See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320703095

Passive Techniques of Digital Image Forgery Detection: Developments and

Challenges

Chapter in Lecture Notes in Electrical Engineering · January 2018

DOI: 10.1007/978-981-10-4765-7_29

CITATIONS READS

20 8,392

2 authors:

Santoshini Panda Minati Mishra

Fakir Mohan University Fakir Mohan University
3 PUBLICATIONS 34 CITATIONS 43 PUBLICATIONS 262 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Santoshini Panda on 08 November 2018.

The user has requested enhancement of the downloaded file.

 Passive Techniques of Digital Image Forgery
Detection: Developments and Challenges
Santoshini Panda1*, Minati Mishra2
1,2
P. G. Department of Information & Communication Technology
Fakir Mohan University, Balasore, Odisha
1
[email protected]
2
[email protected]

Abstract. Photographs and images play an important role in our life but, in this
technology era, equipped with powerful, low cost and easy to use photo editing tools,
people often forge photographs. This practice has posed a question mark on the
trustworthiness of images. Because carefully edited and forged images are very hard
to be distinguished from their genuine and original copies therefore, forgery detection
and separation of the forged images from the innocent ones has become a challenging
issue for image analysts. Image forgery detection procedures are generally classified
into two broad categories; the active and the passive detection techniques. This paper
presents a state of the art review of different passive forgery detection techniques
those are proposed by different authors over time.

Keywords: Copy-move forgery, Cloning, Splicing, watermark.

1. Introduction

Today, photo editing has become a common practice among people because of the
easy to use and freely available image editing tools. Though, all edited images are not
forged but some of those are. When forged images are available in huge numbers, it is
important to detect and separate the genuine copies from the forged lots. Image
forgery detection methods are generally classified into two categories, the active
detection methods and/ or the passive methods. The active methods are generally
termed as authentication techniques and are based on digital signatures or watermarks.
The major drawback of active image authentication is, for verification of the
authenticity of an image, a watermark or a digital signature need to be embedded into
an image at the time of capture or immediately after the image is captured [1]. Passive
forgery detection is an alternative to active authentication which requires no active
information available for the purpose of authentication. These techniques detect
forgery analyzing the image statistics in the absence of watermarks as well as the
original image for comparison.

1.1 Image Manipulation Techniques

Though all image manipulation or image editing procedures do not fall into the
image forgery category still it is certain that all forged images undergo some sort of
manipulation. The corrective manipulations techniques or the image enhancement
techniques such as contrast and brightness adjustments, noise reduction techniques etc
may not be included into forgery categories unless otherwise the manipulations fake
some facts or make some changes to the image contents so as to convey some
misleading information to the viewers [2]. According to authors of [1], image
manipulation techniques can be divided into two categories: content preserving and
content altering. Each of these techniques is further divided into different categories
as shown in figure 1. Authors of [2] argue that steganography can also be classified as
an image manipulation technique as that alters the image content invisibly. The
subsequent sections of the paper discuss various types of image forgery and different
forgery detection techniques.

2. Types of Image Forgery

Image forgery is categorized into four types such as, copy-move/ cloning, splicing,
retouching and morphing [3]. Copy-move or cloning is procedure where a piece of an
image is copy- pasted into some area of the image itself to create the forgery whereas,
splicing is another method in which parts of two or more images are stitched together
to create the forged image. Retouching is achieved by adjusting color, sharpness,
brightness, noise, contrast etc. Morphing is a procedure that transfers an image to a
different one through seamless transition between two images [4]. In figure 2 shown
are four images out of which the second image is a spliced image obtained from first
image and the fourth is a clone of the third.

