Hough Signature based Authentication of image through Daubechies ...

S4 : 84

Hough Signature based Authentication of image through Daubechies Transform technique (HSADT)

Hough Signature based Authentication of image through Daubechies Transform technique (HSADT) Madhumita Sengupta1 and J. K. Mandal2 1

2





Life Member, CSI Department of Computer Science and Engineering, University of Kalyani, Kalyani, Nadia-741235, West Bengal, India; Phone: 9432145902; Fax: 033-2580-9617; Email: [email protected]. Life Member, CSI Department of Computer Science and Engineering, University of Kalyani, Kalyani, Nadia-741235, West Bengal, India; Phone: 9434352214; Fax: 033-2580-9617; Email: [email protected].

In this paper a 4 x 4 Daubechies transform based authentication technique termed as HSADT has been proposed to authenticate gray scale images. The cover image is transformed into the frequency domain using 4 x 4 mask in a row major order using Daubechies transform technique, resulting four frequency subbands AF, HF, VF and DF. The highest frequencies are selected to generate signature through Hough transform. Two to four bits of information from secret signature is embedded into four subbands in non sequential manner based on hash function. Experimental results are computed and compared with the existing authentication techniques such as Li’s, SCDFT, Region-Based method and other similar techniques based on Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR), Image Fidelity (IF) and Universal Quality Image (UQI) which shows improved performance in HSADT. Index Terms: Authentication, Steganography, Daubechies transform technique, Hough transform, Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR), Frequency domain, Information hiding. I. Introduction Digital medium stores massive amount of information in an incredibly less space. The critical human psychologies impart brain to identify the object related information inside digital data. With a prior knowledge human eye is able to identify the actor and actresses from the running videos. Any audio can also be identified correctly by the name of the singer. But in general the paintings or photographs cannot hold the identification clue of the painter or the photographer. Many techniques are available in today’s digital world to protect ownership/verify the authenticity of digital content both in spatial and frequency domain. The proposed work is a frequency domain based technique termed as HSADT, where self authentication of digital image is achieved without external information. Various parametric tests are performed and results obtained are compared with existing techniques like Yuancheng Li’s method [1], Region-Based method [2], SCDFT [3], SAWT [4], SADCT [5], IAHTSSDCT [6] and AWTDHDS [7] based on Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR) and Image Fidelity (IF) analysis [8] to show a consistent relationship with the quality perceived by the HVS (Human Visual System). Section II deals with the overall scheme, the discrete Daubechies transforms technique [9, 14] is given in Section

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013

III. Secret message computation through Hough transform is given in section IV. Embedding and authentication process is outlined in section V. Section VI analyzed the results of proposed work and comparisons with existing techniques. Conclusions are drawn in section VII and that of references are cited at end.

II. The Scheme The HSADT is divided into two major phases. First ‘encryption technique’ to authenticate a cover image at sender side with self generated signature through Hough transform, followed by the generation of stego-image. Next ‘authentication technique’ at receiver side to verify the originality of the image. Two transformation techniques are associated with this scheme to enhance the robustness of the system. Daubechies transformation is used to convert image from spatial domain to frequency domain and Hough transform generates the unique secret image from cover image. Source image of dimension N x N passes through forward Daubechies transform (FDT) based on 4 x 4 non overlapping masks in a row major order to generate the corresponding frequency coefficients such as AF, HF, VF and DF as shown in figure 3 and figure 4. Out of the four ‘frequency coefficient’

XXX

verify the hniques are tness of the nvert image and Hough from cover

ugh forward overlapping rresponding F as shown „frequency erate unique m. Eight to bedded into the hash technique is

Daubechies transform (FDT) based on 4 x 4through non overlapping Source Source image image of of dimension dimension N N x x N N passes passes through forward forward masks in a transform row major(FDT) orderbased to generate the corresponding Daubechies on 4 x 4 non Daubechies transform (FDT) based on 4 x 4 non overlapping overlapping frequency acoefficients such as to AF, HF, VFthe and DF as shown masks masks in in a row row major major order order to generate generate the corresponding corresponding in figure coefficients 3 and figure 4. Out HF, of VF theand four „frequency frequency such as AF, frequency coefficients such as AF, HF, VF and DF DF as as shown shown coefficient‟ sub-bands the DF is selected to generate unique in figure 3 and figure 4. Out of the four „frequency Madhumita in figure 3 Sengupta, and figure et.4.al. Out of the four „frequency secret signature imagethethrough Hough transform. Eight to coefficient‟ sub-bands DF is selected to generate unique coefficient‟ sub-bands the DF is selected to generate unique sixteensignature bits of secret image perHough mask transform. are embedded into secret image through Eight to secret signature image through Hough transform. Eight to sub-bands is selected to generate unique coefficients the of DF frequency sub-bands based embedded on thesecret hash sixteen sixteen bits bits of of secret secret image image per per mask mask are are embedded into into signature image through Hough transform. Eight to sixteen function. Schematic representation of encryption technique is coefficients of frequency sub-bands based on the hash coefficients of frequency sub-bands based on the hash bits of secret image mask are embedded into technique coefficients shown in Schematic figure 1. per function. representation of encryption is function. Schematic representation of encryption technique is of frequency sub-bands based on the hash function. Schematic shown shown in in figure figure 1. 1. representation of encryption technique is Forward shownDaubechies in Fig. 1. Original Image N xImage N Original Original Image NxN NxN Generate signature Generate Generate signature signature

Mask of 4x4 in Transform (FDT) row order Forward Daubechies Maskmajor of 4x4 in Forward Daubechies Mask of 4x4 in Transform (FDT) row major order Transform (FDT) row major order Apply Hough Split coefficients into transform on DF sub-bandsinto Apply Hough Splitfour coefficients Apply Hough coefficient Split coefficients AF, HF, VF andinto DF transformsub-band on DF four sub-bands transform on DF four sub-bands coefficient sub-band AF, HF, VF and DF coefficient sub-band AF, HF, VF and DF Embed two/ four bits of Apply Inverse Stego-image information per bits band Daubechies Embed two/ four of Apply Inverse (N x N) Stego-image Embed two/ four bits of Apply Inverse based on hash function Transform (IDT) Stego-image information per band Daubechies information per band Daubechies (N x N) based on hash function Transform (IDT) (N x N) Transform (IDT) based on hash function Figure 1. Schematic representation of encryption technique

