Fake Image Identification using Image Forensic Techniques

Satyendra Singh, Rajesh Kumar

Departments of Electronics & Communication, JK Institute of Applied Physics and Technology, University of Allahabad,

Prayagraj, UP, India

Keywords: Image forensic, Photo Forgery Detection, Copy-move Image Forgery, Splicing

Abstract: These days’ digital images are a popular way of information sharing. it is known that the uses of images

have diverse areas such as news, magazine, medical, education and entertainment etc. this popularity of

digital images creates an opportunity for the researchers to ensure trustworthiness of an images. Due to all

these important utilities, the digital image has been successfully making it place in society. In this era of

digitalization image editing software is available in cheap mobile devices. That can generate fake image

easily. Fake images have been used by some organisations for influencing election, and violence and

rumour in the society. This paper focus on forgery detection techniques of fake images and understand the

usability of proposed technique for better accuracy of fake image detection.

1 INTRODUCTION

Digital image is a collection of picture

elements(pixel), digital images used in various areas

such as social media, news courtrooms,

entertainment, political campaigns and so on. The

use of digital image is growing very fast today,

digital image has become an important means of

sharing information, even in a small meeting or an

important event, we all use digital cameras or

camera enabled mobile devices. Capture every

moment and share the photo of that moment on

social media. But with the speed of digital images

are becoming popular among us, the trend of fake

image is also increasing. It is difficult to ensure the

integrity and authenticity of the image due to

tempering of the images.

Nowadays image manipulation application

software is easily available on cheap mobile devices.

And by using some powerful software to manipulate

image, without leaving obvious visual clues. Image

authenticity problems occur in application that is

courtroom, mass media and so on. Concerning

authenticity problems of digital images. Therefore,

image an important issue in forensic science.

The digital image manipulation is widely used in

bad aim and hiding original information of an image

has a long history. The history of photo forgery is

very long. Photo has been used by people for a long

time, but the image forgery has been used for the

bad purpose. In below Figure1 shown a photo of

Mao, Stalin, Hitler, Castro, Mussolini, and Brezhnev

times. In this figure photograph is manipulated for

better poses to erasing enemies or bottle of beer

(

Kim et al., 2012)

Figure 1: farid@cs.dartmouth.edu 1

In the times of Stalin, the image manipulation is

needed long hours of lengthy work in darkroom, but

in the present day any one can manipulate image in

few seconds by using photo editing software like

adobe photoshop that cannot be easily detected. The

authentication of photo tampering is necessary to

secure image communication process and

trustworthy of an image.

This picture was taken in the G20 summit in

Hamburg, Germany. The images that are given

bellow official picture of G20 summit and showing

US president Trump on the side lines or lots of

images showing US president and Vladimir Putin

meet first time. In Figure 2 is a fake image, making

the round on social media in 2017. Photo in Figure 3

is the original Getty image.

Singh, S. and Kumar, R.

Fake Image Identiﬁcation using Image Forensic Techniques.

DOI: 10.5220/0010563000003161

In Proceedings of the 3rd International Conference on Advanced Computing and Software Engineering (ICACSE 2021), pages 95-100

ISBN: 978-989-758-544-9

Figure 2: Fake image, making the round on social media

in 2017.

Figure 3: Original Getty image

1.1 Image Forgery

Photo forgery is describe as deleting, adding, and

changing few main features from digital image

without leaving any sign (

Qureshi et al., 2014)

. Image

forgery is created many challenges to ensure

trustworthy of an image.

Types of digital image forgery

Recent years, many ways have been employed to

temper with digital images. Several types of digital

photo tampering are existing.

Some common types of image forgeries are:

1. Copy-move photo tampering

2. Splicing

3. Image Morphing

4. Image Retouching.

1.1.1 Copy-Move Photo Tampering

The digital photo tampering technique is easy to

implement and difficult to detect (

Abidin et al., 2019)

In this tampering some features of a photo are copies

and pasted to another area in the same photo. In

digital photo copied part can be any types (

Qureshi,

2014)

. In figure 4 shown an example of copy move

forgery.

