Abstract: A digitally altered image, often leaving no visual clues of having been tampered with, can be indistinguishable from an authentic image. The tampering, however, may disturb some underlying statistical properties of the image. Under this assumption, we propose five techniques that quantify and detect statistical perturbations found in different forms of tampered images: (1) re-sampled images (e.g., scaled or rotated); (2) manipulated color filter array interpolated images; (3) double JPEG compressed images; (4) images with duplicated regions; and (5) images with inconsistent noise patterns. These techniques work in the absence of any embedded watermarks or signatures. For each technique we develop the theoretical foundation, show its effectiveness on credible forgeries, and analyze its sensitivity and robustness to simple counterattacks.

@phdthesis{popescu05,
  school       = {Department of Computer Science, Dartmouth College},
  url          = {http://www.cs.dartmouth.edu/farid/publications/apthesis05.pdf},
  author       = {Alin C. Popescu},
  year         = {2005},
  title        = {Statistical tools for digital image forensics},
  address      = {Hanover, NH},
}