Investigation of Image Forensic Techniques to Determine Faked Images

Transcription

1 World Journal of Technology, Engineering and Research, Volume 2, Issue 1 (2017) Contents available at WJTER World Journal of Technology, Engineering and Research Journal Homepage: Investigation of Image Forensic Techniques to Determine Faked Images Praveen Y. Chitti a*, Dinesha H.A. b, K.Prabhushetty c a Research Scholar, Dept. of CSE, JCE- VTU, Belagavi, India b Associate Professor, Dept of CSE,JCE - VTU, Belagavi, India c Professor, Dept. of ECE, SGBIT-VTU, Belagavi, India Keywords Determine Faked Images Image Forensic Techniques Image Processing Photoshop Images A B S T R A C T Vast numbers of images in a day, and most of them have been digitally altered in some way. Few have altered basic color corrections, few are instagram filter effects. Some of them have got updated full on Photoshop jobs. It is to completely transform the subject for negative applications. It turns out humans are not in recognizing when an image has been manipulated, even if the change is substantial. Hence to identify these faked photos, a digital forensic technique can be employed. Many digital forensic techniques have been proposed in literature to identify faked and original images. These techniques are bring-out in our research as a study report for further analysis and to identify the research gap. We represent the study report on existing image forensic techniques, its research gap, and our analysis to get current issues WJTER All rights reserved. 414

2 I. INTRODUCTION At present world, everybody has an advanced camera. Billions of computerized pictures were taken. Some of these pictures are utilized for purposes other than family photograph collections or website improvement. On the popularity of advanced photography, producers of photo altering technologies rapidly make up for lost time force. These are available for less expensive and simpler to utilize so user friendly in certainty that anybody can utilize them to improve their pictures. Altering, if done legitimately, can incredibly improve the presence of the photo; increment its effect to the watcher and better pass on the craftsman's message. In any case, where is the moment that a narrative photo ends up plainly subjective gem. While for most purposes altering pictures is more than alright, certain sorts of photos are never to be controlled. Computerized pictures are routinely given to news editors as a feature of occasion scope. Computerized pictures are exhibited to courts as confirmation. For news scope, certain kinds of changes or alterations such as editing, cropping, mixing, modifying could possibly be adequate. Pictures introduced as court proves must not be controlled at all; else they lose believability as adequate proof [1]-[3]. In present digital era powerful graphical editors and well-equipped image manipulation techniques make it tremendously easy to alter original images. Many alterations are highly impossible to diagnose by an inexperienced person eye. It can even escape the verification of experienced skilled editors of prestigious news media as well. Indeed, even an outstanding skilled forensic specialist may miss few of the indications of a forge, potentially allowing forged images to be acknowledged as a evidence in law of court. Significant camera makers endeavored to address the issue by presenting frameworks considering secure advanced authentications. The reason for these frameworks was the capacity to demonstrate that pictures were not changed in the wake of being captured by the camera. Clearly went for photograph columnists and editors, this framework was additionally utilized as a part of lawful cases as bona fide court confirm. The approach looks awesome on paper. The main issue, it doesn't work. The clearly faked pictures effectively breezed through the legitimacy test by the individual producers' check programming which conveys us to the inquiry [2]. If human specialists are experiencing considerable difficulties deciding if a picture was adjusted, and if existing authentication based genuineness confirmation frameworks can't be depended upon, would it be advisable for us to simply abandon the very issue? This paper shows another probabilistic approach permitting programmed genuineness investigation of an advanced picture. The arrangement utilizes various calculations examining diverse parts of the advanced picture, and utilizes a neural system to create a gauge of the picture's realness, or giving the likelihood of the picture being forged [4]-[8]. This paper has been organized in following manner: Section 2 represents the review on image forensic and its existing techniques to identify forged images. Section 3, represents the review on Copy Move Forgery and its detection techniques. Section 4 discusses the research gap in the field of image forensics so far. Section 5 concludes the paper along with future enhancements. II. REVIEW ON IMAGE FORENSIC AND ITS EXISTING TECHNIQUES In this era of digitalization, everyone is surrounded by digital contents like digital image, digital video with the usage of electronic gadgets. Recent technologies have made camara 415

