CAMID: An assuasive approach to reveal source camera through inconspicuous evidence

Abstract

Identifying the source camera of digital images is an important problem due to the abundance of such images nowadays. Although photo response non-uniformities (PRNUs) can accurately identify the source camera, seam-carving techniques have been introduced to hide the cameras’ identity by invalidating the positional correspondence assumption in the PRNU. Especially, the seam carving techniques create irregular carving patterns, which makes source camera identification from seam-carved images challenging. In this paper, we propose a feature-based method called CAM1D, to identify the source camera of seam-carved images. CAM1D constructs a series of camera signatures to enhance robustness and reliability, rather than relying on a single PRNU for identification. Useful features are extracted from the correlation between a series of camera signatures and seam-carved photos, giving high source camera identification accuracy. Experimental results using the confusion matrix and F1 score confirm the effectiveness of CAM1D.