Figure 1. Schematic representation of encryption technique Fig. 1 : Schematic representation of encryption 1. Schematic representation of encryption technique During Figure decoding the stego-image passes through FDT again technique rd

to generate fourthefrequency sub-bands. From FDT every 3 During decoding passes through During decoding the stego-image stego-image passespasses through FDT again again rd During of decoding thetwo/four stego-image through FDT coefficients each frequency band secret bitsFrom are extracted. to generate four sub-bands. every 33In rd to generate four frequency sub-bands. From every again to generate four frequency sub-bands. From every 3rd total eight/sixteen bits oftwo/four information per mask, is extracted coefficients of each band secret bits are extracted. In coefficients of each each band band two/four two/foursecret secretbits bitsare areextracted. extracted. In In coefficients of and compared withbits DFofcoefficient generated signature bits, total eight/sixteen information per mask, extracted total eight/sixteen bits of information per per mask, mask, isis is extracted extracted total eight/sixteen bits of information computed through Hough transform at signature receiver bits, for and compared with DF coefficient generated and coefficient generated signature bits, and compared comparedwith withDFDF coefficient generated authentication. The schematic representation is signature given for in computed through Hough transform at receiver computed through Hough transform for bits, computed through Hough transformatat receiver receiver for figure 2. authentication. The schematic representation is given in authentication. representation is in authentication. The The schematic schematic representation is given in figure 2. Fig. 2.2. figure Forward Stego-Image Daubechies Forward Forward Transform (FDT) Stego-Image Daubechies Stego-Image Daubechies Transform (FDT) Transform (FDT) Extract secret bits from frequency coefficients Extract secret bits from Extract secret bits from based on hash function frequency coefficients frequency coefficients based on hash function based on hash function

(1) (1) (1) (2) S4 : (2) 85 (2) where h[n] is a sequence of low-pass impulse response filter coefficients and g[n] is of a low-pass sequence impulse of high-pass impulse where h[n] is a sequence response filter where low-pass impulse impulse response response filter filter where h[n] h[n] is is aa sequence sequence of of low-pass response filter coefficients, and the coefficients are: impulse coefficients and g[n] is a sequence of high-pass coefficients is aa sequence sequence of of high-pass high-pass impulse impulse coefficients and and g[n] g[n] is response filter coefficients, and the coefficients are: response responsefilter filtercoefficients, coefficients,and andthe thecoefficients coefficientsare: are: h[0] = (1+SQRT (3))/ (4*SQRT (2)), h[0] = (1+SQRT (3))/ (4*SQRT (2)), h[1] == (1+SQRT (3+SQRT (3))/ (3))/ (4*SQRT (4*SQRT (2)), (2)), h[0] h[0] == (1+SQRT (3))/(4*SQRT (4*SQRT(2)), (2)), h[1] = (3+SQRT (3))/ h[2] = (3+SQRT (3-SQRT (3))/ (4*SQRT (2)), h[1] (3))/ (4*SQRT (2)), h[1] (3))/ (4*SQRT(2)), (2)), h[2] = = (3+SQRT (3-SQRT h[3] = (1-SQRT (3))/ (3))/(4*SQRT (4*SQRT (2)), (2)), h[2] = (3-SQRT (3))/ (4*SQRT h[2] = (3-SQRT (3))/ (4*SQRT (2)), h[3] = = (1-SQRT (4*SQRT g[0] h[3], g[1](3))/ = -h[2], g[2] =(2)), h[1] and g[3]= -h[0]. h[3] h[3] = = (1-SQRT (1-SQRT (3))/ (3))/ (4*SQRT (4*SQRT (2)), (2)), g[0] = = h[3], h[3], g[1] ==-h[2], g[2] = =h[1] and g[3]= -h[0]. g[0] g[1] -h[2], g[2] h[1] g[3]= g[0]On = h[3], g[1] = -h[2], g[2]Forward = h[1] and and g[3]= -h[0]. -h[0]. calculation of the Daubechies Transform On calculation of the Forward Daubechies Transform (FDT), the coefficients generated are grouped into four subOn calculation of the Forward Daubechies Transform (FDT), the coefficients generated are Daubechies grouped intoTransform four subOn calculation of the Forward bands, as shown in figure 3 and figure 4 respectively. (FDT), the coefficients generated are grouped into four bands, as in figure 3 and figure 4 respectively. (FDT), theshown coefficients generated are grouped into four subsubbands, as shown in figure 3 and figure 4 respectively. bands, as shown in figure 3 and figure 4 respectively. Average / Lower Average / Lower Frequency (AF)

Horizontal / middle Horizontal middle frequency/ (HF)

Vertical / middle frequency (VF)

Diagonal / Higher Frequency (DF)

Frequency (AF) Average Average // Lower Lower Frequency Frequency (AF) (AF)

Vertical / middle frequency (VF) Vertical Vertical // middle middle frequency frequency (VF) (VF)

frequency (HF) Horizontal Horizontal // middle middle frequency frequency (HF) (HF) Diagonal / Higher Frequency (DF) Diagonal Diagonal // Higher Higher Frequency Frequency (DF) (DF)

Fig. 3 : Daubechies Coefficient in separate bands after FDT.in separate bands after FDT. Figure 3. Daubechies Coefficient Figure 3. Daubechies Coefficient in separate bands after FDT. Figure 3. Daubechies Coefficient in separate bands after FDT.

Separate coefficients AF, HF, coefficients VF and DF AF, Separate Separate coefficients AF, HF, VF and DF HF, VF and DF

Apply Hough transform on DF coefficient sub-bandon to Apply Hough transform Apply Hough transform on generate secret signature DF coefficient sub-band to DF coefficient sub-band to generate secret signature generate secret signature Generate secret bits stream Generate secret bits stream Generate secret bits stream

Compare No Compare No Compare No Authentic . Unauthentic Yes Authentic . Unauthentic Authentic . Unauthentic Yes Yes Figure 2. Schematic representation of authentication technique 2. Schematicrepresentation representation of authentication technique Fig.Figure 2 : Schematic of authentication Figure 2. Schematic representation of authentication technique 2 technique

III. DISCRETE DAUBECHIES TRANSFORM TECHNIQUE III. Discrete Daubechies Transform Technique

Every transformation equation exists with a pair. Everywavelet transformation exists with Daubechies has no explicitequation function expression. The a pair. Daubechies wavelet has no explicit function expression. The scaling functions and wavelet functions are defined by the scaling (1) functions and wavelet functions are defined by the equations and (2) respectively.

equations (1) and (2) respectively.

(1)

(2) where h[n] is a sequence of low-pass impulse response filter coefficients and g[n] is a sequence of high-pass impulse response filter coefficients, and the coefficients are: h[0] = (1+SQRT (3))/ (4*SQRT (2)), h[1] = (3+SQRT (3))/ (4*SQRT (2)),

Figure Representation of subbands after FDT on Baboon image Fig. 4 4. : Representation of subbands after FDT on Figure 4. Representation of subbands after FDT on Baboon image Baboon image Figure 4. Representation of subbands after FDT on Baboon image

Inverse Daubechies transform IDT is a similar operation as FDT, where frequency coefficients are transformed to spatial domain to generate stego-image on embedding.