3 cheaper by promoting cameraa facilitated devices such as mobile phones, laptops, tablets and so on. People across the globe are uploading millions of images per minute. In many cases, these contents are making life easy in some way. This technology advancement is also facilitating high quality image editing software s, mobile apps, web interfaces that can make undetectable changes in images i and videos easily. Currently, many image editing tools are available in market which can easily use for image forgery. Few of the popular tools are i) GIMP, ii) Affinity Photo, iii) Acorn, iv) Adobe Lightroom, v) Adobe Photoshop, and etc [20]. Additionally, there are numerous image editing apps available in play store for android phones such as i) Pixlr, ii) PicsArt, iii) Instagram, iv) Cymera, v) Aviary, vi) Snapseed, and etc [9]. These tools/apps usage may be for positive purpose or negative purpose. Negative usage most possibly for creating some rumours, influence political fight, unethical reasoning, illegal purposes and etc. These negative moves are creating high demands for forensic analysis of images to prove their originality. In earlier stenography approach, when the source is confirmed, the image origin and its integrity can be authenticated by using digital watermark and it is a part of active forensic analysis. In many cases, the sources of images are unknown. That means except images, we don t have any background information and it is a part of passive/blind forensic analysis. At present market demands for blind forensic analysis and in our review article we emphasis more on passive forensic analysis techniques [10 14] and its challenges. Camara Pixel Geometric Format Statistical Physical Color filter array Resembling Reflection Shadow JPEG, GIF, MPEG, PNG PCA, LDA 2D, 3D Sensor imperfection Thumbnail, copy -move Calibration, Principle point CFA Interpolation Light Environment, whether Camara Response Figure 1: Various forms of Digital Forensic Tools Before going in depth details of blind forensic, let us present the two ways of fake images that can be created i) splicing and ii) copy move. In splicing, two or more images of different type are used to create a fake image. In copy move, some image content of same image is misused to make fake image. These forgery ways and its detection represented in section 3. When generating the realistic looking fake images, various operations are carried out such as contrast and brightness enhancement, median filtering, resembling, cropping and etc. Forensic understanding of these types of operations also helps it detecting forgery. On top of this understanding, many anti-forensics techniques [15-20] have been developed to hide the artefacts of forgery. However, to understand the image forgery, literature presents the various phenomena such as i) camera based forensics, ii) pixel based forensics, iii) statistical based forensics, iv) Geometric based forensics and physics based forensics which are represented in Figure 1 [21]. Many forensic techniques have been proposed in literature which is represented in table 1[21]. Many digital forensic techniques and its focus along with its special features and constraints are represented. 416

4 Table 1: On-going Research Techniques Digital Forensic Technique Steganography Dataset: normal images e.g. Lena, Barbara,Baboon JPEG compression Dataset: Corel, NJIT, NRCS,UCID Preserving predict ion direction Dataset: UCID-v2 corpus Detection of interest point Dataset: Image manipulation Dataset -segmentation -image hash Dataset: CASIA Sourc e/aut hors Sun et al. Luo et al. Li et al. Zandi et al. Pun et al. Visual descriptors, Carvalho et statistical approach, K-Nearest al. neighborhood Dataset: DSO-I, DSI-I -Sliding window Korus et al. Linear Filtering Connoter et JPEG Compression al. Support Vector Machine Dataset: UCID -No comparative analysis -iterative process leading to storage complexity Not applicable for other forms of imageattacks Median filter, Gaussian Model Dataset: UCID-v2 corpus Expectation- Maximization, Segmentation Dataset: MICC- F600 Artifacts of histogram, JPEG compression Dataset: BOSS Public dataset, UCID Key-point injection, region classification, Dataset: INRIA Holidays, UCID Addresse d Problem Securing information Analysis of JPEG error Anti-forensic (hiding artifacts of lossy compression) Copy-move attack Tampered re gion localization Impersonating image regions Spec ial Feat ure Good retention of PSNR Effective operation towards error analysis Higher success rate -Simple Detection Technique -Faster Response Time Good Precision over colored image -Higher accuracy Detection of -supports both tampered double Analyzing 99.6% chains of accuracy operators. Recovery of original tampering operation Fan et al. Artifact hiding Good Accuracy, simple approach Li et al. Cao et al. Costanzo et al. Copy-move attack Forged image to be detected using contrast Identification of SIFT keypoints Good Accuracy High performance Solves majority of key- point based problems Constraint Outcomes not benchmarked Computational complexity is not addressed -involves computational cost -leads to iterative approach. - computationally complex -Only effective for splicing Highly de Involves higher computational cost Involves higher computational cost due to iterative process Doesn t addresses complexities associated with dataset Outcomes not benchmarked 417