Figure 4: Copy move forgery image.

1.1.2 Splicing

Splicing is another type of technique of image

tampering. In this tempering of image forgery two or

more images are used for creating tampered photo

(

Tembe et al., 2017)

. In this technique some feature of

photo is copied from one photo and pasted to

another photo. Image splicing is difficult to detect

than copy move forgery (

Majumder et al., 2018)

Figure 5 shows an example of an image splicing,

here left side is natural photo and right side (b) is the

spliced photo.

Figure 5: Splice image.

1.1.3 Image Morphing

In morphing forgery, morphing is applied in two

images. In this technique shape of photo is change to

one form shape in another photo (

Elaskily et al.,

2017)

. Figure 6 an example of photo morphing photo

(a) is original photo of Hillary Clinton, photo (b) is

morphed photo and (c) is original photo of Donald

Trump.

ICACSE 2021 - International Conference on Advanced Computing and Software Engineering

(A) (B) (C)

Figure 6: A is original image of Hillary Clinton, B

Morphed-image and C is original photo of Donald Trump.

1.1.4 Image Retouching

This is less dangerous type of photo forgery than

other types of photo forgery techniques. In this

technique original photo do not changes, but some

enhancement or reduces feature of original photo

such as light, background and color changing to

divert attention about an objective in photo. Image

retouching technique is popular among smart phone-

based camera users and magazine photo editors

(

Mankar et al., 2015

). Fig.7. is an example of image

retouching.

on title should be in 11-point bold with initial letters

capitalized, aligned to the left with a linespace

exactly at 12-point, hanging indent of 1,0-centimeter

and with an additional spacing of 10-point before

(not applicable right after a subsection title) and 10-

poi

Figure 7: Retouching image.

2 LITERATURE REVIEW

(Eduard et al., 2019) present a system for detection

of splicing forgery of digital image forgeries. In here

method based on VGG-16 CNN. Author use CNN

based method for classification of images in two

parts first one is original and second forgery. And

classification is used image fragments. In this paper

used CASIA dataset for experiment. In this dataset

images are partitioned into training and test sample

in the ratio of 80:20. All the images are additionally

compressed by using JPEG algorithm. Proposed

CNN based algorithm used on compressed images of

quality factor(Q). accuracy on Q=90 and Q=80 is

67.1% and 66.3% respectively. And in this paper,

the accuracy of proposed algorithm is 97.8% for

fine-tuned model and 96.4% accuracy for zero stage

trained. Limitation of this algorithm, it has narrow

range of accuracy. Where forge image is compressed

by JPEG compression algorithm.

(Ghoneim et al., 2018) in here, author describe a

system for medical photo tamper detection for smart

healthcare system. Author proposed system is used

for checking authenticity of an image and identify

that photos related to healthcare are not change. The

proposed forgery system consists of noise extraction

pattern, SVM and ELM classifier and the realization

of multiresolution regression filter. In this paper the

proposed system is tested on three different

databases, two databases CASIA 1 and CASIA 2 has

original photo and other one DDSM database

mammogram. The proposed method achieve

accuracy 98% for original photos and 84.3% for

medical photos. The method achieved best

performance, when add the score of two classifiers.

(Selvaraj et al., 2020) in this study, author

proposed an improved key point-based copy move

forgery detection system and using a sensitivity

based clustering approach. Author find that

sensitivity-based clustering performs well in

comparison to existing agglomerative clustering and

DBSCAN algorithm. The suggested approach is also

resistant to a variety of geometric attacks, such as

rotation, composition and scaling.

(Dixit et al., 2020) proposed method has proved,

if picture is obtained from various datasets with

varied features. This method has proven to be

successful in detecting forgeries. For image level

and pixel level detection, the suggested approach has

shown encouraging results. The proposed method

shows robustness against composite attacks.

Proposed approaches to detect tampered pictures

that can withstand and a wider range of distortion

parameters, such as non-affine transformations, in

the future.