IV. Secret Message Computation through Hough Signature Significant image orientation features can be identified and extracted through Hough transform by analyzing lines in an image [10]. Hough transform maps the individual pixels from image domain into a shape in parameter domain [11] that’s creates a butterfly signature. Forward Daubechies

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013

Significant Significant image image orientation orientationfeatures featurescan canbe beidentified identifiedand and (a) Horizontal (b) Vertical coefficients (c) Diagonal (a) Horizontal (b) Vertical coefficients (c) Diagonal extracted through Hough transform by analyzing lines in an after FDT after FDT coefficients after FDT extracted through Hough transform by analyzing lines in an coefficients coefficients after FDT after FDT coefficients after FDT (HF sub-image) (VF sub-image) (DF sub-image) image [10]. Hough transform maps the individual pixels from (HF sub-image) (VF sub-image) (DF sub-image) image [10]. Hough transform maps the individual pixels from image domain into a shape in parameter domain [11] that‟s image domain into a shape in parameter domain [11] that‟s creates a butterfly signature. Forward Daubechies DF Hough Signature based Authentication of image through creates butterfly signature. Forward Daubechies DF S4 : 86 aCSI Daubechies Transform technique (HSADT) 3 JOURNAL VOL. transform XX, NO. XX, with XXXXXXXXX coefficients are passed COMPUTING, through Hough a range coefficients are passed through Hough transform with a range of 0 to 100 pixel value, as selected for proposed work, of 0 to 100 pixel value, as selected for proposed work, dependent onInverse hash function. Daubechies transform IDT is a similar operation as DF coefficients are passed through Hough transform with a dependent on hash function. Grayscale diagonal coefficient sub-image on passing FDT, where frequency coefficients are transformed to spatial range of 0 to 100 pixel value, as selected for proposed work, Grayscale diagonal coefficient sub-image on passing through Hough transformation with range, andcoefficient origin as domain to generate embedding. dependent on hash function.stego-image Grayscale diagonal through Hough transformation with on range, and origin as (d) Threshold image of (e) Threshold image of (f) Threshold image of center of image, generates a matrix of rho verses theta, by eq (d)HF Threshold image of (e)VF Threshold image of (f)DF Threshold image of sub-image on passing through Hough sub-image sub-image sub-image center of image, generates a matrix of transformation rho verses theta,with by eq HF sub-image VF sub-image DF sub-image 3range, and eq 4 where limits are 1≤ theta( θ ) < Pi and –N ≤ rho( IV. S ECRET M ESSAGE C OMPUTATION THROUGH H OUGH and origin as center of image, generates a matrix of 3 )and eq 4 where limits are 1≤ theta( θ ) < Pi and –N ≤ rho( SIGNATURE ρrho ≤ N. verses theta, by eq 3 and eq 4 where limits are 1≤ theta(q) ρ ) ≤ N.

< Pi and –N ≤Significant rho(r) ≤ N. image orientation features can be identified and (a) Horizontal (b) Vertical coefficients (c) Diagonal extracted through Hough transform by analyzing lines in an coefficients after FDT after FDT coefficients after FDT (3) (HF sub-image) (VF sub-image) (DF sub-image) image [10]. Hough transform maps the individual (3) pixels from image domain into a shape in parameter domain [11] that‟s (4) creates a butterfly signature. Forward Daubechies DF (4) (g) Hough signature of (h) Hough signature of (i) Hough signature of coefficients are passed through Hough transform with a range (g)HF Hough signature of (h)VF Hough signature of (i)DF Hough signature of sub-image sub-image sub-image of 0 to of 100ρ (rho) pixel value, as value selected proposed On calculation for each of for θ (theta) and work, HF sub-image VF sub-image DF sub-image (ii) Couple image On ofof (rho) forofeach each valueθofof θ (theta) and Oncalculation calculation (rho) for value (theta) and dependent onρrhash function. incrementing the matrix value ρ verses aq butterfly like (ii) Couple image incrementing the matrix value of ρ verses θ a butterfly like incrementing the matrix value of r verses q a butterfly like Grayscale diagonal coefficient sub-image on passing unique structure is generated which is termed as butterfly throughas Hough transformation with unique structure isis generated which is istermed as as butterfly structure generated which termed butterfly signature treated secret image as shown in range, figure 5and fororigin all as (d) Threshold image of (e) Threshold image of (f) Threshold image of center of image, generates a matrix of rho verses theta, signature treated as secret image as shown in figure 5 for all signature treated as secret image as shown in figure 5 for allby eq HF sub-image VF sub-image DF sub-image the three sub-bands. 3 and eq 4 where limits are 1≤ theta( θ ) < Pi and –N ≤ rho( the the three three sub-bands. sub-bands. ρ ) ≤ N.

(3) (a) Horizontal

(b) Vertical coefficients (c) Diagonal after FDT after FDT (b) Vertical coefficients coefficients (c) Diagonal Hough signature (h) Hough signature of (i) Hough signature of (HF(g)sub-image) sub-image) (DF sub-image) coefficients after FDT of (VF after FDT coefficients after FDT HF sub-image VF sub-image DF sub-image (VF sub-image) (DF sub-image) On calculation of ρ (rho) for each value of θ (theta) and (HF sub-image) (ii) Couple image (a) Horizontal (b) Vertical coefficients (c) Diagonal incrementing the matrix value of ρ verses θ a butterfly like coefficients after FDT after FDT coefficients after FDT (a) Horizontal (b) Vertical coefficients (c) Diagonal unique structure is generated which(DFissub-image) termed as butterfly (HF sub-image) (VFafter sub-image) coefficients after FDT FDT coefficients after FDT signature treated as secret image as shown in figure 5 for all (HF sub-image) (VF sub-image) (DF sub-image)

(4) coefficients after FDT (a) Horizontal

the three sub-bands.

(d) Threshold image of (e) Threshold image of (f) Threshold image of sub-image sub-image sub-image (d)HF Threshold image of (e)VF Threshold image of (f)DF Threshold image of HF (a) sub-image sub-image DF sub-image Horizontal (b)VF Vertical coefficients (c) Diagonal (d) Threshold image of (e) Threshold image of (f) Threshold image of HF sub-image VF sub-image sub-image (d) Threshold image of (e) Threshold image of (f) DF Threshold image of (a) Horizontal VF sub-image (b) Vertical coefficientsDF sub-image (c) Diagonal HF sub-image coefficients after FDT (HF sub-image)

after FDT (VF sub-image)

coefficients after FDT (HF sub-image)

after FDT (VF sub-image)

coefficients after FDT (DF sub-image)

coefficients after FDT (DF sub-image)

(g) Hough signature of (h) Hough signature of (i) Hough signature of HF sub-image VF sub-image DF sub-image (g) Hough signature of (h) Hough signature of (i) Hough signature of (iii) Image HF sub-image VFMap sub-image DF sub-image (h) Hough signature of (g) Hough signature of (i) Hough signature of HF sub-image VF sub-image DF sub-image (g) Hough signature of (h) Hough signature of (i) Hough signature of (i) Baboon Image (d) Threshold image of (e) Threshold image of Threshold image of HF sub-image VF sub-image DF (f) sub-image HF sub-image (i) BaboonVF sub-image DF sub-image Image

(d) Threshold image of HF sub-image

(e) Threshold image of VF Map sub-image (iii) Image

(f) Threshold image of DF sub-image

(h) Hough signature of VF sub-image

(i) Hough signature of DF sub-image

Figure 5. Different sub-bands images of various coefficients with respective threshold image and their signature generated through Hough transform for Figure 5. DifferentBaboon, sub-bands images various coefficients with respective Couple andofMap image. threshold image and their signature generated through Hough transform for Baboon, Couple and Map image.

(g) Hough signature of HF sub-image

(iii) Map Image (g) Hough signature of HF sub-image

(h) Hough signature of VF sub-image (i) Baboon Image

(i) Hough signature of DF sub-image

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013

Fig. 5 : Different sub-bands images of various Figure 5. with Different sub-bands images of various coefficients with respective coefficients respective threshold image and their threshold image and their signature generated through Hough transform for signature generated through Hough transform for Baboon, Couple and Map image. Baboon, Couple and Map image.

