.73 24.06MSG4 3.67 42.84 0.89 30.66MSG5 4.79 41.69 1.09 36.83

Average 3.24 43.71 0.76 25.88

Table 11: Algorithm- KAMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 2.03 45.64 0.49 15.98MSG2 2.41 44.76 0.60 20.24MSG3 3.57 43.01 0.74 23.02MSG4 3.83 42.63 0.89 29.29MSG5 4.93 41.53 1.09 35.22

Average 3.36 43.51 0.76 24.75

MESSAGE TYPE: WORD FILE (DOCX)Table 12: Algorithm- KMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.432 53.84 0.168 6.244MSG2 0.678 51.75 0.2532 9.684MSG3 0.831 50.93 0.2975 11.36MSG4 0.91 50.4 0.3329 12.96

Average 0.71 51.73 0.26 10.06

International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

29

Table 13: Algorithm- KIMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.63 50.91 0.16 5.02MSG2 0.96 49.10 0.25 7.84MSG3 1.16 48.30 0.29 9.19MSG4 1.28 47.81 0.32 10.49

Average 1.01 49.03 0.25 8.13

Table 14: Algorithm- KAMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.67 50.48 0.16 4.77MSG2 1.01 48.73 0.24 7.48MSG3 1.21 47.96 0.29 8.78MSG4 1.34 47.49 0.32 10.04

Average 1.06 48.67 0.25 7.77

MESSAGE TYPE: PLAIN TEXT FILE (TXT)

Table 15: Algorithm- KMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.176 57.61 0.0798 2.948MSG2 0.645 51.88 0.2605 9.997MSG3 0.841 50.72 0.3262 12.66MSG4 1.502 48.08 0.5344 21.17

Average 0.79 52.07 0.30 11.69

Table 16: Algorithm- KIMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.26 54.70 0.07 2.36MSG2 0.92 49.32 0.25 8.09MSG3 1.18 48.22 0.31 10.25MSG4 2.09 45.72 0.51 17.11

Average 1.11 49.49 0.29 9.45

Table 17: Algorithm- KAMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.27 54.32 0.07 2.24MSG2 0.95 48.99 0.25 7.73MSG3 1.22 47.93 0.31 9.80MSG4 2.14 45.49 0.51 16.37

Average 1.15 49.18 0.28 9.04

MESSAGE TYPE: pdf FILETable 18: Algorithm- KMLA

MSG FILE MSEC PSNRC AFCPV PBC

MSG1 0.493 53.37 0.1889 7.002

MSG2 0.557 52.63 0.2226 8.446

MSG3 0.786 51.1 0.3006 11.58

Average 0.61 52.37 0.24 9.01

Table 19: Algorithm- KIMLAMSG FILE MSEC PSNRC AFCPV PBC

MSG1 0.68 50.62 0.18 5.63MSG2 0.80 49.89 0.21 6.81MSG3 1.10 48.53 0.29 9.38

Average 0.86 49.68 0.23 7.27

Table 20: Algorithm-KAMLA

MSG FILE MSEC PSNRC AFCPV PBCMSG1 0.72 50.20 0.18 5.37MSG2 0.84 49.51 0.21 6.49MSG3 1.14 48.22 0.29 8.97

Average 0.90 49.31 0.22 6.94

The average of all the above metrics was compared as shown

in the following charts from Figure 6 .to Figure 9

(a) (b)

(c) (d)

Figure 6: MESSAGE TYPE: IMAGE FILE

International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

30

(a) (b)

(c) (d)

Figure 7 : MESSAGE TYPE : WORD FILE ( DOCX)

(a) (b)

(c) (d)

Figure 8 : MESSAGE TYPE: pdf FILE

(a) (b)

(c) (d)

Figure 9: MESSAGE TYPE : TEXT FILE ( TXT)

6. CONCLUSIONIn this paper, two new algorithms are proposed to increasethe capacity of the Cover image. In KMLA [1], the capacityof the cover image was increased by 42 over and abovenormal LSB substitution method. In the proposed newalgorithm, KIMLA the capacity is increased by whooping123 % over LSB method. In the advanced algorithmKAMLA, this capacity is raised further to 146 %. Themaximum MSE for both these algorithms is well below 4%.Amongst these three algorithms, the Percentage Byteschanged decreases with increase in capacity. The AFCPV forKIMLA and KAMLA is less than KMLA giving moreimperceptibility in the stego images.

