Problem Statement

Background

More specifically, a comprehensive overview of four main types of forgery detection techniques such as image splicing, copy-move, resampling, and retouching detection is given.Various existing methods have been reviewed in each category and observed that existing techniques suffer from one or more following limitations.

• Detection Accuracy
• High computation complexity
• Vulnerable against various attacks such as rotation, scaling, JPEG compression, blurring, and brightness adjustment,etc.
• A lot of false matches with regular background.
Apart from abovementioned limitations, one major issue of these detection techniques is the limited scope of utilization, for example, method developed for copy move forgery cannot work with image splicing or resampling and vice versa. No detection method can be used as a solution for detecting all kind of forgeries.

Relevance

Hence, there is a great need to develop a robust, sophisticated forgery detection technique which could eliminate aforementioned limitations. Furthermore, researchers may extend these techniques to detect forgeries in videos. Automation of quality factor determination is a major future direction for this research. Restoration of forged JPEG regions will also be investigated in the future.

Objectives

In this work we propose a methodology to adjust the fashioned picture at the degree of small scale examples to trick a cutting edge fraud finder. At that point, we examine on the adequacy of the proposed system as a component of the degree of information on the fabrication identification calculation. We propose a strategy to modify the forged image at the level of micro-patterns to fool a state-of-the-art forgery detector. Then, we investigate on the effectiveness of the proposed strategy as a function of the level of knowledge on the forgery detection algorithm using ML and abstract concept’s of deep learning.