5 dataset. Next section describes the copy move forgery operations and its detection techniques. III. REVIEW ON COPY MOVE FORGERY The previous section has discussed various forms of digital forensic tools and many digital forensic techniques along with its special features and constraints. Few of the significant research contributions towards addressing the problems associated with image forgery and varied forms of algorithms and approaches have been undergone. Copy Move Forgery and its detection techniques are highlighted in this section and represented in table 2[22]. It focuses on the forgery operations that are performing and tampering detection techniques that are addressing those operations. Table 2: Copy Move Forgery and Its Detection Techniques Source Forgery Operations on Images Digital image tamper detection techniques- A comprehensive study [8] Digital image tampering- A threat to security[9] [10] Detecting image splicing using merged features in chromo space[18] Image forgery detection based on semantic[19] Tampering and copy move forgery detection using SIFT feature[11] Efficient copy move forgery detection for detection for digital images[13] Survey of image forgery detection[14] Comparison and analysis of photo image forgery detection techniques [15] Retouching, spelling,copypaste, cropping, cloning Copy move, resize, image splicing, noising, blurring Tamper detection techniques Edge blurring Laplace filter, PCA, DCT,DWT,SVD Image splicing local/global DCT SRM,CASIA V2 dataset blurring compression and resize Copy move Framework semantic ontology commonsense knowledgebase Copy move, block, feature based methods. Copy move, image splicing Copy move, splicing,resize, cropping cloning Copy move, copy create, copy paste,pca,dct,dwt,sif T Statistical & block characteristics Pixel, format, camera physically, geometric based. JPEG compression analysis, edge detection, localization 418

6 Copy move image forgery detection method using steerable pyramid transform and texture descriptor[21]-[23] fields. IV. Copy move SPL,LBP Localization SIFT MIFT localization The next section highlights the discussion on research gap that are exists in image forensic Discussions After reviewing the digital forensic tools, techniques, and the silent feature of the image forgeries and detection, it is strongly recommend that there is a necessity of more strengthening the research direction. Research issues that are not yet being considered with image forensics to be focused are represented with descriptions. This section will describe such points to represents the significant research gap. Table 3 represents the research gap points to be considered along with its descriptions [22]. Table 3: Research gap highlights Research Gap /Points to be considered Higher Dependence towards Dataset Low Focus on Algorithm Complexity Non-equilibrium between Detection and Quality Lack of Benchmarked Outcomes Description It is still an open end questions about the sustainability of an existing algorithms of image forgery detection using dataset when swapped by real-time captures. There is no research work that has used both dataset and real-time captures to show robustness. At present, none of the existing techniques have claimed to offer cost effective algorithm performance without any deviation to image forgery detection mechanism. Hence, there is no balance between image forgery detection accuracy with algorithm complexity performance. At present, we don t find much work in literature where the image quality is being emphasized to a larger extent. Ignoring the image quality will be quite detrimental if in future any iterative based algorithm is applied over larger epoch values. it is quite a difficult task to rate the effectiveness of any presented research work just on the basis of experimental data. The dataset Dresden Image dataset is reputed for its benchmarking capabilities, but quite a few of the existing research techniques have used them. V. CONCLUSION The fundamentals of image-forensics domain and its detection technology are discussed. Types of image forensic and forgeries techniques are highlighted. We represented the study points and review on image forensic and its existing techniques to identify forged images. Various forms of digital forensic tools, ongoing research on different techniques are listed. Copy move forgery and its tampering techniques have been discussed. We represented the review on Copy Move Forgery and its detection techniques. At last we highlighted the research gap in the field of image forensics which are not done so far. VI. FUTURE WORK 419

