An Improved JPEG Image Blocking Artifact Detector

Abstract

Abstract Sharing digital images on social media has become very common these days. People must check for authenticity to share images on social media websites. The shared image may be forged intentionally or unintentionally and can defame someone's reputation, leading to severe events, such as public riots. Thus, authentication of digital images which are posted on social media websites is of paramount importance. Our social media should be intelligent enough to check on these forged images such that no false information spreads around society. Many image forgery detection algorithms have been used by big social groups based on JPEG compression artifacts, but these may not work well in the presence of anti-forensics. JPEG compression is the most widely used standard in social media these days. Two important artifacts, quantization and blocking artifacts, are being exploited by various experts for forensic analysis. JPEG anti-forensic techniques clear away these artifacts to fool forensic detectors. This work presents a novel technique derived from the inter-block interdependence of DCT coefficients for ferreting out JPEG-blocking artifacts in the presence of anti-forensics. In the case of JPEG images, the cropping operation shifts the blocking artifacts within the block, changing the inter-block interdependence. We propose to take advantage of this change to ferret out the blockiness in an image that will help the forensic analyst detect forgery. The proposed method can detect blocking artifacts even if anti-forensic operations are applied and take the intelligence of social media to a step up. A set of different and reproducible experiments have been conducted over a large set of images. It has been observed that the proposed detector outperformed the existing ones in ferreting out blocking artifacts in altered (anti-forensically) JPEG images.

Keywords

• Blocking artifacts
• Image-forensics
• JPEG Compression