Digital Image Tamper Detection Based on Multimodal Fusion of Residue Features - Advanced Concepts for Intelligent Vision Systems - page 78

Digital Image Tamper Detection Based on Multimodal
Fusion of Residue Features
Girija Chetty 1 , Julian Goodwin 2 , and Monica Singh 2
1 Faculty of Information Sciences and Engineering,
University of Canberra, Australia
[e-mail]
2 Video Analytics Pty. Ltd. Melbourne, Australia
Abstract. In this paper, we propose a novel formulation involving fusion of
noise and quantization residue features for detecting tampering or forgery in video
sequences. We reiterate the importance of feature selection techniques in
conjunction with fusion to enhance the tamper detection accuracy. We examine
three different feature selection techniques, the independent component analysis
(ICA), fisher linear discriminant analysis (FLD) and canonical correlation analysis
(CCA) for achieving a more discriminate subspace for extracting tamper
signatures from quantization and noise residue features. The evaluation of
proposed residue features, the feature selection techniques and their subsequent
fusion for copy-move tampering emulated on low bandwidth Internet video
sequences, show a significant improvement in tamper detection accuracy with
fusion formulation.
Keywords: image tampering, digital forensics, feature selection, image fusion.
1 Introduction
Digital Image tampering or forgery has become major problem lately, due to ease of
artificially synthesizing photographic fakes- for promoting a story by media channels
and social networking websites. This is due to significant advances in computer
graphics and animation technologies, and availability of low cost off-the-shelf digital
image manipulation and cloning tools. With lack of proper regulatory frameworks and
infrastructure for prosecution of such evolving cyber-crimes, there is an increasing
dissatisfaction about increasing use of such tools for law enforcement, and a feeling
of cynicism and mistrust among the civilian operating environments.
Another problem this has lead to, is a slow diffusion of otherwise extremely
efficient image based surveillance and identity authentication technologies in real-
world civilian operating scenarios. In this paper we propose a novel algorithmic
framework for detecting image tampering and forgery based on extracting noise and
quantization residue features, their transformation in cross-modal subspace and their
multimodal fusion for intra-frame and inter-frame image pixel sub blocks in video
sequences. The proposed algorithmic models allow detecting the tamper or forgery in
low-bandwidth video (Internet streaming videos), using blind and passive tamper
detection techniques and attempt to model the source signatures embedded in camera