Fingerprinting using JPEG Quantization Tables
The paper Digital Image Ballistics from JPEG Quantization by Hany Farid describes how digital stills can be attributed to camera makes due to certain differences in the implementation of JPEG compression.
Most digital cameras produce stills in JPEG format. JPEG is a lossy data compression algorithm. The algorithm uses a quantization table which allows manufacturers to fine-tune quality and file size of an image.
Harid examines stills taken by 204 different makes of digital cameras. After comparing the quantization tables he is able to identify 62 unique models. The rest groups into clusters of 2 to 28 models, whereas models of the same manufacturer and product line tend to form a group.
According to Hamid quantization tables used by image processing software significantly differ from those used by cameras. This could be exploited to prove editing of a digital image.