4

CSI JOURNAL COMPUTING, VOL. XX, NO. XX, XXXXXXXXX

4 : 87 S4

CSI JOURNAL COMPUTING, VOL. XX, NO. XX, XXXXXXXXX

Madhumita et.AND al. AUTHENTICATION V. Sengupta, EMBEDDING CSI JOURNAL COMPUTING, VOL. XX, NO. XX, XXXXXXXXX

Embedding done for authentication and authentication V. isEMBEDDING AND AUTHENTICATION Embedding and Authentication isV.done at destination end to verify the originality. Discrete EMBEDDING AND AUTHENTICATION Embedding V. is done for authentication and authentication Daubechies Transform performs as preprocess of embedding Embedding is done for authentication and authentication is done Embedding at destination end to the originality. Discrete rd is done forverify authentication and authentication inisDaubechies transform domain. end In tothis process of of embedding 3 done at destination verify the originality. Discrete performs preprocess embedding is done Transform at destination end toasverify the originality. Discrete (a) Original Image rd Daubechies Transform performs as preprocess of embedding coefficient of every subband is used so that no two in transform In performs this process of embedding 3 Boat Daubechiesdomain. Transform as preprocess of embedding incoefficient transform In this is process ofso embedding 3rd rd(a) Original Image consecutive positions aresubband tamper by secret bits. The positions ofdomain. every used that no two in transform domain. In this process of embedding 3 Boat coefficient every subband issingle used thatare no two consecutive (a) Original Image selected forofembedding in tamper asubband mask Pso consecutive positions are byso secret bits. The positions 10, P 12, P30 coefficient of every is used that no and two Boat positions are tamper by secret bits. The positions selected for P32selected as consecutive shown figure 6 in with different for in embedding a single mask are P10bits. , P12The , P30positions and positions are tamper bycolor. secret embedding in a single mask are P , P , P and P as shown 10 diagonal 12mask 30 are PFor shown for in figure 6 withindifferent color. experimental purpose only Hough selected embedding a single P 32, Psignature 32 as 12, P30 and in figure 6 with different color.only diagonal Hough10signature For purpose P32 experimental as shown in figure 6 with different of various dimensions are used. On color. embedding Hough experimental purpose only diagonal Hough of For various dimensions are used.only Ondiagonal For experimental purpose Hough signature signature generated by DF sub-image 8 embedding bit per 16 Hough bytes is (a) Original Image signature of various dimensions are 8used. On16 embedding signature generated by DF sub-image bit per bytes isbits (a) OriginalClock of various dimensions are used. On embedding Hough Image needed and the payload became 0.5 bpB. Embedding 12 Hough signature generated by sub-image DF bpB. sub-image 8per bit16 per 16 is Clock needed and thegenerated payload became 0.5 Embedding 12 bits signature by DF 8 bit0.7 bytes (a) Original Image per mask or 16 bits per mask payload becomes to 1.0 bpB. bytes is needed the payload became 0.5 bpB. Embedding per mask or and 16 and bits mask payload becomes 0.7 to 1.0 bpB. Clock needed the per payload became 0.5 bpB. Embedding 12 bits To obtain the payload of 0.7 bpB dimension of signature 12 per mask or 16 bits per mask payload becomes Tobits obtain the payload of 0.7 bpB dimension of signature per mask or 16 bits per mask payload becomes 0.7 to 1.00.7 bpB. should bebe151 and that of for for0.7 1.0bpB bpB thesize size toshould 1.0To bpB. To xobtain the of dimension 151 x 151 151 and payload that of 1.0 bpB the ofofof obtain the payload of 0.7 bpB dimension of signature signature should be 181 x 181 for the cover image signature should be 151 x 151 and that of for 1.0 bpB the signature x 181 coverbpB image of ofof shouldshould be 151bex 181 151 and thatforof the for 1.0 the size size of signature should be 181 x 181 for the cover image dimension 512 x 512. dimension 512 should x 512. be 181 x 181 for the cover imageof of signature dimension 512 x512 512. dimension x 512.

PP0000 P01 PP0202 PP0303 HFHF AF AF Band AF PP1010 P00P11 P01PP12 Band Band HF 1313 P03Band 12 P02PP Band P P P P 10P 11P 12PP 13 DF Band P P VF 20 21 22 23 P P VF DF 20 21 22 23 Band VF PP30 P20P P31 DF 31 P21P Band Band P3232 P22PP3333 P23Band 30 Band Band P P P P 30 31 32 33 Figure 6. Single mask representation of four subbands 6. mask Single mask representation ofof four subbands Fig. 6 :Figure Single representation four subbands Figure 6. Single mask representation of four subbands

(c) Stego-Image Boat

(f) Extracted Secret

(b) secret

(c) Stego-Image

(f) Extracted

(b) secret signature (c) Stego-Image Clock (f) Extracted Secret signature Clock Secret (b) secret (c) Stego-Image (f) Extracted signature Clock Secret

(a) Original Image (b) secret (c) Stego-Image (f) Extracted (a)Elaine Original Image signature (b) secret (c) Stego-Image Secret (f) Extracted Elaine (a) Original Image (b) secret (c) Stego-Image (f) Extracted Elaine signature Elaine Secret Elaine signature Elaine Secret

(b) secret (c) Stego-Image (f) Extracted signature Jet Secret (f) Extracted (b) secret (c) Stego-Image (b) secret (c) Stego-Image (f) Extracted signature Secret signature JetJet Secret

(b) secret (c) Stego-Image (f) Extracted signature Map Secret (b) secret (c) Stego-Image (f) Extracted (b) secret (c) Map Stego-Image (f) Extracted signature Secret

signature

Map

Secret

(b) secret (c) Stego-Image (f) Extracted signature Space Secret (b) secret (c) Stego-Image (f) Extracted Space signature Space Secret (a) Original Image (b) secret (c) Stego-Image (f) Extracted

at receiver side without attack.

Space

(a) Original Image (b) secret (c) Stego-Image (f) Extracted Baboon signature Baboon Secret (a) Original Image (b) secret (c) Stego-Image (f) Extracted Baboon signature Baboon Secret

(b) secret signature

(b) secret signature

(b) secret (c) Stego-Image (f) Extracted signature Boat Secret (b) secret (c) Stego-Image (f) Extracted signature Boat Secret

(a) Original (b) secret (c) Stego-Image (c) Stego-Image (f) Extracted (a) Original ImageImage (b) secret (f) Extracted Couple signature Secret Couple signature Couple Couple Secret (a) Original Image (b) secret (c) Stego-Image (f) Extracted Couple signature Couple Secret