Fig 1. Types of Image Manipulation

Fig2. a, c: Original Images, b: Spliced Image, d: Cloned Image

3. Techniques of Digital Image Forgery Detection

Basically, the image forgery detection procedures fall into two classes. One is
active detection and another is passive detection. Active detection requires in-built
digital signature and/or watermarking where as passive detection does not require any
of these two. It works by analyzing the content and statistics of the image [5]. Image
content based detection methods are further divided into the splicing detection and
cloning detection methods. Cloning detection techniques are generally based on
comparisons and matching. This falls into two major classes namely; the exhaustive
search method and the overlapping block matching (OLBM) techniques. The
exhaustive search method necessitates every possible part of an image to be compared
with every other part of the image to detect a match. This gives rise to very high time
complexity. Just like the exhaustive search method, simple OLBM method also can
detect clones in a forged image, if the image is not subjected to further post
processing such as intensity variation, noise contamination, compression etc. The time
complexity of this method is less than the time complexity of the exhaustive search
but, still remain as high as O (b2B2) where, b x b is the block size and B= (M-b+1) (N-
b+1), M x N being the size of the image. In addition to the time complexity, the other
dilemma with this method is choosing a right block size. If larger the blocks lead to
smaller detection accuracy then smaller blocks are resulted with higher of false
positive rates. The false positives are reduced by measuring the block shift and the
search time can be improved using Vectorization and lexicographic sorting. Further
reduction in dimension and improvement in search time is achieved by using DWT
and SVD methods. DCT and PCA based methods provides robustness against
intensity changes [6]. Coming to robust detection methods, the Luo, Huang and Qiu
[7] suggested seven feature based detection technique provides robustness against
JPEG compression, noise attacks and blurring but fails to detect intensity variant
clones whereas, the four feature based intensity invariant detection model for JPEG
compressed images (IIDMJPEG) proposed by Mishra and Adhikary detects intensity
variant clones in JPEG compressed images as well as provides robustness against
noise attack and blurring [8]. In figure 3 is presented a classification structure of
various digital image forgery detection methods.

Fig3. Classification of Forgery Detection Techniques

3.1 Passive Detection: Non-robust Methods

The passive forgery detection techniques have evolved through several phases
from the non-robust methods to highly robust techniques. The non-robust techniques
fail to detect forgery when the forged image has been subjected to lossy compression,
blurring, rotation, scaling noise attacks etc. On the other hand, the robust methods
succeed in detecting forgeries even if the tampered image is subjected to one or more
of these above mentioned post processing operations. This and the following sub-
section present a brief record of the evolution of the passive detection methods.
Jessica Fridrich, David Soukal and Jan Lukas [9] give the idea about efficient
detection method. First one is an exhaustive search method where an image and its
circularly shifted version are compared for matched parts. Though this is an effective
method the computational complexity of this method is very high it impractical for
making it impossible for practical use even for medium-sized images. The second
method suggested by the authors is the autocorrelation method where the forgery is
localized by the peaks corresponding to the cloned and the original parts in an image.
The Authors also suggested an overlapping block matching method that involves
vectorization and lexicographic sorting. In this method, a square block of b×b pixels
slides through an image one pixel at a time to give (M-B+1) (N-B+1) blocks. These
blocks are matched to locate the clones. The steps of block based exact match are
given in figure 4.

Fig 4. Forgery detection through block matching

S.Murali et al. [10] in their paper suggested a DCT based method for detecting
forgery in a JPEG compressed image and a standard deviation based edge detection
method. Zhouchen Lin [11] have proposed a robust method based on double
quantization (DQ) effect and have applied a trained SVM (Support Vector machines)
to take tampering decision. According to the authors DQ effect shows periodic peaks
and valleys in the histogram of DCT coefficients and hence can be applied for
tampering detection. Ruchita Singh, Ashish Oberoi and Nishi Goel [12] have used a
DCT and SIFT based method for feature extraction and forgery detection. Zimba and
Xingming [13] proposed two similar algorithms based on PCA and DWT. The
method proposed by Popescu [14] divides an N×N image into K=(N-b+1)2
overlapping. Each block is reshaped into b2 long row vectors and inserted into a K×b2
feature matrix. To improve the time complexity, DWT is applied to the blocks and
hence, the size of the feature vectors is reduced. They also proposed a method [15]
based on principal component analysis and Eigen value decomposition (PCA-EVD)
that reduces the dimension of the feature vectors and improves the computation time.
Babak Mahdian and Stanidlav Saic [16] proposed a blur invariant method that
successfully detects clones even if the cloned regions are blurred and noisy. Zaho
Junhong [17] presented an LLE (Locally Linear Embedding) - a non-liner dimension
reduction technique that detects copy-move forgery as well as fused edges. Sam T.
Roweis et al [18] have also used LLE for tamper localization. Ramandeep Kaur [19]
has used Local binary pattern (LBP) - a texture descriptor for feature extraction and
have used similarity criterion and Euclidean distance threshold for detection of clones.

