CHAPTER 1

INTRODUCTION
1.1 Image Encryption

Image Encryption techniques try to convert original image to another image that is
hard to understand; to keep the image confidential between users, in other word, it is
essential that nobody could get to know the content without a key for decryption. The
process of encoding plain text messages into cipher text messages is called encryption.and
the reverse process of transforming cipher text back to plain text is called as decryption.
Image and video encryption have applications in various fields including internet
communication, multimedia systems, medical imaging, Tele-medicine and military
communication. olor images are being transmitted and stored in large amount over the
Internet and wireless networks, which take advantage of rapid development in multimedia
and network technologies. In recent years, plenty of color image encryption approaches
have been proposed. !ntil now, various data encryption algorithms have been proposed
and widely used, such as "#$, %$", or I&#" most of which are used in text or binary
data. It is difficult to use them directly in multimedia data and inefficient for color image
encryption because of high correlation among pixels. 'or multimedia data are often of high
redundancy ,of large volumes and require real-time interactions.

'igure (.( Image #ncryption

(
#ncryption of images with the traditional encryption algorithms such as
%$", &#$ etc. was found inapt due to some typicality)s of images such as its bulk si*e as
also the correlation amongst the pixels . This gave rise to a new area of research for
encrypting images. #ncryption of images may broadly be classified based on the nature of
recovered image as either lossy or lossless image encryption. This classification resulted
in the following two different lines of approaches being adopted for maintaining
confidentiality of images.
1.2 Image Encryption (using eys!
This approach is basically similar to the conventional encryption methods
which involved using an algorithm +and a key, to encrypt an image. $ome of the proposed
techniques for encrypting images use -&igital $ignatures. , -haos Theory. , -/ector
0uanti*ation. etc. to name a few. There are some inherent limitations with these
techniques; they involve use of secret keys and thus have all the limitations as regards key
management. In addition, in some cases the available keys for encryption are limited
+restricted key space,. "lso high computation involved in encryption as also weak security
functions are also an issue . 1owever the greatest strength of most of these schemes is that
the original image is recovered in totality.
1." Image #p$itting
This approach, in a very basic form, involves splitting an image at the pixel
level into multiple shares + two or more,, such that individually the shares convey no
information about the image, but a qualified set of these shares will help regenerate the
original image +at least partially,. "di $hamir in (232 is credited for introducing the idea
of dividing a secret data into 4 random shares. In (225, 6aor and $hamir , using this as
the basis, proposed the concept of -/isual ryptography., which involves secret sharing
of an image by dividing it into multiple shares. 7any variations to the scheme proposed in
have been researched to overcome its limitations, each having their own merits and
demerits. &espite the advancements made in this line of research, the quality of the
recovered secret images still remains an area of concern due to the poor quality of these
recovered images+including loss of contrast and colors,. &espite its limitations the

4
greatest strength of these schemes is that firstly, there is no requirement of key
management and secondly the decryption involves no computation.
1.% Hy&ri' Approac(
In this approach using some kind of an encryption key the image is split
into random shares. Inc*e et al.proposed the concept of sieves for encrypting images. $ieve
is typically a binary key. The original image is placed over the sieve. 8ixels from the
original image situated above a hole of the sieve goes through and form one share of the
image. The pixels that stay on the sieve on a black pixel will form the other share. 'rom
the analysis of the various cryptographic approaches for images, it is appreciated that the
essentials for any cryptographic scheme would involve low computation cost, recovery of
original image, absence of keys and robustness. 1ence these motivations guide us to take a
novel approach.
1.) *isua$ cryptograp(y
/isual ryptography is an emerging cryptographic technique which allows
visual information +pictures, text, etc., to be encrypted in such a way that the decryption
can be performed by human visual system, without the aid of computers. It uses a simple
algorithm unlike the complex. It needs neither cryptography knowledge nor complex
computation. /isual cryptography technique +for black and white images, is introduced by
6aor and $hamir in (229 during #!%:%;8T)29. 'or security concerns, it also ensures
that hackers cannot perceive any clues about a secret image from individual cover images.
"ny visual secret information +pictures, text, etc, is considered as image and encryption is
performed using simple algorithm to generate n copies of shares depending on type of
access structure schemes. The simplest access structure is the 4 out of 4 scheme where the
secret image is encrypted into 4 shares and both needed for a successful decryption. These
shares are random dots without revealing the secret information. <asic visual cryptography
is expansion of pixels. /isual cryptography is a method of sharing a secret image among
a group of participants, where certain group of participants is called as qualified group who
may combine their shares of the image to obtain the original, and certain other group is
defined as forbidden group who cannot obtain any information on the secret image, even if
they combine knowledge about their parts. The scheme gives an easy and fast decryption