Fabrication isis done through onon Fabrication donethrough throughANDing/ORing ANDing/ORingoperation operation Fabrication is done ANDing/ORing operation on stream of signature with the source/cover bits based on hash Fabrication is done through ANDing/ORing operation on stream of signature with the source/cover bits based on hash stream of signature with the source/cover bits based on hash function. stream of signature with the source/cover bits based on hash function. function. function. VI. Discussions RESULTS AND DISCUSSIONS VI. Results and (a) Original Image VI. deals R ESULTS AND D ISCUSSIONS This section with the AND results of computation on VI. RESULTS DISCUSSIONS Jet This section deals with the results of computation on (a) Original Image Image embedding self generated signature. PGM [12] onon (a) Original This section dealsdeals withhidden the the results of Ten computation Thisself section with results of computation Jet Jet embedding generated hidden signature. Ten PGM [12] images have been taken andhidden HSADT issignature. appliedTen onTen each. All embedding self generated hidden PGM [12] embedding self generated signature. PGM [12] images have been taken and HSADT is applied on each. coverimages images arebeen of 512 x and 512 HSADT in dimension asonshown inAll have is applied each. images have been taken and is applied each. All cover images are oftaken 512 xHSADT 512 in dimension as on shown inAll figure 7 and the secret signature generated through Hough cover images are of 512 x 512 in dimension as shown cover are secret of 512signature x 512 in dimension as shown figureimages 7 and the generated through Houghinin transform of DF coefficient sub-image is of three different figure 7 and the secret signature generated through Hough figure 7 andof the signature generated through Hough transform DF secret coefficient sub-image is of three different dimensions 128ofx 128, 151 x 152, sub-image and 181 x 181. transform DF coefficient is of three different (a) Original Image transform of DF coefficient sub-image is of three different dimensions 128 x 128, 151 x 152, and 181 x 181. Map The visual impact image/ original dimensions 128 x over 128, the 151cover x 152, and 181 x 181.image on (a) Original Image dimensions 128 x 128, 151 x 152, and 181 x 181. The visual impact over the cover is image/ image embedding secret image/message shownoriginal in figure 7. on (a) Original Map Image The the visual impact over the cover image/ original image Map on embedding the secret is original shown in figure The visual impact over image/message the cover image/ image on Figure 7.(a). is the original image with figure 7.(b). the secret embedding the secret image/message is shown in figure 7. 7.Hough Figure transform 7.(a). secret is the original image with figure the7. embedding image/message is shown in7.(b). figure generated signature 181 x7.(b). 181 Figurethe 7.(a). is the original image with figure thein secret secret Hough transform generated signature 181 x 181 in Figure 7.(a). is the original image with figure 7.(b). secret dimension, figure 7.(c) stego-image after embedding Hough transform generated signature 181 the x secret 181 in dimension, figure 7.(c) stego-image after embedding secret Hough transform generated signature 181 x image 181 secret in image/message and figure is the extracted secret dimension, figure 7.(c)7.(d) stego-image after embedding image/message figure 7.(d) is the extracted secret image dimension, figure 7.(c) stego-image embedding at receiver side and without image/message andattack. figure 7.(d) is after the extracted secretsecret image(a) Original Image Space at receiver side without attack. (a) Original Image image/message and figure 7.(d) is the extracted secret image at receiver side without attack.

(a) Original Image Baboon

4

(c) Stego-Image Baboon

(f) Extracted Secret

signature

Space

Secret

(a) Original Image (b) secret (c) Stego-Image (f) Extracted Tank signature Tank Secret (a) Original Image (b) secret (c) Stego-Image (f) Extracted Tank signature Tank Secret

(a) Original Image Tank

(b) secret signature

(c) Stego-Image Tank

(f) Extracted Secret

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013

S4 : 88

CSI JOURNAL CSI JOURNAL COMPUTING, COMPUTING, VOL. XX, VOL. NO.XX, XX,NO. XXXXXXXXX XX, XXXXXXXXX

Hough Signature based Authentication of image through Daubechies Transform technique (HSADT) 5

5

5

CSI JOURNAL COMPUTING, VOL. XX, NO. XX, XXXXXXXXX

(a) Original (a) Original Image Image (b) secret(b) secret(c) Stego-Image (c) Stego-Image (f) Extracted (f) Extracted Truck Truck signaturesignature Truck Truck Secret Secret

(c) Extracted (c) Extracted secret from secret from (b) Attacked stego-image Figure 7.Figure Original 7. Original image, secret image, signature secret signature grneratedgrnerated through Hough throughtransform Hough transform (b) Attacked stego-image attacked attacked stego-image stego-image Fig. 7of: DF Original image, secret signature grnerated (a) Stego-image (a) Stego-image of Map of Map MSE 171.924549 MSE 171.924549 of DF coefficient coefficient sub-image, sub-image, Embedded/Stego-image Embedded/Stego-image and extracted and extracted secret from secret from MSE 534.647629 MSE 534.647629 PSNR 25.777425 PSNR 25.777425 by receiver. by receiver. through Hough transform of DF coefficient sub-image, (a) Original Image (b)stego-image secret stego-image (c) Stego-Image (f) Extracted PSNR 20.850127 PSNR 20.850127