(Meena et al., 2020) author describe the Fourier

Mellin Transform (FMT) and SIFT algorithm for

copy move forgery detection technique. The

proposed method shows very satisfactory results

under various geometric transformations because the

FMT and SIFT descriptors are rotation and scaling

invariant in nature. Author find that the proposed

technique works very good in some special

condition like scaling with factor 50% - 200%, and

compression in JPEG with a quality factor up to 20.

3 MATERIALS AND METHODS

There are some images and some methods to find

the images are fake or not.

Fake Image Identiﬁcation using Image Forensic Techniques

3.1 About Image

These images have been taken from Iranian Missile

test images in the form of jpg. In figure 8 is the

original image of missile test and in figure 9 is the

tampered image of missile test.

Figure 8: Original image of Iranian missile test.

Figure 9: Fake image of Iranian missile test.

3.2 Methods

In the Experimental purpose Scale invariant feature

transformation (SIFT) is used for copy move tamper

detection. In this method some steps are applied

firstly give input original or forged image converts

input image data into SIFT features named sift

descriptor, in second steps this feature covert into

clustering SIFT then matching clustering results and

finally take decision on forged or original image.

Table 1: Comparison of various tampering Detection techniques

SSN Title Methods Detection

Domain

Advantage

1 Tampered and computer-

generated face photos

identification based on deep

learning (Dang et al., 2020).

Deep learning -

basedframework.

Tampered and

computer-

generated face

photos

detection.

Flexible, computationally

efficient, and robust against

imbalanced dataset.

2 Deep learning on digital photo

splicing detection using CFA

artifacts ( Hussien et al., 2020).

Deep learning

using color filter

array.

Digital image

splicing.

Accuracy is 95.5%

3 Detecting fake images on social

media using machine learning

(AlShariah et. al.).

Deep learning

technique via

CNN.

Detecting fake

image on social

media

Accuracy is 97%

4 Copy-move tamper detection

using SURF feature extraction

and SVM supervised learning

technique (Dhivya et al., 2020).

Speeded up

robust feature

(SURF) and

SVM

Copy move

forgery

Accuracy is 95.5%

5 Median filtering forensics in

digital photos based on

frequency-domain features (Liu

et al., 2017).

A novel

frequency domain

feature

Medium

filtering

detection

Reduce computing and merit for

mass processing data in real

time.

6 Copy-move forgery detection

using SIFT algorithm (Huang et

al., 2008).

SIFT Copy move

forgery

This method has good

performance on different types

of post photo processing (such

as rotation noise, scaling etc.)

ICACSE 2021 - International Conference on Advanced Computing and Software Engineering

4 EXPERIMENTAL RESULTS

To validate our proposed method, we performed

experiment for detecting copy move forgery using

scale invariant feature transformation (SIFT)

algorithm. SIFT algorithm applied on copy move

forged image. Following two figures are show

experimental results. Figure 10 shows results of

DoG pyramid images and Figure 11 result of copy

move forge images.

Figure 10: Results of DoG pyramid images.

Figure 11: Result of detected feature of copy-move forged

images.

For experimentation purpose MATLAB 2015a

student version and window 10 operating system,

8gb RAM and processor intel core i5 has been used.

In above figure green mark region using key points

and blue mark region is accurate selected key points

approximation. In this figure green and blue mark

region show copy part of same image. SIFT is better

method to detect copy move forgery.

5 CONCLUSIONS

In this work various types of digital image

tampering identification techniques are studied and

tested. For testing copy-move tampered image

Scale invariant feature transformation algorithm has

been used and experimented and experimental

results show that it is better, as compared to another

image forgery detection techniques. The main aim of

this study is to be understanding the various image

forgery detection techniques. Further this study

helps to the beginners for understand fundamental

steps involved in digital image forensic.

REFERENCES

Abidin, Arfa Binti Zainal, et al. (2019). "Copy-Move

Image Forgery Detection Using Deep Learning

Methods: A Review." 2019 6th International

Conference on Research and Innovation in

Information Systems (ICRIIS). IEEE.

AlShariah, N. M., Khader, A., & Saudagar, J. Detecting

Fake Images on Social Media using Machine

Learning.