=
process that is done by stacking the shares onto transparencies to reveal the shared image
for visual inspection. It does not require very complicated cryptographic mechanism and
computation.
1.+ App$ication o, -isua$ cryptograp(y
Today the growth in the information technology, especially in computer
networks such as Internet, 7obile communication, and &igital 7ultimedia applications
such as &igital camera, handset video etc. has opened new opportunities in scientific and
commercial applications. <ut this progress has also led to many serious problems such as
hacking, duplications and malevolent usage of digital information. <eing a type of secret
sharing scheme, visual cryptography can be used in a number of applications including
access control. 'or instance, a bank vault must be opened every day by three tellers, but for
security purposes, it is desirable not to entrust any single individual with the combination.
1.. #cope an' o&/ecti-e
The main ob>ective of this approach is to provide conventional image
encryption schemes without using keys and a secret image is split into multiple random
images and with minimum computation cost the original secret image can be retrieved
back. The original secret image can be retrieved in totality. This approach employs
$ieving, &ivision and $huffling to generate random shares such that with minimal
computation, the original secret image can be recovered from the random shares without
any loss of image quality. The scope of this pro>ect is to encrypt the image without use of
encryption keys and recover the original secret Image from random shares without any loss
of image quality. . This approach employs $ieving, &ivision and $huffling to generate
random shares such that with minimal computation, low storage and bandwidth
requirements.

9
CHAPTER 2
PROPO#ED TECHNI0UE
2.1 propose' system
:ur proposed technique involves splitting an image into multiple shares. The shares
so generated reveal no information about the original secret image and to retrieve the secret
image all the shares are required. The proposed technique is implemented with the $&$
algorithm and involves three steps. In step one +$ieving, the secret image is split into
primary colors. In step two +&ivision, these split images are randomly divided. In step
three +$huffling, these divided shares are then shuffled each within itself. 'inally these
shuffled shares are combined to generate the desired random shares. The various steps
involved in generating two random shares are depicted in 'igure 4.(.
'igure 4.( $teps involved in generating two %andom $hares


5
?hile representing colors, additive and the subtractive color models are the most preferred
models. In the %@< or the additive model, the three primary colors i.e. %ed, @reen, <lue
are mixed to generate the desired colors. The colors as visible on the computer monitor are
an example of the additive model. $imilarly when using the 7; or the subtractive model,
the colors are represented by the degree of the light reflected by the colored ob>ects. In this
scheme yan +, 7agenta +7, and ;ellow +;, pigments are used to produce the desired
range of colors. This model is extensively used in printers.
$ince our proposed techniques involves computation during the encryption and
decryption stages and the results are to be viewed on the computer monitors hence it is
natural for us to use the additive color model. since the shares were printed on
transparencies, hence subtractive model was the natural choice for such applications.
:n a monitor an image may be thought as ?idth A 1eight 4-dimensional matrix, with
each entry in the matrix representing a pixel value. #ach of these pixels are a series of bits
composed of values representing the %@< values. B bit+4 bits each for %,@,<,, (C bits +9
bits each for %,@,<,, 49 bits++B bits each for %,@,<,, 9B bits +(C bits each for %,@,<, etc.
are some of the commonly used %@< schemes. 'igure 4 represents the representation of
%D@D< values for an individual pixel. If x be the number of bits used for representing any
primary color, then a total of 4=x colors can be represented by mixing the three primary
colors. The values of each primary color will then vary from E to +4x-(,.
The scheme that we present here is a +*, *, threshold scheme i.e. for retrieving a secret
image that has been divided into * shares all * shares are required. 6o shares individually
convey any information about the secret image, nor do a combination of subset of random
shares, the original image will only be retrieved from the complete set of random shares.
The scheme implemented using the $&$ +$ieve, &ivision, $huffle, algorithm involves the
following three stepsF

C
2.2 #ie-ing
$ieving as the name suggests involves filtering the combined %@< components
into individual %, @ and < components +refer 'igure =,. The granularity of the sieve
depends the range of values that %D@D< component may take individually. To make the
process computationally inexpensive, sieving uses the A:% operator.