Truck signature Truck Embedded/Stego-image and extracted secretSecret from Hit on Hit image on image while transmitting while transmitting over unsecure over unsecure networknetwork (c) secret from 8.Figure Original Original image, attacked image, attacked stego-image stego-image and the extracted and theExtracted extracted secret image secret image stego-image bygrnerated receiver. Figure 7. Original image, secret signature through Hough transform Figure Fig. (b)image, Attacked stego-image 8 8.: Original attacked stego-image and the after „encryption‟ after „encryption‟ at sender at sender side is side a very is common a and very common problem in in(a) Stego-image of Map from attacked from stego-image. stego-image. attacked stego-image MSEattacked 171.924549 of DF coefficient sub-image, Embedded/Stego-image extracted secretproblem from MSE 534.647629 extracted secret image from attacked stego-image. today‟stoday‟s hi-techhi-tech world.stego-image world. Hit canHit can of any be ofform any like formstamp, like stamp, PSNR 25.777425 bybe receiver. PSNR 20.850127 crop, crop, cutpaste and small paste portion small portion of image ofunsecure image over the over same the same Calculation Calculation of average of average MSE, PSNR MSE, and PSNR IF and for ten IF for images ten images Hitcut onand image while transmitting over network image image and many and more. many This more. type Thisoftype hitover isof commonly hit is commonly known known with two with bits twosecret bits per secret band per per band mask perPSNR have maskbeen haveIF done been to done to Hit on image while transmitting unsecure network Calculation of average MSE, and for ten images after ‘encryption’ at sender side is a very common problem in Figure 8. Original image, attacked stego-image and the extracted secret image as external as external attack at or attack sabotage or sabotage overisan over image. an image. problem in embed embed secret secret image image of from dimension of dimension 128per x 128 128. xThe 128. average The average after „encryption‟ sender side a very common attacked stego-image. with two bits secret per band mask have been done to today’s hi-tech world. Hit can be of any form like stamp, crop, HSADT HSADT is world. a robust is aHit robust one forbe oneauthentication/ownership foranyauthentication/ownership MSE for MSEembedded for embedded image image is 0.929503 is 0.929503 that ofthat PSNR of PSNR is is today‟s hi-tech can of form like stamp, cut and paste small portion of image over the same image and secret image of0.999953 dimension 128 x 128. The average MSE verification verification against against such attacks. such attacks. Theoftechnique The technique proposed proposed is 48.4654052 is embed 48.4654052 dB, IF of dB, is 0.999953 IF is and UQI and is UQI 0.999579 as ten given as given crop, cut and paste small portion image over the same Calculation average MSE, PSNR andisIF0.999579 for images many more. This type of hit is commonly known as external for embedded image is 0.929503 that of PSNR is 48.4654052 sustainable sustainable against against various various kinds kinds attack for example as for known example an inanwith tableinI. tablebits I. secret per band per mask have been done to image and many more. This typeof ofattack hitofisas commonly two attack or sabotage overMSE an image. image corrupted corrupted with with 4090.09 MSE PSNR and PSNR 12.01dB 12.01dB can canembed To calculate To the result theof for result higher forishigher payload, bit three secret bit secretI. dB, IF iscalculate 0.999953 and UQI 0.999579 as given in table asimage external attack or sabotage over 4090.09 anand image. secret image dimension 128payload, xthree 128. The average HSADT isauthenticated a robust one for authentication/ownership also be also authenticated be by using by HSADT using HSADT as shown as shown in figure in 8 figure 8 per band per per band mask per / mask four bit / four secret bit per secret band per per band mask per has mask also has also HSADT is a robust one for authentication/ownership MSE for image is 0.929503 of PSNR is bit To embedded calculate the result for higherthat payload, three verification against such attacks. Thestego-images. technique proposed isbeen embedded. with other with such other cases such of cases attacked of attacked stego-images. been embedded. To achieve To achieve the payload the payload of 0.7 bpB of 0.7 the bpB secret the secret verification against such attacks. The technique proposed is 48.4654052 dB, IF is 0.999953 and UQI is 0.999579 as given per mask / that four bit band perbpB mask sustainable against various kinds of attack as for example animagesecret image of 151 of x band 151 is xper 151 taken is and taken andof that tosecret achieve of toper achieve 1.0 bpB 1.0 sustainable against various kinds of attack as for example an in table I. image corrupted with MSE 4090.09 and PSNR 12.01dB candimension dimension of been authenticating of embedded. authenticating image image taken taken is 181 is x 181 181 181bpB has also To achieve the payload ofx0.7 image corrupted with MSE 4090.09 and PSNR 12.01dB can To calculate the result for higher payload, three bit secret through which an which increase anofincrease of 0.3 bpB of 0.3 hasbpB been hasand achieved. been achieved. Theto achieve The also be authenticated by using HSADT as shown in figure 8through secret image 151 151 is taken also be authenticated by using HSADT as shown in figure 8 results perthe band per / in four bitxtable per band per that maskofhas also results are given aremask ingiven table IItable and IIsecret andIIItable respectively. III respectively. with other such cases of attacked stego-images. 1.0 bpB dimension of authenticating image taken is 181 x 181 with other such cases of attacked stego-images. been To achieve the of 0.7 bpBsome the secret To embedded. increase To increase the payload the payload of payload 0.2of bpB 0.2 some bpB more more image of 151 x 151 is taken and that of to achieve 1.0 bpB through which an increase of 0.3 bpB has been achieved. degradation degradation within within permissible permissible limit islimit allowed is allowed which is which 1.89 is 1.89The ofofgiven authenticating image taken 181 xFurther 181 in dimension terms in of terms MSE and MSE-4.80 dB -4.80 dB terms in of terms PSNR. ofis PSNR. Further results are inand table IIinand table III respectively. through which anpayload increase of 0.3 bpB has been achieved. Theof increases increases in payload in form 0.7 form bpB 0.7 to bpB 1.0 to bpB 1.0 (increases bpB (increases of (c) Extracted (c) Extracted secret from secret from To increase the payload of 0.2 III bpB some more degradation (b) Attacked (b) Attacked stego-image stego-image results in table II and table payload payload byare 0.3given by bpB0.3 more) bpB MSE more) increases MSE increases byrespectively. 8.24byand 8.24 that and of that of (a) Stego-image (a) Stego-image of of attacked attacked stego-image stego-image MSE 65.408474 MSE 65.408474 within permissible limit is allowed which terms of Baboon Baboon MSE 83.694336 MSE 83.694336PSNR To increase the payload of 0.2 bpBis 1.89 somein more decreases PSNR decreases by 5.95 by dB. 5.95 dB. PSNR 29.974463 PSNR 29.974463 PSNR 28.903843 PSNR 28.903843 degradation permissible is allowed whichincreases is 1.89 in MSE andwithin -4.80 dB in termslimit of PSNR. Further TABLE I. TABLE DATA I. ONDAPPLYING ATA ON APPLYING HSADT HSADT OVER 10 OVER IMAGES 10 WITH 0.5WITH BPB 0.5 BPB in payload terms ofform MSE0.7 and dB in terms ofIMAGES PSNR. Further bpB-4.80 to 1.0 bpB (increases of payload by 0.3 Cover Image Cover increases in Image payload form 0.7 PSNR bpB to 1.0 IF bpB (increases of MSE MSE PSNR IF UQI UQI (c) Extracted secret from bpB more) MSE increases by 8.24 and that of PSNR decreases 512 x 512 512 x 512 (b) Attacked stego-image payload by 0.3 bpB more) MSE increases by 8.24 and that of attacked stego-image (a) Stego-image of (a) Baboon (a) Baboon 0.919235 48.496537 48.496537 0.9999510.999951 0.9997560.999756 MSE 65.408474 by 5.95 dB.0.919235 MSE 83.694336 Baboon PSNR decreases by 5.95 dB. (b) Boat (b) Boat 0.9920850.992085 48.165317 48.165317 0.9999480.999948 0.9997800.999780 (c) Clock(c) Clock 0.7481120.748112 49.391139 49.391139 0.9999800.999980 0.9998900.999890 Table I : Data on applying HSADT over 100.5images (d) Couple (d) 48.106090 48.106090 0.999938 0.999622 0.999622 TABLE I. Couple D1.005707 ATA ON1.005707 APPLYING HSADT OVER 100.999938 IMAGES WITH BPB (e) Elaine(e) Elaine 0.9563370.956337 48.324694 48.324694 0.999954 0.999954 0.999781 0.999781 with 0.5 BPB (c) Extracted (c) Extracted secret from secret from Cover Image (f) Jet (f) Jet 0.8869860.886986 48.651637 48.651637 0.9999720.999972 0.9991310.999131 (b) Attacked (b) Attacked stego-image stego-image MSE PSNR IF UQI (a) Stego-image (a) Stego-image of of attacked attacked stego-image stego-image (g) 512 Mapx(g) Map 0.9119720.911972 48.530988 48.530988 0.9999730.999973 0.9997120.999712 MSE 4090.092205 MSE 4090.092205 512 Baboon Baboon MSE 942.828369 MSE 942.828369 Cover Image MSE PSNR IF (h) Space (h) Space 0.855579 0.855579 48.808201 48.808201 0.999950 0.999950 0.999448 0.999448 PSNR 12.013473 PSNR 12.013473 (a) Baboon 0.919235 48.496537 0.999951 0.999756 UQI PSNR 18.386477 PSNR 18.386477 (i) Tank Tank 1.012505 1.012505 48.076834 48.076834 0.999944 0.999944 0.999334 0.999334 512 x(i)512 (b) Boat 0.992085 48.165317 0.999948 0.999780 (j) Truck (j) Truck 1.006512 1.006512 48.102615 48.102615 0.999917 0.999917 0.999338 0.999338 (c) Clock 0.748112 49.391139 0.999980 0.999890 (a) Baboon 0.919235 48.496537 0.999951 0.999756 Average Average 0.929503 0.929503 48.4654052 48.4654052 0.999953 0.999953 0.999579 0.999579 (d) Couple 1.005707 48.106090 0.999938 0.999622

