Detection of JPEG image forgeries using passive-blind detection methods

Project leader: Mahdian
Supported by: Grant Agency of the Czech Republic, No. P202/10/P509.
Duration: 2010 - 2011


In today's digital age, images have become easy to manipulate and edit even for non-professional users. This is mainly due to the advent of low-cost, high-performance computers, more friendly human-computer interfaces (HCI), and the availability of many powerful and easy to control image processing and editing software packages. The digital information revolution and issues concerned with multimedia security have also generated several approaches to authentication and tampering detection. Generally, these approaches can be divided into active and passive-blind approaches. This project focuses on the blind image authentication approach, as it is regarded as a new direction and in contrast to the active approach, it works in the absence of any digital watermark or signature or any prior information about the image or its source. We focus on JPEG images as JPEG file format is the most widely used image format. Particularly, we propose methods for exposing forgeries by using JPEG quality analysis, measuring inconsistencies of blocking artifact, quantization tables analysis and source camera identification. Our aim is to minimize the successful abuse of forgeries in the never-ending game between image forgery creators and image forgery detectors by developing novel and more sophisticated mathematical and computational algorithms.