'igure 4.4 'lowhart of the $ieving operation

3
2." Di-ision
1aving filtered the original image into the %, @ and < components, the next step involves
dividing the %, @ and < components into * partsD shares each.
% +%
"
, %
<
, %

,----------------, %
G
,
@ +@
"
, @
<
, @

,----------------, @
G
,
< +<
"
, <
<
, <

,----------------, <
G
,
?hile dividing it is ensured that each element in %
"-G
, @
"-G
and <
"-G
is assigned values
randomly, such that the entire domain is available for randomi*ed selection; in case x H B,
then individual elements should be randomly assigned a value varying from E- 455. The
shares so generated should be such that +%
"
, %
<
, %

,--------------- %
G
, should regenerate %
and similarly for @D< components .
'igure 4.= 'lowhart of the &ivision operation

B
2.% #(u,,$ing
The random shares created by division in no way exhibit any resemblance to
the original image, but as a second step towards randomi*ing the generated shares i.e. %"-
G, @"-G and <"-G , we perform the shuffle operation. This involves shuffling the
elements in the individual shares. The sequence in which the elements within the shares are
shuffled depends on the value of one of the other shares generated from the same primary
color. In other words %< decides how %" is shuffled, % decides how %< is shuffled,
------------ %G decides %G-( is shuffled and %" decides how %* is shuffled. The shuffling
operation uses the comparison operator on the I$< of the determining element to decide
the shuffle sequence.
'igure 4.9 'low hart of the shuffle operation

2
2.) Com&ine
1aving carried out the above three operations the generated shares are combined to
generate the final * random shares +%$,.
%andom share " - +%
"- shuffle
, @
"- shuffle
and <
"- shuffle
,
%andom share < - +%
<- shuffle
, @
<- shuffle
and <
<- shuffle
,
%andom share - +%
- shuffle
@
- shuffle
and <
- shuffle
,
The random shares so generated individually convey no information about the secret
image, however to recover the original image all the random shares would be required.
'igure 4.5 'low hart of the ombine operation.

(E
CHAPTER "
#D# A12ORITH3
".1. A12ORITH3
1. #ie-ing
Input $ecret Image
$ieve+$ecret Image,
:utput +%, @, < components,
2. Di-ision
n H total number of pixels + E to n-(,
%i D @i D <i H individual values of the ith pixel in the %, @, < components
* H total number of random shares
x Hnumber of bits representing each primary color
maxJval H 4x
%epeat 4 for %, @, < component
4+a, for i H E to +n-4,
K
for share k H " to +G-(,
%ki H %andom+E, maxJval,
"ggrJ$umi H %ki
L
%*i H+ maxJval M %i N +"ggrJ$umi O maxJval,, O maxJval
". #(u,,$e
%epeat for %"-G, @"-G and <"-G +all generated shares,
for k H " to G
K

((
%k-shuffle H %k
8tr'irst/ac H (
8trIast/ac H n-(
'or i H ( to +n-(,
K If +%+kM(,+i-(, is even,
K %+k-shuffle, 8tr'irst/ac H %ki
8tr'irst/ac MM, iMM
L
#lse
K %+"-shuffle, 8tr'irst/ac H %"i
iMM, 8trIast/ac --
L L L
%. Com&ine
'or k H " to G
%$k H +%k-shuffle A:% @k-shuffle A:% <k-shuffle,
Thus at the end of the above process we have %andom shares +%$" ,%$< -------------------
%$k,.

