A Robust Watermarking method based on. Compressed Sensing and Arnold scrambling. Veena V K, Jyothish Lal G, Vishnu Prabhu S, Sachin Kumar S, Soman ...
A Robust Watermarking method based on Compressed Sensing and Arnold scrambling Veena V K, Jyo this h Lal G, Vishnu Prabhu S, Sachin Kumar S, Soman K P
Centre For Excellence in C o mputational Engineering and Net w orking Amrita Vishwa V idyapee tham Coimbatore-641112, India
veena-vk89(Q:gmail.com
Abstract--Watermarking is a technique for information hid
leads to waste of data calculation. Besides compression, the
ing, which is used to identify the authentication and copyright
secrecy of the data is also having prim ary importance. Hence
protection. In this paper, a new method of watermarking scheme is proposed, which uses both Compressed Sensing and Arnold
scrambling m e thod for efficient data compression and encryption.
encrypt ion of the compress ed data is performe d follow ing the compression. This step is performed cither by conventional
Comprcssh'c sensing tcchniquc aims at thc rcconstruction of
cryptograph i c a lgo rithms or s ome custom
sparse signal using a small number of linear measurements.
pression and encryption
Compressed measurement.. are then encrypted using Arnold
transform. The proposed encryption scheme is computationally more secure against investigated attacks on digital multimedia signals. Keywords-Watermarking, Compressh'e sensing, Arnold scram bling Method, Encryption, Decryption.
1.
com
me thod is proposed to improve the secrecy of w at ermarked image. The se curity is don e in two stages namel y Co mpres s ive Sensing and Arnold scrambling. Compressive s ampling uni l1es
the
sampling, compressio n and encryption of the data to ,
be sent. This is achieved by collecting linear measurements
y=Ax of a sparse signal
is deve lopi ng
design j oint
In this paper, a new
x
where 'A: is the linear transform
carrying certain regu lar iti es . The lin e ar measurements 'y' are
INTRODUCTION
No waday s te chno log y
schemes [I].
fu ncti on of sensing matrix 'A:
at a rate faster than
pseudo-random
entries is
[2]. The matrix 'A: having
ge nerate d us ing a secret key, w hi ch
multimedia signals has increased more in the internet The security as
is known only to the sender and receiver. Thus Compressed
pect plays a major role in many applications such as video
representation of original signal. During the second level of
speed of light and the amount of redundant digital .
surveillance, confidential data tr ansmis sion in military and
medical fie lds , as the dig ita l
sig nals
are transmitted throug h
insecure transmission channel or network. So to discourage the
unauthorized c opying and distribution of the digital
Sensing measurements can be considered as an encrypted securi ty, the compre ssed and encry pt ed data obtained from the
first
level is subj ect ed t o Arno l d
scrambling This method of .
encryption scrambles the data in a certain I1xed number of
data, one
ite ration , which is known only to the authe nticated receiver.
has to protect intelligence information and intellectual property
Thus the receiver is able to decrypt the original data with
of the digital media. Watermarking is a unique technique for protection of intellectual property
ri ghts. The owners hip
information is hidden in the multimedia object, which can be
further
extracted or decoded for authentication purposes. It can
be in the form of text, image or any logo and the application
the aid of these secret keys . Expe rim ental results shows that
the combined security
of Co mpre ssive Sen sing a nd Arn old
scrambling proposed in this paper is computationally more
secure
against inv estigated attacks on sensitive data.
This paper is organized as follows. Section 2 gives a brief
decides which one to be used as watermark.
introduction a b out watermarking. Section 3 pr ovides a brief
is b and limited, it is desirable to compress the digital data prior to tnmsrnission for efficient util ization of memory/storage resources and co mm unic a tio n bandwidth. The conventional e ncoding process starts with
Compressive Se nsi ng scheme. Section 4 explains Arnold scrambling method. Section 5 describes the pr o po sed enc rypt ion and decryption scheme. S ect ion 6 gives the experimental results and final ly section 7 concludes the
sampling of the signal and then the sampling values are
paper.
As the transmission channel
quantized and coded for transmission or storage. The decoding
process is the reverse of encoding . But conventional appro ach
intro du cti on abo ut
II. WATERMARK EMBEDDING SCHEME
s trict l y fo llows N yquist Sa mpl ing theorem which states that
A watennar k is a s eco n dary i m age , label or m es sage that
the sampling frequency should be greater than or equal to
is embedded to the original image. The objective of the
bandwidth, making more complication in
watermarking scheme is to embed some piece of distinguished
the hardware design. In addition to that, lots of little coeffi
information within an image. So the modified output is a
cients are discarded during conventional compr es si o n , which
wat e rmarked image. The different algo r ithms used are discrete
tw o tim es me ssage
978-1 -4673-2322-2/1 2/$31.00 ©2012 IEEE 105
TABLE I Two DIMENSIONAL ARNOLD TRANSFORM PERIOD WITH DIFFERENT DEGREE N
Fourier, Cosine, wavelet transformation and fractal approaches for digital images. In general, the watermark is integrated into the image compo nents by a factor that allows amplification of the watermarking values in order to obtain the best results.The principle of correlation is used in detecting the watermark in the image.
5 6 7 10 12 8 N 32 40 48 50 56 60 Period 24 30 12 150 24 60
N 2 Period 3
3 4
4 3
9 10 11 12 30 5 64 100 12 125 48 150 60 250 8 6
12 12
16 12 128 256 96 192
24 12
25 50 480 512 120 384
It automatically detects the presence of watermark in the image based on a specific correlation level by comparing with original image. This method usually enables the in tegration of one bit watermark information. On integrating more information, the watermarking detector generates a bit sequence corresponding to a specified correlation level. In this paper, digital watermarking is done in the spatial domain by embedding the text information in the bth bit plane of the
using linear programming techniques whose computational complexity is of the order of O(N). To recover back x from y, it is required to estimate sparse solution to y=fW, where
min
III. COMPRESSIVE SENSING SAMPLING reconstructing the signals and images from fewer samples.
¢\II. Once
0 is known with the
IIfl11 subject toy
=
no
(2)
Thus the problem get changed into linear programming which is quite straight forward.
Compressive Sensing achieves this by two principles: sparsity
IV. ARNOLD SCRAMBLING METHOD
and incoherence. A.
=
be expressed as
image.
Compressive Sensing theory is the process of acquiring and
n
knowledge of \II x can be recovered. The sparse solution can
Arnold transformation is posed in the research of Arnold
Sparsity of the signal and incoherence
ergodic theory, which is also called cat face transformation.
Compressive sampling normally relies on the sparse nature
It is a chaotic transformation proposed by V.I Arnold. It
of the signal. Most of the natural signals are sparse in nature
is used in the digital image scrambling process because of
and therefore can be represented as in a concise manner
its iterative periodicity, good decentralization etc. Therefore
when they are expressed in the proper basis \II . Incoherence
it finds application in the field of image encryption, digital
gives the idea that object or data with sparse representation
watermarking etc. The transform process realign the pixel
in \II must be spread out in the domain in which they are
matrix by means of clipping and splicing. Two dimensional
acquired. More precisely, the sensing/sampling waveform have
Arnold transformation is defined as follows:
an extremely dense representation in \II . Let x be a real valued signal of length N which can be E
represented as x=\IIO where \II
RNxN is an orthonor
mal basis matrix(or dictionary) that can provide a K sparse
1 representation of x E RNx and 0 E RNx 1 can be well approximate using only K <