Abstract: Dirt particles on a digital camera's sensor may leave marks on images taken by the camera. These marks may be used to identify the camera used to take the picture. Our earlier work empirically determined the effects that various camera parameters (such as lens focal length and image compression) have on the position and appearance of such marks. The earlier paper suggested that a grid should be overlaid on the images to be compared. This paper analytically determines the minimum number of marks that should coincide in a test shot taken by the camera and the picture that needs to be linked to the camera. The minimum number of cells depend on the dimensions of the overlaid grid, and the statistical significance with which the match need to be made.

@article{Oliver:2008aa,
  author       = {Martin S. Oliver},
  journal      = {South African Computer Journal},
  volume       = {41},
  url          = {http://mo.co.za/forbib.htm},
  year         = {2008},
  title        = {On the probability of a false positive match for digital camera identification based on sensor dirt location},
  pages        = {21--28},
}