Abstract: Feature-based camera model identification plays an important role in the toolbox for image source identification. It enables the forensic investigator to discover the probable source model employed to acquire an image under investigation. However, little is known about the performance on large sets of cameras that include multiple devices of the same model. Following the process of a forensic investigation, this paper tackles important questions for the application of feature-based camera model identification in real world scenarios. More than 9,000 images were acquired under controlled conditions using 44 digital cameras of 12 different models. This forms the basis for an in-depth analysis of a) intra-camera model similarity, b) the number of required devices and images for training the identification method, and c) the influence of camera settings. All experiments in this paper suggest: feature-based camera model identification works in practice and provides reliable results even if only one device for each camera model under investigation is available to the forensic investigator.

@inproceedings{Gloe:2009aa,
  publisher    = {Springer Verlag},
  author       = {Thomas Gloe and Karsten Borowka and Antje Winkler},
  url          = {http://www.springerlink.com/content/b21343w4h1587n08/},
  series       = {Lecture Notes in Computer Science},
  booktitle    = {Information Hiding, 11th International Workshop, IH 2009, Darmstadt, Germany, June 8-10, 2009, Revised Selected Papers},
  volume       = {LNCS 5806},
  year         = {2009},
  editor       = {Stefan Katzenbeisser and Ahmad-Reza Sadeghi},
  address      = {Berlin, Heidelberg},
  title        = {Feature-based camera model identification works in practice: results of a comprehensive evaluation study},
  pages        = {262--276},
}