An SVD-based Fragile Watermarking Scheme With
Grouped Blocks
Qingbo Kang
Chengdu Yufei Information
Engineering Co.,Ltd.
610000 Chengdu, China
Email: qdsclove@gmail.com
Ke Li, Hu Chen
National Key Laboratory of Fundamental
Science on Synthetic Vision.
Sichuan University, 610000 Chengdu, China
Email: likeneill@gmail.com, huchen@scu.edu.cn
Abstract—This paper proposes a novel fragile watermarking
scheme for digital image authentication which is based on
Singular Value Decomposition(SVD) and grouped blocks. The
watermark bits which include two types of bits are inserted
into the least significant bit(LSB) plane of the host image
using the adaptive chaotic map to determine the positions. The
groped blocks break the block-wise independence and therefore
can withstand the Vector Quantization attack(VQ attack). The
inserting positions are related to the statistical information of
image block data, in order to increase the security and provide
an auxiliary way to authenticate the image data. The effectiveness
of the proposed scheme is checked by a variety of attacks, and
the experimental results prove that it has a remarkable tamper
detection ability and also has a precise locating ability.
Keywords—Image Authentication, Tamper Detection, Fragile
Watermarking, Singular Value Decomposition
I. INTRODUCTION
With the tremendous development of information technol-
ogy, especially in network communication and multimedia,
digital images have a paramount role in our daily life. How-
ever, the digital images can be easily modified and tampered
with the help of powerful image processing software. In fact,
lots of people can easily manipulate images that may bring
about human casualty or financial loss [1]. So maintaining
the authenticity and integrity of digital images has become a
considerable aspect of many organizations [2], [3]. The image
authentication schemes can fall into two types according to
the methods they are based on: cryptography based schemes
[4], [5] and fragile watermark based schemes [6]–[8]. Image
authentication schemes based on cryptography calculate a
message authentication code (MAC) from images using a hash
function, and they can detect if an image has been modified,
but they don’t have the ability to locate the modified regions
[9]. In a scheme which is based on fragile watermark, the
host images that need to be protected are embedded with
the watermark which is often generated using either image
features extracted from host image or the generated random
values, when the image need to be authenticated, the original
watermark is then extracted from the watermarked image to
detect the tampered regions [10]. Walton proposed the first
fragile watermark-based authentication schemes [11]. It only
provides very limited tamper detection [12]. Holliman and
Memon proved that schemes which are block-wise independent
are vulnerable to vector quantization attack [13]. The Vector
Quantization attack(VQ attack) means the counterfeit image
can be reconstructed using a vector quantization code-book
generated from a set of watermarked images, because all
blocks in the images are authenticated, hence the counter-
feit images are authentic with the watermarking scheme. To
withstand the VQ attack, researches proposed a number of
schemes. [12], [14] proposed the fragile watermarking schemes
that use the chaotic pattern to generate the different image
and then map it into a binary image eventually insert into
the LSB bit-plane of the host image. Since the corresponding
watermark of the modified pixel value may be consistent with
the original watermark, result in the failure detection to the
tampered image. In this case, localization of the tampered
regions is incomplete, and detection of the tampering pattern
is imprecision [15].
Singular value decomposition(SVD) is a kind of effective
method of algebraic feature extraction. It can not only capture
the basic structure of the data in the matrix, but also reflect
the algebraic essence of the matrix. These excellent features
make it have a wide application in signal processing, image
compression, pattern recognition and other fields. SVD also
has been widely used in robust watermarking field [16]. In the
scheme proposed by Sun et.al [17], performing SVD in the
spatial domain, and the watermark is embedded by quantizing
the largest SV of an image block. However, this method is
vulnerable to VQ attack. In this paper, a novel SVD-based
watermarking scheme for image authentication is proposed.
The blocks of the host image are disturbed with the help of
Arnold scrambling method. Then, all scrambled image blocks
are divided into grouped blocks. For each block, two types of
watermark bits are embedded, one for the block itself, the other
for the grouped blocks. An adaptive chaotic image pattern is
generated using the logistic map for each block to determining
the embedded position of the watermark bits. The use of the
watermark bits of the grouped blocks is to break the block-
wise independence and withstand the VQ attack.
II. S
INGULAR VALUE DECOMPOSITION AND CHAOTIC
MAPS
A. Singular Value Decomposition
In linear algebra, the SVD is a factorization of a real or
complex matrix. Formally, any real or complex m × n matrix
172
2nd International Conference on Information Technology and Electronic Commerce (ICITEC 2014)
978-1-4799-5299-1/14/$31.00 ©2014 IEEE Dalian, China