Localization of Tampering Created with Facebook Images by Analyzing Block Factor Histogram Voting

Localization of Tampering Created with Facebook Images by Analyzing Block Factor Histogram Voting

Archana V. Mire (Sardar Vallabhbhai National Institute of Technology, Surat (SVNIT), Surat, India), Sanjay B. Dhok (Visvesvaraya National Institute of Technology (VNIT), Nagpur, India), Narendra. J. Mistry (Sardar Vallabhbhai National Institute of Technology, Surat (SVNIT), Surat, India) and Prakash D. Porey (Visvesvaraya National Institute of Technology (VNIT), Nagpur, India)
Copyright: © 2015 |Pages: 22
DOI: 10.4018/IJDCF.2015100103
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Facebook images get distributed within a fraction of a second, which hackers may tamper and redistribute on cyberspace. JPEG fingerprint based tampering detection techniques have major scope in tampering localization within standard JPEG images. The majority of these algorithms fails to detect tampering created using Facebook images. Facebook utilizes down-sampling followed by compression, which makes difficult to locate tampering created with these images. In this paper, the authors have proposed the tampering localization algorithm, which locates tampering created with the images downloaded from Facebook. The algorithm uses Factor Histogram of DCT coefficients at first 15 modes to find primary quantization steps. The image is divided into BXB overlapping blocks and each block is processed individually. Votes cast by these modes for conceivable tampering are collected at every pixel position and the ones above threshold are used to form different regions. High density voted region is proclaimed as tampered region.
Article Preview

1. Introduction

Facebook (FB) provides plug-in for face detection, but doesn’t check tampering present in the image. Due to the absence of tampering localization applications, the probability of an image being tampered increases. Tampering consists of covering or embedding false realities, where a segment from the source is replicated to the destination image. The destination image which undergoes aligned double compression is called as a background image. As 8X8 DCT grid cannot be aligned to the primary compression grid in tamapered region, it lacks double compression artifacts. JPEG based algorithms search double compression artifacts and declare the absence as a tampered region. There are different fingerprints recommended for double compression artifacts based on the DCT coefficient histogram. When an image is uploaded on Facebook, EXIF information shows that the current antiquities such as ownership, copyright, site, date, camera, lens, etc. are removed. Facebook provides a high quality uploading option which compresses images without changing the resolution. Without this Facebook option, an image undergoes a down sampling followed by compression, which removes prior compression artifacts and creates new. Single compression artifacts are removed by down sampling process and are introduced again by recompression. If a person downloads an image from the Facebook, tampers and recompress, then double compression artifacts get introduced in the untampered region. JPEG based algorithms fail to detect tampering where primary compression quality is greater than secondary (Lin, He, X. Tang, & C. Tang,2009; Farid, 2009; Bianchi & Piva, 2010; Bianchi & Piva, 2011; Bianchi & Piva, 2012; Zach, Riess, & Angelopoulou, 2012). Facebook downsamples and compresses the images with high quality while maintaining its visual effects. JPEG based tampering localization algorithms fail here due to this downsampling and high compression quality. Up to the best of our knowledge, tampering made with Facebook images has not been addressed by any of the JPEG based algorithm.

Previous work in this area includes, Farid (2009) who recompressed the double compressed image at different quality level and subtracted it from the same. He found the difference minimum, at the quality level of primary compression. This effect, called as ghost effect, identifies the primary compression quality of a double compressed image. As compression grid used for tampering cannot be aligned, leads to higher difference in tampered region with respect to untampered region, at primary compression quality. This high difference region can be used to identify tampering present in double compressed image. Zach et al. (2012) proposed the method to check difference images and located tampering present in the double compressed image. Lin et al. (2009) plotted histogram of double quantized and single quantized coefficients. They identified the periodicity of empty and over concentrated bins in the histogram of double quantized coefficients. They computed period of the histogram and figured out tampering posterior probability map for every 8 X 8 block. Bianchi & Piva (2010) proposed that the coefficients of double compressed images have a tendency to cluster at the DCT grid position of primary and secondary compression. They plotted DC coefficient histogram by shifting grid position at different location within 8 X 8 blocks and computed integer periodicity map from Discrete Fourier Transform of histograms.

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 10: 4 Issues (2018): 1 Released, 3 Forthcoming
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing