EXIF-white balance recognition for image forensic analysis

Abstract

Due to the lack of post-processing resistance, traditional forensic methods are vulnerable to cascade image manipulations, e.g. copy-and-paste operation followed by high compression. Different from these traditional methods, a new forensic method that has the ability to resist multiple types of post-processing, is proposed by using white balance from the EXchangeable Image File format (EXIF) header. We first extract image quality metrics between each two combination of one original image and twelve re-balanced images. By regularizing the eigen spectrum of image quality metrics, the compact set of image eigen features is then selected for recognizing different EXIF-white balance modes via the SVM classifier. The experimental results show that the proposed method has the ability to resist the influence of high compression or heavy downsampling in both theoretical and realistic scenarios. Furthermore, thanks to image eigen features affected by cascade image operations, it is possible to lead to a wrong white balance mode. Thus, we use the EXIF-white balance parameter as a manipulator indicator for forgery detection. Based on the forgery photos in practice, the proposed evidence can detect cascade manipulated images which are subject to copy-and-paste followed by different white balance post-processing operations, high compression or heavy downsampling.