Dixit, A., & Bag, S. (2020). Composite attacks-based

copy-move image forgery detection using AKAZE and

FAST with automatic contrast thresholding. IET

Image Processing, 14(17), 4528-4542.

Dhivya, S., Sangeetha, J., & Sudhakar, B. (2020). Copy-

move forgery detection using SURF feature extraction

and SVM supervised learning technique. Soft

Computing, 1-12.

Dang, L. M., Min, K., Lee, S., Han, D., & Moon, H.

(2020). Tampered and computer-generated face

images identification based on deep learning. Applied

Sciences, 10(2), 505.

Elaskily, M. A., Aslan, H. K., Elshakankiry, O. A.,

Faragallah, O. S., Abd El-Samie, F. E., & Dessouky,

M. M. (2017, November). Comparative study of copy-

move forgery detection techniques. In 2017 Intl Conf

on Advanced Control Circuits Systems (ACCS)

Systems & 2017 Intl Conf on New Paradigms in

Electronics & Information Technology (PEIT) (pp.

193-203). IEEE.

Eduard, A., & Shashkin, A. (2019). Journal of Physics:

Conference Series.

Ghoneim, A., Muhammad, G., Amin, S. U., & Gupta, B.

(2018). Medical image forgery detection for smart

healthcare. IEEE Communications Magazine, 56(4),

33-37.

Hussien, N. Y., Mahmoud, R. O., & Zayed, H. H. (2020).

Deep Learning on Digital Image Splicing Detection

Using CFA Artifacts. International Journal of

Sociotechnology and Knowledge Development

(IJSKD), 12(2), 31-44.

Huang, H., Guo, W., & Zhang, Y. (2008, December).

Detection of copy-move forgery in digital images

using SIFT algorithm. In 2008 IEEE Pacific-Asia

Workshop on Computational Intelligence and

Industrial Application (Vol. 2, pp. 272-276). IEEE.

Kim, H. J., Lim, S., Moon, J., Kim, B., & Jung, E. S.

(2012). A photographic forensic case study: Myths,

principles and techniques. Mathematical and computer

modelling, 55(1-2), 3-11..

Kashyap, A., Parmar, R. S., Agrawal, M., & Gupta, H.

(2017). An evaluation of digital image forgery

detection approaches. arXiv preprint

arXiv:1703.09968.

Liu, A., Zhao, Z., Zhang, C., & Su, Y. (2017). Median

filtering forensics in digital images based on

frequency-domain features. Multimedia tools and

applications, 76(21), 22119-22132.

Meena, K. B., & Tyagi, V. (2020). A hybrid copy-move

image forgery detection technique based on Fourier-

Fake Image Identiﬁcation using Image Forensic Techniques

Mellin and scale invariant feature transforms.

Multimedia Tools and Applications, 79(11), 8197-

8212.

Mankar, S. K., & Gurjar, A. A. (2015). Image forgery

types and their detection: A review. International

Journal of Advanced Research in Computer Science

and Software Engineering, 5(4), 174-178.

Majumder, M. T. H., & Al Islam, A. A. (2018, December).

A Tale of a Deep Learning Approach to Image

Forgery Detection. In 2018 5th International

Conference on Networking, Systems and Security

(NSysS) (pp. 1-9). IEEE.

Qureshi, M. A., & Deriche, M. (2014, February). A review

on copy move image forgery detection techniques.

In 2014 IEEE 11th International Multi-Conference on

Systems, Signals & Devices (SSD14) (pp. 1-5). IEEE.

Rothe, P. R., Asthankar, P. P., & Rothe, J. P. Image

Forgery Detection Based on SURF and Machine

Learning Classifier.

Selvaraj, P., & Karuppiah, M. (2020). Enhanced copy–

paste forgery detection in digital images using scale‐

invariant feature transform. IET Image Processing,

14(3), 462-471.

Tembe, A. U., & Thombre, S. S. (2017, February). Survey

of copy-paste forgery detection in digital image

forensic. In 2017 international conference on

innovative mechanisms for industry applications

(ICIMIA) (pp. 248-252). IEEE.

ICACSE 2021 - International Conference on Advanced Computing and Software Engineering

100