Development of a Robust Algorithm to Detect Multi-Cloning (Copy-Move) Forgeries

Abstract

Digital image forensic is an emerging discipline that signifies a never ending struggle against image forgery. In this modern Era of advanced computation, tampering of images can be effortlessly accomplished by a number of available economical editing software‟s like Adobe Photoshop, Corel Draw etc. This poses a need for establishing techniques in order to verify the integrity and authenticity of digital images. There are many types of image forgery; cloning/copy-move attack is one of them. Cloning forgery is specific type of tampering that uses a portion of original image as source to hide or duplicates certain features within the same image. This type of tampering is considered to be most advanced and has become researcher‟s point of interest in recent years. So the research in this dissertation is carried on this paradigm. The research carried out during this dissertation is divided in two phases. The first phase focuses on development of a novel and robust model, which can identify single and multiple blind cloning forgeries in a given image while second phase deals with comparison of developed model with previously developed techniques in the field of digital multi-cloning detection. The proposed methodology utilizes colored image unlike previously developed methods which worked on gray-scale. Local binary pattern label along with clustering is used to minimize false positive rate. The evaluation of the proposed method has been done on data set, MICC-F220, which is approved by IEEE transactions of information security and image forensics. All tests on algorithm and experiments have been carried out using MATLABR2012a. Afterwards, a comprehensive yet detailed comparison of previously developed copy-move methodologies with the proposed technique is presented. Parameters like time complexity, effects of post processing, rates, sensitivity, specificity & accuracy etc. are compared. A concluding summary of this Master‟s Thesis together with an outlook on future suggestion completes this work.