3.2 Passive Detection: Robust Techniques

Jessica Fridrich, David Soukal and Jan Lukas [9] in their paper have also suggested a
technique robust against JPEG compression along with the techniques discussed in
the above sub-section. This method is same as that of the exact match with one
difference that it considers of quantized DCT coefficients for comparison instead of
the pixel representations for finding similarity.
N. Muhammad et al. in their paper [20] have proposed a dyadic Wavelet Transform
(DyWT) and DWT based robust and improved forgery detection model. Wang X. et
al. [21], in their paper, have suggested a DWT, DCT and Eigen vector based method
that is invariant to JPEG compression and additive noise. In this method first DWT
and DCT are applied to the image blocks and then the resulting coefficients are
multiplied to form the Eigen vectors. Block match is calculated measuring the mean
and variance of distances between Eigen values of Eigen vectors for forgery
localization. Jing Zhang et al. [22] have combined phase correlation with DWT for
their detection model. They have calculated spatial offset between copied and pasted
region to compute the difference between the image and its shifted version. This
method is claimed to reduce time complexity and is robust to jpeg compression. G. Li
et al. [23] also have also suggested a DWT and SVD based model where SVD
(Singular Value Decomposition) have been applied to the blocks of the low frequency
sub band then the SV vectors are lexicographically sorted to locate the clones. This
method is robust to JPEG compression up to quality level 70.

PCA based robust technique proposed by Popescu and Farid [24] efficiently detects
duplicated region with an computational cost in the order of O(NlogN) but is sensitive
to noise or lossy compression. In paper [25], Amtullah et al. have used a faster and
robust to noise speeded up robust feature (SURF) – a rotation and scale invariant key
point detector and descriptor based algorithm. To identify the duplicated regions, the
authors of [26] have combined KD-tree with SURF. A K-dimensional tree or KD-tree
is a binary tree with nodes as k-dimensional points and is common technique in
nearest neighborhood search. A KD-tree with N-nodes needs Olog2N search
operations. Prerna.C et al. [27] also used KD-tree with SHIFT and RANSAC
(Random Sample Consensus) algorithms. RANSAC algorithm has been used to find
out the unreliable key points. This technique is robust to additive noise and JPEG
compression. There are many more publications [28], [29] on KD-tree based methods
but all are not included here. Authors of [30], [31] have detected clones based on
SIFT algorithm where they have extracted the SIFT descriptors and then matched
those to identify the forgery. The SIFT (scale invariant features transform) algorithm
works in four steps such as, the scale space extrema (SSE) detection, the key-point
localization, orientation assignment and key-point description. This method is claimed
to be robust against noise attack, JPEG compression, rotation and scaling.

W. Luo, J. Huang and G.Qiu [7] in their paper introduced a seven feature based
method that is robust against noise, blur attacks and lossy compression. Najah
Muhammad et al. [32] have used multi-scale segmentation and denoising based
efficient technique for clone detection. This method is claimed to be robust against
noise and blurring attacks. In the paper hybrid copy-move forgery detection technique
using regional similarity indices [33], the authors developed a forgery detection using
local fractal dimension for image segmentation and estimating SSIM (Structural
Similarity Index Measure) between each block pair in each segmented region to
localize the forged regions. S.Bayram et al. [34] have suggested a FMT (Fourier-
Mellin transform) based robust to noise, blur, rotation, scaling and JPEG compression
method. They have used counting bloom filters (CBF) instead of lexicographic sorting
for computation time improvement. Solario et al. [35] have suggested a one
dimensional descriptor invariant to reflection and rotation based on log polar co-
ordinates. In this method, the pixels of overlapping blocks are represented in log polar
co-ordinates and summed along the angle to obtain the descriptor.
3.3 Comparison of Different Forgery Detection Methods
A comparison of various important forgery detection techniques evolved from time to
time is given in the table 1 below.

Table 1: Comparison of Different Forgery Detection Techniques

Method used Paper Advantages Disadvantages

serial
7- feature based [7] Robust to blurring, noise, lossy Not tested for rotation and
robust algo compression scaling
[8] Robust to blurring, noise, lossy
4- feature based Not tested for rotation and
compression. Can detect intensity variant
IIDMJPEG scaling -
clones
Exhaustive [9] Time complexity. Cannot
Detects cloned images without post-
Search and detect if changes is subjected to
processing.
Autocorrelation post processing operations.
[10] Detect forgery in JPEG compressed Fails in case of rotation,
DCT
image. scaling.
[11] Fast, Robust against JPEG
Not tested for rotation,
DCT and DQ compression and various forgery
scaling, intensity change.
methods, fine grained detection.
[12] Reduced time complexity, robust
DCT and SIFT Not tested for rotation
against rotation, scaling and noise.
DWT, PCA, [13] Reduced feature dimension. Better Fails in case of rotation,
EVD accuracy. scaling and heavy compression.
[15] Reduced the dimension and improved
PCA and EVD -
computational time.
[16] Robust against noise, JPEG
BLUR -
compression, blurring.
[17][18] Detects copy-move forgery as well as
LLE -
fused edges.
DWT-DCT and [21] Invariant to JPEG compression and
-
Eigen Vector additive noise
DWT, Phase [22] Reduce time complexity and robust
-
correlation to JPEG compression
[23] Robust to JPEG compression up to Not invariant to rotation and
DWT-SVD
QF 70, less time complexity. scaling.
[24] Sensitive to noise and lossy
PCA Efficient method, low false positives
compression
[25] Invariant to rotation and scaling.
SURF -
Faster and robust to noise
[27] Robust to additive noise and JPEG
KD-tree, SIFT -
compression
[30] [31] Robust against noise attack, JPEG
SIFT -
compression, rotation and scaling.
[34] Robust to blurring, noise, scaling, Cannot detect forgeries
FMT lossy compression and transitional which have rotation of above 10
effects degrees and scaling of 10%
log polar [35]
invariant to reflection and rotation -
descriptor

