Source Camera Identification Techniques: a Survey

EasyChair Preprint no. 10586

Abstract

Successful investigation and prosecution of major crimes like child pornography, insurance claims, movie piracy, traffic monitoring, and scientific fraud among others, largely depends on the availability of water-tight evidence to prove the case beyond any reasonable doubt. When the evidence required in investigating and prosecuting such crimes involves digital images/ videos, there is a need to prove without an iota of doubt the source camera/device of the image in question. Much research has been reported to address this need over the past decade. The proposed methods can be divided into brand or model-level identification or known imaging device matching techniques. This paper investigates the effectiveness of the existing image/video source camera identification techniques, which use both intrinsic hardware artifacts-based techniques like sensor pattern noise, lens optical distortion and software artifacts-based techniques like colour filter array, and auto white balancing, to determine their strengths and weaknesses. Publicly available benchmark image/video datasets and assessment criteria to quantify the performance of different methods are presented and the performance of some of the existing methods is compared. Finally, directions for further research on image source identification are given.

Keyphrases: Camera brand source identification, Camera color filter array, Camera model source identification, Image lens optical distortion, sensor pattern noise, Source camera identification

@Booklet{EasyChair:10586,
  author = {Chijioke Emeka Nwokeji and Akbar Sheikh Akbari and Antoliya Gorbenko and Iosif Mporas},
  title = {Source Camera Identification Techniques: a Survey},
  howpublished = {EasyChair Preprint no. 10586},

  year = {EasyChair, 2023}}