7 Our future research direction will be towards addressing this research gap. We propose the innovative solutions the address these issues. VII. ACKNOWLEDGEMENTS Our sincere thanks to Dr. K.G. Vishwanath, Principal and Director, Jain College if Engineering, Belagavi, for the encouragement. REFERENCES Peng B, Wang W, Dong J, Tan T (2017) Optimized 3D LightingEnvironment Estimation for Image Forgery Detection. IEEE TransInf Forensics Secur 12: Lv Y, Shen X, Chen H (2011) An improved image blind identificationbased on inconsistency in light source direction. J Supercomput58(1): Saleh SQ, Hussain M, Muhammad G, Bebis G (2013) Evaluation ofimage Forgery Detection Using Multi-scale Weber Local Descriptors.International Symposium on Visual Computing, pp Carvalho T, Faria FA, Pedrini H, Toreres RS, Rocha A (2016)Illuminant-based transformed spaces for image forensics. IEEETrans Inf Forensics Secur 11(4): Li L, Xue J, Wang X, Tian L (2013) A robust approach to detect digital forgeries by exploring correlation patterns. Pattern Anal Appl18(2): Forensics Secur 7(5): Gaborini L, Bestagini P, Milani S, Tagliasacchi M, Tubaro S (2015)Multi-clue image tampering localization IEEE Int Work InfForensics Secur WIFS pp Chierchia G, Poggi G, Sansone C, Verdoliva L (2014) A bayesian-mrf approach for PRNU-based image forgery detection. IEEE Trans InfForensics Secur 9(4): Yu L, Han Q, Niu X, Yiu SM, Fang J, et al. (2016) An improved parameter estimation scheme for image modification detectionbased on DCT coefficient analysis. Forensic Sci Int 259: Bianchi T, Piva A (2011) Analysis of non-aligned double JPEGartifacts for the localization of image forgeries. IEEE Int Work InfForensics Secur WIFS. 13. Kaimal AB, Manimurugan S, Anitha J (2013) A modified anti-forensic technique for removing detectable traces from digitalimages Int. Conf. Comput. Commun. Informatics IEEE p Cao G, Zhao Y, Ni R, Tian H (2010) Anti-Forensics of Contrast Enhancement in Digital Images. ACM Multimed Secur Work p Fan W, Wang K, Cayre F, Xiong Z (2014) JPEG anti-forensics with improved tradeoff between forensic undetectability and imagequality. IEEE Trans Inf Forensics Secur 9(8): Fan W, Wang K, Cayre F, Xiong Z (2013) JPEG anti-forensics using non-parametric DCT quantization noise estimation and naturalimage statistics. Proc first ACM Work Inf hiding Multimed Secur -IH&MMSec 13: Fang Z, Wang S, Zhang X (2009) Image splicing detection using camera characteristic inconsistency. Multimed Inf Netw SecurMINES 1: Jiang Y, Zeng H, Kang X, Liu L (2013) The game of countering JPEGanti-forensics based on the noise level estimation. In: 2013 Asia- Pacific Signal Inf. Process. Assoc Annu Summit Conf IEEE p Kirchner M, Fridrich J (2010) On detection of median filtering indigital images. Media Forensics Secur II, USA, p Agarwal S, Chand S, Skarbnik N (2016) SPAM revisited for medianfiltering detection using higher-order difference. Secur Commun Networks 9(17): Shashidhar TM, Dr. KB Ramesh, Reviewing the Effectivity Factor in Existing Techniques of Image Forensics, International Journal of Electrical and Computer Engineering (IJECE), Vol. 7, No. 6, December 2017, pp. 3558~ C.Rajalakshmi, Dr.M.Germanus Alex, study of image tampering and review of tampering detection techniques, International Journal of Advanced Research in Computer Science, Valume 8, july-august