4. Summary and Conclusion

Even after a lot of research has been carried out during the last decade, passive
forgery detection still continues to be an open research area. Because copy-move
forgery or cloning is a technique commonly used by manipulators to forge digital
images so, copy-move forgery detection techniques form one of the most important
classes of passive detection techniques. The overlapping block matching (OLBM)
method suggested in 2003 by Fridrich, A. J. et al. was one of the important
developments in the field of cloning detection. Many improvements have been
suggested by various researchers from time to time to improve the time complexity of
the algorithm as well as to make the detection algorithm robust against post
processing operations such as changes in contrast , brightness and color, noise and
blurring attacks, lossy compressions, geometric transformations such as rotation,
scaling and translation to certain extends but still there exists the need to develop an
efficient algorithm that will be able to detect forgeries even after multiple post
processing operations and those have been subjected to more than 10% scaling and
rotated by an angle greater than 10 degree.

References:
[1] Haouzia A., Noumeir R.: Methods for image authentication: a survey.
Multimedia Tools and Applications, 39(1), pp.1–46, (2008).
[2] Mishra M., Adhikary M.C.: Digital Image Tamper Detection Techniques - A
Comprehensive Study. International Journal of Computer Science and
Business Informatics, 2(1), pp.1-12, (2013).
[3] Kaur H., Kaur K.: A Brief Survey of Different Techniques for Detecting
Copy-Move Forgery. International Journal of Advanced Research in
Computer Science and Software Engineering, 5(4), pp. 875-882, (2015).
[4] Shah H., Shinde p., and Kukreja J.: Retouching Detection and Steganalysis.
International Journal of Engineering Innovation & Research, 2(6), pp. 487-
490, (2013).
[5] Gupta A., Saxena N., and Vasistha S.K.: Detecting Copy-Move Forgery
Using DCT. International Journal of Scientific and Research Publications,
3(5), pp. 1-4, (2013).
[6] Mishra M., Adhikary M. C.: Detection of Clones in Digital Images.
International Journal of Computer Science and Business Informatics, 9(1),
pp. 91-102, (2014).
[7] Luo, W., Huang, J., and Qiu, G.: Robust detection of region-duplication
forgery in digital image. In: Proceedings of 18th International Conference on
Pattern Recognition IEEE, vol. 4, pp.746-749, ( 2006).
[8] Mishra M., Adhikary M.C.: Robust detection of Intensity Variant Clones in
Forged and JPEG compressed Images. ANVESA, 9(1), pp. 48-60, (2014).
[9] Fridrich, J., Soukal, D., and Lukas, J.: Detection of copy-move forgery in
digital images. In Proceedings of Digital Forensic Research Workshop,
(2003)
[10] Murali S., Govindraj B., Chittapur, Prabhakara H. S., and Basavaraj S.
Anami.: comparison and analysis of photo image Forgery detection
techniques. IJCA, 2 (6), pp. 45-56, (2012).
[11] Zhouchen L.: Fast, automatic and fine-grained tampered JPEG image
detection via DCT coefficient analysis. Pattern Recognition, 42(11), pp.
2492-2501, (2009).
[12] Singh R., Oberoi A., and Goel N.: Copy-Move Forgery Detection on Digital
Images. International Journal of Computer Applications, 98 (9), (2014).
[13] Zimba M., and Xingming S.: DWT-PCA (EVD) based copy-move image
forgery detection. International Journal of Digital Content Technology and
its Application, 5 (1), pp. 251–258, (2011).
[14] Popescu A.C., and Farid H.: Exposing digital forgeries by detecting traces of
resampling. IEEE Transactions on Signal Processing, 53 (2), pp. 758-767,
(2005).
[15] Zimba M., Xingming S.: Fast and robust image cloning detection using block
characteristics of DWT coefficients. International Journal of Digital Content
Technology and its Application, 5(7), pp. 359–367, (2011).