(4
CHAPTER %
ANA14#I# AND CO3PARI#ON
%.1 Comparison 5it( Image Encryption
Iast few decades have seen lots of schemes being proposed for image encryption using
keys, some of the prominent ones have been here. 7anniccam and <ourbakis in (224
proposed an image encryption and compression scheme using $"6 language. The
scheme was fundamentally based on chaos theory. 1owever this was applicable to only
grey scale images. $imilarly Ain and hen in 4EEB following up on the work of , proposed
a two stage image encryption scheme. $tep one involved fusion of the original image and
the key image and step two involved encryption of the fused image using 1enon chaotic
system. hen, 1wang and hen in 4EEE proposed the use of /ector 0uanti*ation +/0, for
designing a cryptosystem for images. In /0 images are first decomposed into vectors and
followed by sequential encoding of these vectors. Thereafter traditional cryptosystems
from commercial can be used.
%.2 Comparison 5it( Image #p$itting
The idea of Image splitting more often referred to as /isual ryptography $chemes +/$,
involves splitting a secret image into n random shares such that these shares individually
reveal no information about the secret image +but for its si*e, but a qualified subset of the
shares+as specified by the encrypter, when stacked up reveal the secret image. The random
image shares +qualified set, are merely printed on transparencies and stacked up revealing
the original image,. The ma>or issues which restrict its employment is the poor quality of
the recovered image limited color representation etc. 7any research papers have been
published using this approach, starting from a binary imag moving to greyscale image and
finally employing it to color images. Though with each subsequent research paper the
quality of the recovered image improved, however, but for no other scheme was able to
completely recover the original image from the shares. ?hen evaluating the performances

(=
of these suggested solutions they are often evaluated on performance measures such as
contrast, accuracy, security, computational complexity etc. Thus an ideal solution would
regenerate the original image from the shares in terms of colors and contrast, it would also
have to be secure and computationally inexpensive.
%." Comparison 5it( Hy&ri' Approac(
In this approach using some kind of an encryption key the image is split into random
shares. Inc*e et al proposed the concept of sieves for encrypting images. $ieve is typically
a binary key. The original image is placed over the sieve. 8ixels from the original image
situated above a hole of the sieve goes through and form one share of the image. The pixels
that stay on the sieve on a black pixel will form the other share. 'rom the analysis of the
various cryptographic approaches for images, it is appreciated that the essentials for any
cryptographic scheme would involve low computation cost, recovery of original image,
absence of keys and robustness.
%.% Comparison 5it( Ot(er #c(emes
Image encryption may be classified as lossy D lossless image encryption.
The conventional /$ schemes all generate a degraded image quality of the recovered
image and hence some modifications to /$ often referred to as /ariants to /isual $ecret
$haring schemes have also been proposed. 1ence a true comparison of our scheme would
involve comparing it to the other proposed /$s as also its variants. 7ost of the digital
cameras today support 49 bit true color schemes and upwards, hence it is natural that most
of the secret sharing schemes would need to support 49 bit color schemes. $ome schemes
do not support 49 bit true color scheme. :ur scheme along with Tsai et.al scheme supports
49 bit true color schemes. "nother important factor is how the si*e of the share increases
with increase in the number of shares and the number of colors. This is a very critical
factor when considering the bandwidth constraint i.e. transmitting the shares on the net as
also the storage si*e of each of these shares. In the extended Thien and Iin)s scheme
supporting true color, the si*e of each share increases three times. $imilarly in Iukac and
8lantonis +n,n, threshold scheme each share becomes 4n-( times larger, thus with increase

(9
in number of shares i.e. n, the si*e of the share doubles for each new participant. In our
scheme the si*e of the random share is not a function of the number of colors in the image
or the number of shares. The si*e of the random share thus is always constant i.e. equal to
the si*e of the secret image. Thus the proposed schemes performs better on the bandwidth
and storage requirement parameters.
In our proposed technique both during encryption and decryption the
computation cost is low since the ma>ority of the operations use logical A:%, :% and
"6& operators. Tsai et.al scheme involves = steps, initial training, encoding and decoding.
The initial training phase involves 8rincipal omponent "nalysis +8", and 'orward
6eural 6etwork +''6,. The initial training phase itself involves heavy computation cost
though the encoding In our proposed scheme there are no keys involved and hence there is
no key management. "ll that is required is to transmit one of the random shares on a secret
channel while transmitting the rest on an unsecure channel. In Tsai et.al scheme the
decoding step involves use of a weighted matrix < generated during the training phase and
a seed Ps) used in the encryption phase, thus handling of these two secret elements raises
issues similar to key management in an encryption algorithm. In Tsai et.al scheme the
quality of the recovered image is almost similar to the original secret image, however the
fact remains that the recovered image is not same as the original secret image. In our
scheme the recovered image is an exact replica of the original image as no data is lost
during the sieving division and shuffling operations. The results were validated using
6ormali*ed orrelation +6,. 6 is used to measure the correlation between the original
secret image and the recovered images from the random shares.

