Detectable Tampering of JPEG Anti-Forensics

Inhalt:
Many forensic techniques recently tried to detect the tampering and manipulation of JPEG compressed images that became a critical problem in image authentication and origin tracking. Some techniques indicated that a knowledgeable attacker can make it very hard to trace the image origin, while others indicated that portions of the compressed image that has been compressed at different quality factor quantization matrices are distinguishable if they are recompressed at a higher quality factor quantization matrix (with less quantization steps). In this paper, we pursue the idea of recompressing forensically suspect-able images with different compression parameters. We use different quantization matrix sizes that would indicate a DCT projection at different frequencies (horizontally, vertically, and diagonally), and would make it easier to track any tampering or hacking footprints. We show that a JPEG compressed image can make these footprints distinguishable if recompressed with a smaller size quantization matrix. Illustrative examples are presented.