[16] Mahdian B., and Saic S.: Detection of copy–move forgery using a method
based on blur moment invariants. Forensic Science International, 171(2), pp.
180-189, (2007).
[17] Junhong, Z.: Detection of Copy-Move Forgery Based on one Improved LLE
Method. In: Advanced Computer Control (ICACC) IEEE, vol. 4, pp.547-
550, (2010).
[18] Roweis S.T., Saul L.K.: Nonlinear dimensionality reduction by locally linear
embedding. Science, 290(5500), pp. 2323 -2326, (2000).
[19] Kaur R.: Copy-Move Forgery Detection Utilizing Local Binary Patterns.
International Journal of Emerging Technologies in Computational and
Applied Sciences, 7(3), pp.290-294, (2013).
[20] Muhammad, N., Muhammad, H., Muhammad, G., and Bebis, G.: Copy-
Move Forgery Detection using Dyadic Wavelet Transform. In: Computer
Graphics, Imaging and Visualization (CGIV), Eighth International
Conference on IEEE, pp.103-108 (2011).
 [21] Wang, X., Zhang, X., Li, Z., and Wang, S.: A DWT-DCT based passive
forensics method for copy-move attacks. In: 2011 Third International
Conference on Multimedia Information Networking and Security, IEEE,
pp.304-308, (2011).
[22] Zhang, J., Feng, Z., and Su, Y.: A New Approach for Detecting Copy-Move
Forgery in Digital Images. In: Communication Systems, 11th IEEE Singapore
International Conference on IEEE, pp.362-366, (2008).
[23] Li, G., Wu, Q., Tu, D., and Sun, S.: A Sorted Neighbourhood Approach for
Detecting Duplicated Regions in Image Forgeries based on DWT and SVD.
In IEEE International Conference on Multimedia and Expo, pp.1750-1753,
(2007)
[24] Popescu A, Farid H.: Exposing digital forgeries by detecting duplicated
image regions. Technical Report TR2004-515, Department of Computer
Science, Dartmouth College, pp.1-1, (2004)
[25] Amtullah S., and Koul A.: Passive Image Forensic Method To Detect Copy-
Move Forgery In Digital Images. IOSR-JCE, 16(2), pp.96-104, (2014).
[26] Shivakumar B.L., and Baboo L.D.S.S.: Detection of Region Duplication
Forgery in Digital Images Using SURF. International Journal of Computer
Science Issues, 8(4), (2011).
[27] Prerna C., Percy G. J., Angaline S., and Thanga B. I.: A key-point based
copy-move forgery detection. International journal of advanced information
science and technology (IJAIST), 12(12), pp.175-180, (2013).
[28] Sagawa, R., Masuda, T., and Ikeuchi, K.: Effective nearest neighbor search
for aligning and merging range images, In: 3-D Digital Imaging and
Modeling, Fourth International Conference on IEEE, (2003) 54-61
[29] Bentley J.L.: Multidimensional binary search trees used for associative
searching. Communications of the ACM, 18(9), pp. 509-517, (1975).
[30] Huang, H., Guo, W., and Zhang, Y.: Detection of Copy-Move Forgery in
Digital Images Using SIFT Algorithm. In: Computational Intelligence and
Industrial Application, Pacific-Asia Workshop on IEEE, vol. 2, pp.272-276,
(2008)
[31] Lowe D.G.: Distinctive Image features from Scale-Invariant Key points.
International Journal of Computer Vision, 60(2), pp. 91-110, (2004).
[32] Muhammad, N., Hussain, M., Muhammad, G., and Bebis, G.: A Non-
Intrusive Method for Copy-Move Forgery Detection, In: International
Symposium on Visual Computing. Springer Berlin Heidelberg, pp.516-525,
(2011).
[33] Oommen R.S., and Jayamohan M.: A Hybrid Copy-Move Forgery Detection
Technique Using Regional Similarity Indices. International Journal of
Computer Science and Information Technology (IJCSIT), 7(4), pp.127-134,
(2015).
[34] Bayram, S., Sencar, T., and Memon, N.: An Efficient and Robust Method
For Detecting Copy-Move Forgery. In: IEEE International Conference on
Acoustics, Speech and Signal Processing, pp.1053-1056, (2009).
[35] Solorio, S. B., and Nandi, A.K.: Passive forensic method for detecting
duplicated regions affected by reflection, rotation and scaling. In: Signal
Processing Conference, 17th European, IEEE, pp.824–828, (2009).

View publication stats