| Title |
Building Camera Source Identification Image Database through Croudsourcing |
| Authors |
길민호(Minho Gil) ; 강상욱(Sang-ug Kang) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.3.61 |
| Keywords |
Digital forensics; Image database; Source camera identification; Photo response non uniformity; Convolution neural network |
| Abstract |
As the convenience and utilization of digital cameras increase, various studies are being conducted to identify source cameras used for shooting from captured images. However, as deep learning technology has begun to be used in this field, there is a lack of image databases necessary for experiments to develop or verify the proposed algorithm. Moreover, the existing image database does not support the latest camera model or only provides data on one of digital camera and smartphone camera, and thus lacks reality. In this paper, we intend to show the validation results of the Sangmyung Image Database (SMDB) collected through the crowdsourcing and share data. SMDB consists of about 12,000 images taken with 9 digital cameras and 26 smartphone cameras of the latest model, and is divided into Flatfield, Dark, Natural images, which can be used to analyze noise pattern of image sensors or identify source cameras. In addition, SMDB shows similar levels of camera source identification accuracy to existing image databases such as FODB, VISION, and Dresden, even though collected by an unspecified number of ordinary people through crowdsourcing, satisfying both the reliability of the procedure and the accuracy of the results. |