A robust forensic method based on Scale-Invariance Feature Transform

Nowadays, the digital forensic techniques for digital images are developed with the origin identification and integrity verification functions for security reasons. Currently, most developed methods are based on the idea of extracting high frequency noise as the feature for investigating purpose. However, these noise can be easily destroyed by the variety of the image processing manipulations, which are commonly applied in forging an image. Due to the weak robustness, these feature based forensic methods greatly limited in the practical applications. In this paper, we propose a content based forensic method by applying the Scale-Invariance Feature Transform (SIFT) to the images. This method can identify the copy-and paste areas by comparing the similarity between the contents in the two areas. Experiment result proves that this method is robust to rotation, cropping and resizing, add-in noise and blurring.