Exposing digital image forgeries is a major challengefor photography authentication and forensics investigation, which aimsat proving the authenticity of digital photos. In this paper, weintroduce a digital image forensics approach based on color coherenceto exposing copy-paste-blur forgeries. We first discuss the inherentcolor coherence introduced by imaging pipeline, and then obtain severalstatistical features from the coherence characterization. Using the SVMclassifier, we reveal traces of copy-paste-blur tampering inphotographic forgeries. Experimental results indicate that the proposedmethod can effectively expose the copy-paste-blur forgeries andlocalize the tampered regions with high accuracy.