$ represents the secret image and % the recovered image. w, h represents the widthDheight
of the photographs and represents the exclusive :% operator. ?e repeated the test over
multiple images, the 6 for all the recovered images was (.EEE. The generated random
shares are highly secure as the spatial correlation between the pixels is eliminated by
employing the randomi*ation function thrice for each pixel value per share

(5
%.) 6uture #cope
The scheme is suitable for authentication based application or where trust cannot be
reposed in any one participant for decision making and a collective acceptance is required
to proceed. " typical scenario for this could be thought of as a secret code which has to be
fed in to commence a nuclear strike; the said code could be converted into an image and
split into random shares, held with the collective decision making body. To retrieve the
secret code random share of all the participants would be required.
CONC1U#ION

(C
In this paper a novel visual cryptographic technique is presented,
which is a combination of the traditional visual cryptographic
scheme and the conventional image encryption schemes. " secret
image is divided into multiple random shares and with minimum
computational cost the true secret image can be retrieved back.
The proposed algorithm has the following advantages
+a, The original secret image can be retrieved without any loss in
colour and contrast
+b, There is no pixel expansion problem and there by storage
requirement per random
share is same as original image
+c, 6o key management problem since we are not using any keys
for encryption
+d, The system is robust to withstand brute force attacks with
lower cost.

(3
RE6ERENCE#
Q(R $ubramanian Salyan, %./ignesh, !>>val1.Thakkar, T."diline 7acriga9 =rd ;ear,
&epartment of of Information Technology, $ri $airam #ngineering ollege hennai,
India.T7I$TIS:$-on ommunication $ecure &ata Transmission In ?ireless 6etwork
T4E(4 I### $tudents) onference on #lectrical, #lectronics and omputer $cience.
Q4R Ain Ghang and ?eibin hen, T" new chaotic algorithm for image encryption ,
International onference on "udio, Ianguage and Image 8rocessing, 4EEB. +I"II8
4EEB,, pp BB2-B24.

Q=R "loka$inha and Sehar $ingh, T" technique for image encryption using digital
signature , :ptics ommunications+4EE=,,4(B+9-C,,pp442-4=9.
Q9R $.$.7aniccam, 6.@. <ourbakis, TIossless image compression and encryption using
$"6 , 8attern %ecognition =9 +4EE(,, pp (442-(495.
Q5R hin-hen hang, 7in-$hian 1wang, Tung-$hou hen, T" new encryption
algorithm for image cryptosystems , The Uournal of $ystems and $oftware 5B +4EE(,, pp.
B=-2(.

QCR ". $hamir,T1ow to share a secret, ommun."7, vol. 44, no. ((, pp.C(4NC(=,
(232.
Q3R 7. 6aor and ". $hamir,T/isual cryptography, in 8roc. #!%:%;8T)29, <erlin,
@ermany, (225, vol. 25E, pp. (N(4, $pringer-/erlag, I6$.
QBR . hang, T.A ;u, T$haring a secret gray image in multiple images , inF8roceedings
of 'irst International $ymposium on yber ?orlds,4EE4,pp.4=E-49E.
Q2R 1.-. ?u, .-. hang, T$haring /isual 7ulti-$ecrets !sing ircle$hares , omput.
$tand. Interfaces (=9 +4B, ,pp. (4=N(=5, +4EE5,.
Q(ER 1anan 7ahmoud 1anan$aad "l-1ulaibah $arah "hmad "l-6aeem, 6ovel
Technique for $teganography in 'ingerprints ImagesF &esign and
Implementation ,&epartment of Information Technology, ollege of omputer and
Information $ciences, entre of #xcellence in Information "ssurance Sing $aud
!niversity, Singdom of $audi "rabia.
Q((R ?ikipediaF httpFDDen.wikipedia.orgDwikiD#ncryption.
Q(4R "rpad Inc*e, T8ixel sieve method for secret sharing Vvisualcryptography
%o#du6et International onference 8roceeding $ibiu49-4C Uune 4E(E, I$$6 4ECB-(E=B,
p. B2-2C.