PSNR 29.974463

PSNR 28.903843

Elaine 0.956337 0.999781 (b) (e) Boat 0.992085 48.324694 48.165317 0.999954 0.999948 0.999780 TABLE TABLE II. OND APPLYING ATA ON APPLYING HSADT HSADT OVER 10 OVER IMAGES 10 WITH IMAGES 0.7 WITH BPB 0.7 BPB (f)II. Jet DATA 0.886986 48.651637 0.999972 0.999131 (c) (g) Clock 0.748112 48.530988 49.391139 0.999973 0.999980 0.999890 Map 0.911972 0.999712 Cover Image Cover Image MSE MSEPSNR PSNR IF0.999950 IF UQI0.999448 UQI (h) Space 48.808201 512 x 512 512 x 512 0.855579 (d) Couple 1.005707 48.106090 0.999938 0.999622 (i) Tank 1.012505 48.076834 0.999944 0.999334 (a) Baboon (a) Baboon 3.211468 3.211468 43.063768 43.063768 0.999827 0.999827 0.999146 0.999146 (j) Truck 1.006512 48.102615 0.999917 0.999338 (e) Elaine 0.956337 48.324694 0.999954 0.999781 3.1504593.150459 43.147065 43.147065 0.9998340.999834 0.9992920.999292 (c) Extracted (c) Extracted secret from secret from (b) Boat (b) Boat (b) Attacked (b) Attacked stego-image stego-image Average 0.929503 48.4654052 0.999953 0.999579 Clock (c) Clock 2.145336 2.145336 44.815850 44.815850 0.999943 0.999943 0.999684 0.999684 (a) Stego-image (a) Stego-image of of attacked attacked stego-image stego-image (c) (f) Jet 0.886986 48.651637 0.999972 0.999131 MSE 49.579098 MSE 49.579098 (d) Couple (d) Couple 2.8648072.864807 43.559850 43.559850 0.9998240.999824 0.9989190.998919 Space Space MSE 371.585205 MSE 371.585205 PSNR 31.177817 PSNR 31.177817 TABLE II. Elaine D3.039127 ATA ON3.039127 APPLYING HSADT OVER 100.999853 IMAGES WITH 0.70.999712 BPB (e) 43.303315 43.303315 0.999853 0.999295 0.999295 PSNR 22.430219 PSNR 22.430219 (g)Elaine Map(e) 0.911972 48.530988 0.999973 (f) Jet (f) Jet 2.2087862.208786 44.689267 44.689267 0.9999290.999929 0.9978550.997855 Cover Image (g) (g) Map 3.014938 3.014938 43.338019 43.338019 0.9999120.999912 0.9990420.999042 (h)Map Space 0.855579 48.808201 MSE PSNR IF0.999950 UQI 0.999448 x(h) 512 (h) 512 Space Space 2.4801372.480137 44.186047 44.186047 0.9998540.999854 0.9983780.998378 (i) Tank 1.012505 48.076834 0.999944 0.999334 (a) Baboon 3.211468 43.063768 0.999827 0.999146 (i) Tank (i) Tank 3.125103 3.125103 43.182160 43.182160 0.999828 0.999828 0.997912 0.997912 (j) Truck (j) Truck 2.998535 2.998535 43.361712 43.361712 0.999753 0.999753 0.997998 0.997998 (b) Boat 3.150459 43.147065 0.999834 0.999292 (c) Extracted secret from (j) Truck 1.006512 48.102615 0.999917 0.999338 (b) Attacked stego-image (c) Clock 2.145336 44.815850 0.999943 0.999684 Average Average 2.8238696 2.8238696 43.6647053 43.6647053 0.999856 0.999856 0.998752 0.998752 (a) Stego-image of attacked stego-image MSE 49.579098 (d) Couple 2.864807 43.559850 0.999824 0.998919 Average 0.929503 48.4654052 0.999953 0.999579 Space MSE 371.585205 (a) Stego-image of Baboon

(b) Attacked stego-image MSE 4090.092205 PSNR 12.013473

PSNR 31.177817

(c) Extracted secret from attacked stego-image MSE 942.828369 PSNR 18.386477

PSNR 22.430219

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013

(e) Elaine (f) Jet (g) Map (h) Space (i) Tank (j) Truck Average

3.039127 2.208786 3.014938 2.480137 3.125103 2.998535 2.8238696

43.303315 44.689267 43.338019 44.186047 43.182160 43.361712 43.6647053

0.999853 0.999929 0.999912 0.999854 0.999828 0.999753 0.999856

0.999295 0.997855 0.999042 0.998378 0.997912 0.997998 0.998752

S4 : 89

Madhumita Sengupta, et. al. Table II : Data on applying HSADT over 10 images with 0.7 BPB Cover Image 512 x 512

MSE

PSNR

IF

UQI

(a) Baboon

3.211468

43.063768

0.999827

0.999146

(b) Boat

3.150459

43.147065

0.999834

0.999292

(c) Clock

2.145336

44.815850

0.999943

0.999684

(d) Couple

2.864807

43.559850

0.999824

0.998919

(e) Elaine

3.039127

43.303315

0.999853

0.999295

(f) Jet

2.208786

44.689267

0.999929

0.997855

(g) Map

3.014938

43.338019

0.999912

0.999042

(h) Space

2.480137

44.186047

0.999854

0.998378

(i) Tank

3.125103

43.182160

0.999828

0.997912

(j) Truck

2.998535

43.361712

0.999753

0.997998

Average

2.8238696

43.6647053

0.999856

0.998752

Table III : Data on applying HSADT over 10 images with 1.0 BPB Cover Image 512 x 512 (a) Baboon

MSE

PSNR

IF

UQI

11.418964

37.554536

0.999386

0.996956

(b) Boat

10.459854

37.935547

0.999448

0.997655

(c) Clock

10.389271

37.964953

0.999725

0.998496

(d) Couple

11.316303

37.593758

0.999305

0.995858

(e) Elaine

10.169678

38.057732

0.999507

0.997638

(f) Jet

8.677208

38.747004

0.999721

0.991871

(g) Map

11.648750

37.468010

0.999658

0.996286

(h) Space

12.152859

37.284019

0.999285

0.992038

(i) Tank

11.426361

37.551724

0.999372

0.992499

(j) Truck

12.951977

37.007443

0.998934

0.991697

Average

11.061123

37.7164726

0.999434

0.995099

A comparative study has also been made between SADCT [5], SCDFT [3], Li method [1], Region–Based [2], WTSIC [13], AWTDHDS [7], IAHTSSDCT [6], SADT [14], ATFDD [9] and SAWT [4] in terms of mean square error, peak signal to noise ratio and payload (bits per Byte). Comparison is done on average of ten PGM images of figure 7, and the computation is given in table IV.