7. REFERENCES[1] Dr. H. B. Kekre, Ms. Archana A. Athawale,

“Information Hiding using LSB Technique withIncreased Capacity”, International Journal ofCryptography and Security, Special issue onSteganography, Oct 2008

[2] http://www.eurojournals.com/ejsr_31_2_01.pdf[3] N. Johnson and S. Jajodia, “Exploring steganography:

seeing the unseen,” IEEE Computer, pp. 26-34,February 1998

[4] T. Sharp, “An implementation of key-based digitalsignal steganography,” in Proc. Information HidingWorkshop, vol. 137, Springer LNCS, 2001, pp. 13–26.

International Conference and Workshop on Emerging Trends in Technology (ICWET 2011) – TCET, Mumbai, India

31

[5] Johnson, Neil F, “Steganography”, 2000:www.jjtc.com/stegdoc/index2.htm

[6] H. B. Kekre, Archana A. Athawale, Pallavi N.Halarnkar , “Increased Capacity of Information Hidingin LSB’s method for text and Image” , InternationalJournal of Electrical, Computer and SystemsEngineering , volume 2, No. 4http://www.waset.org/ijecse/v2/ v2-4-34.pdf (Europe)

[7] H. B. Kekre, Archana A. Athawale, Pallavi N.Halarnkar , “Polynomial Transformation to improveCapacity of Cover Image for Information Hiding inMultiple LSB’s”, International Journal of EngineeringResearch and Industrial Applications (IJERIA),AscentPublications, Volume II, March 2009.Pune (India)

[8] H. B. Kekre, Archana Athawale and Pallavi N.Halarnkar, “Performance Evaluation of Pixel ValueDifferencing And Kekre’s Modified Algorithm ForInformation Hiding In Images”, ACM InternationalConference on Advances in Computing,Communication and Control (ICAC3).2009 (Availableon ACM Portal:http://portal.acm.org/citation.cfm?id=1523103.1523172)

[9] J. Mielikainen, “LSB Matching Revisited,” IEEE SignalProcessing Letters, vol. 13, no.5, May 2006.

[10] D. C. Wu and W. H. Tsai, “A steganographic methodfor images by pixel-value differencing,” PatternRecognition Letters, vol. 24, no. 9-10, pp. 1613-1626,June 2003

[11] H.B. Kekre, Archana A. Athawale, Pallavi N.Halarnkar, “High Payload using High Boost Filtering inKekre’s Multiple LSB’s Algorithm”, 2nd InternationalConference on Advances in Computer Vision andInformation Technology ACVIT 2009, 16th -19thDecember 2009, Aurangabad

[12] H. B. Kekre, Archana A. Athawale, Sudeep D.Thepade, “Clandestine Data Entrenching and Salvaging“National Conference on Information andCommunication Technology 29th Feb and 1st Mar,NCICT-08.

[13] H. B. Kekre, Archana A. Athawale, “Personal DataIngrain and Regain”, NCA, FCRCE, Bandra (W),Mumbai, 16th -17th May 2008.

[14] H.C. Wu, N.I Wu, C.S. Tsai and M.S. Hwang, “ImageSteganographic scheme based on pixel-valuedifferencing and LSB replacement methods”,VISP(152), No. 5, October 2005

[15] H. B. Kekre, Archana A. Athawale, Uttara A. Athawale,“Increased Capacity for Information Hiding usingWalsh Transform“,ICWET-10 International Conferenceand Workshop on Emerging Trends in Technology(ICWET 2010), 26-27 Feb 2010, Available on ACMPortal athttp://portal.acm.org/citation.cfm?id=1741906.1741923

[16] H. B. Kekre, Archana A. Athawale, Pallavi N.Halarnkar, “Increased Capacity and High Security forEmbedding Secret Message in Transform Domain usingDiscrete Cosine Transform”, (Technopath)

[17] H. B. Kekre, Archana A. Athawale, Pallavi N.Halarnkar, “Performance Comparison of DCT andWalsh Transform for Steganography”, InternationalConference and Workshop on Emerging Trends inTechnology (ICWET 2010), 26-27 Feb2010.Site: http://portal.acm.org / citation.cfm? id = 1741906 . 1741921& coll = GUIDE & dl = GUIDE & CFID = 91821685 &CFTOKEN = 15950099

[18] H. B. Kekre, Archana A. Athawale, Dipali Sadavarti ,“A Novel Steganographic Scheme using Discrete SineTransform based on Energy Distribution” Springer-International Conference on Contours of ComputingTechnology (Thinkquest-2010), 13-14th Mar 2010 (tobe uploaded on Online Springer Link )

[19] H. B. Kekre, Archana A. Athawale, Dipali Sadavarti, “Algorithm to Generate Kekre’s Wavelet Transformfrom Kekre’s Transform”, International Journal ofEngineering, Science and Technology, (IJEST)ISSN:0975-5462, Vol. 2(5), 2010, 756-767 website-http://www.ijest.info/docs/IJEST10-02-05-22.pdf(India)