Table IV : Comparison of HSADT with Existing Technique Technique SADCT [5]

Capacity (bytes)

Size of cover image

bpB (bits per Bytes)

PSNR in dB

8192

512 * 512

0.08

56.63

SCDFT [3]

3840

512 * 512

0.12

30.10

Li’s Method [1]

1089

257 * 257

0.13

28.68

Region-Based [2]

16384

512 * 512

0.5

40.79

WTSIC [13]

16384

512 * 512

0.5

42.25

AWTDHDS [7]

16384

512 * 512

0.5

44.87

IAHTSSDCT [6]

16384

512 * 512

0.5

47.48

SADT [14]

16384

512 * 512

0.5

49.69

ATFDD [9]

16384

512 * 512

0.5

49.84

ATFDD [9]

32768

512 * 512

1.0

36.70

SADT [14]

32768

512 * 512

1.0

46.36

SAWT [4]

131072

512 * 512

1.3

36.62

HSADT

16384

512 * 512

0.5

48.46

HSADT

22952

512 * 512

0.7

43.66

HSADT

32761

512 * 512

1.0

37.72

From table IV it is clear that the proposed HSADT is performing better than the existing techniques and compatible with SADT. The computational results while comparing with SADT, HSADT technique seems fragile but HSADT technique is having the features of robust authentication after a forceful sabotage or attack.

VII. Conclusion This paper addressed the issue of image authentication and ownership verification be embedding different dimension of secret signature generated by Hough transform. Authenticity is incorporated by embedding secret signature in each frequency mask of carrier image generated by forward Daubechies transform in randomly generated position based on hash function with key. This Daubechies based authentication technique doesn’t needed any preprocessing or separate adjustment, thus reduces the cost of computation and complexity.

Acknowledgment The authors express deep sense of gratuity towards the

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013

S4 : 90

Hough Signature based Authentication of image through Daubechies Transform technique (HSADT)

Dept of CSE University of Kalyani where the computational resources are used for the work and the PURSE scheme of DST, Govt. of India. References [1] Li Yuancheng, Xiaolei Wang, “A watermarking method combined with Radon transformand 2D-wavelet transform”, IEEE, Proceedings of the 7th World Congress on Intelligent Control and Automation, June 25 - 27, Chongqing, China, 2008. [2] A. Nikolaidis, I. Pitas, “Region-Based Image Watermarking”, IEEE Transactions on Image Processing, Vol. 10, NO. 11, pp. 1721-1740, November 2001. [3] T. T. Tsui, X. –P. Zhang, and D. Androutsos, Color Image Watermarking Usimg Multidimensional Fourier Transfomation, IEEE Trans. on Info. Forensics and Security, vol. 3, no. 1, pp. 1628, 2008. [4] M. Sengupta, J.K. Mandal, “Self Authentication of Color image through Wavelet Transformation Technique (SAWT)”, pp- 151154, ISBN 93-80813-01-5, ICCS 2010. [5] M. Sengupta, J. K. Mandal, “Self Authentication of Color Images through Discrete Cosine Transformation (SADCT)”, IEEE catalog no : CFP1122P-CDR, ICRTIT, Anna University, Chennai, ISBN No-: 978-1-4577-0589-2, 3rd-5th June 2011. [6] M. Sengupta, J. K. Mandal “Image Authentication using Hough Transform generated Self Signature in DCT based Frequency Domain (IAHTSSDCT)”, IEEE, ISED- 2011, Kochi, Kerala, pp324-328, DOI 10.1109/ISED.2011.43, 2011. [7] M. Sengupta, J. K. Mandal, “Authentication in Wavelet Transform Domain through Hough Domain Signature (AWTDHDS)” UGC-Sponsored National Symposium on Emerging Trends in Computer Science (ETCS 2012), Barrackpore, ISBN number 978-81-921808-2-3, pp 61-65, 2012. [8] M. Kutter , F. A. P. Petitcolas, A fair benchmark for image watermarking systems, Electronic Imaging ‘99. Security and

Watermarking of Multimedia Contents, vol. 3657, Sans Jose, CA, USA, January 1999. The International Society for Optical Engineering, http://www.petitcolas.net/fabien/publications/ ei99-benchmark.pdf. (Last accessed on 25th March, 2012). [9] M. Sengupta, J. K. Mandal, “An Authentication Technique in Frequency Domain through Daubechies Transformation (ATFDD)”, International Journal of Advanced Research in Computer Science (IJARCS), Volume 3 No. 4 (July-August 2012), www.ijarcs.info, pp 236-242, ISSN No. 0976-5697. [10] V.F. Leavers, “Shape detection in computer vision using the hough transformation”, Springer-Verlag, Berlin, 1992, 201 pages, ISBN- 3-540-19723-0, Published online by Cambridge University Press 09 Mar 2009, doi: 10.1017/S0263574700016210. [11] P.V. C. Hough. A Method and Means for Recognizing Complex Patterns. US Patent: 3,069,654, Dec. 1962. [12] Allan G. Weber, The USC-SIPI Image Database: Version 5, Original release: October 1997, Signal and Image Processing Institute, University of Southern California, Department of Electrical Engineering. http://sipi.usc.edu/database/ (Last accessed on 25th May, 2011). [13] J. K. Mandal, Madhumita Sengupta, “Authentication /Secret Message Transformation Through Wavelet Transform based Subband Image Coding (WTSIC)”, IEEE, International Symposium on Electronic System Design 2010, pp 225-229, ISBN 978-0-7695-4294-2, Bhubaneswar, India, Print ISBN: 978-1-4244-8979-4, DOI 10.1109/ISED.2010.50, Dec, 20th -22nd, 2010. [14] Madhumita Sengupta and J. K. Mandal, “Self Authentication of image through Daubechies Transform technique (SADT)”, First International Conference on Intelligent Infrastructure, CSI- 2012, 47th Annual National Convention, Organised by: Computer Society of India, Kolkata Chapter, Proceedings published by Tata McGraw Hill Education Private Limited, ISBN(13):978-1-25-906170-7, ISBN(10): 1-25-906170-1, pp- 249252, 1st -2nd, December, 2012.

About the Authors

Madhumita Sengupta (born at Chandigarh, in 1982), M. Tech (CSE, University of Kalyani, 2010), pursuing Ph.D. as University Research Scholar (University of Kalyani), in the field of Transform domain based Image Processing, Steganography. Total number of publications 15.

Prof. (Dr.) Jyotsna Kumar Mandal M. Tech. (Computer Science, University of Calcutta), Ph.D.(Engg., Jadavpur University) in the field of Data Compression and Error Correction Techniques, Professor in Computer Science and Engineering, University of Kalyani, India. Life Member of Computer Society of India since 1992 and life member of cryptology Research Society of India. Ex-Dean Faculty of Engineering, Teachnology & Management, working in the field of Network Security, Steganography, Remote Sensing & GIS Application, Image Processing. 26 years of teaching and research experiences. Nine Scholars awarded Ph.D. one submitted and eight are pursuing. Total number of publications is more than two hundred seventy seven in addition of publication of five books from LAP Lambert, Germany.

CSI Journal of Computing | Vol. 2 • No. 1-2, 2013