Article Preview
TopIntroduction
Nowadays, the widely applied digital imaging devices bring great convince to the people in daily life. At any time, people can capture scenes around them by the portable cameras or the built-in camera in the mobile; the government can achieve 24-hour surveillance by the widely installed CCTV; the journalists can records the 1/24-second-motions by the professional camera. However, the security of the captured digital images remains unprotected and such problem needs urgently investigation by the research and the engineer. The security problem can be summarized as which person/device produces the image and whether the image is modified. As a result, the digital forensic techniques for digital images are developed with the origin identification and integrity verification functions in order to solve the aforementioned problems.
Generally speaking, the forensic techniques extract a fingerprint, which is a digital feature left by the digital imaging device, and compared it to the reference fingerprints representing a set of imaging devices in the database (Dirik, Sencar, & Memon, 2008; Fridrich, 2009; Caldelli, Amerini, Picchioni, De Rosa, & Uccheddu, 2009; Cao & Kot, 2010). Depending on the comparing result, the forensic techniques can identify the origin and verify the integrity of the digital images (Lukas, Fridrich, & Goljan, 2006a; Lukas, Fridrich, & Goljan, 2006b). The framework of forensic techniques is illustrated in Figure 1. The scheme first extracts the digital fingerprint 
of image 
, and then 
is compared to fingerprint 
in the fingerprint database. To identify the origin of the image, the fingerprint database contains the reference fingerprints of a variety of imaging devices. If a correlation between a reference fingerprint and 
is maximal among all correlations and greater than a predefined threshold, then the corresponding device is identified as the source device of the image under investigation. To verify content integrity, the fingerprint database consists of fingerprints on different areas of the reference image. If a reference fingerprint on a special area results in a correlation lower than the threshold, then this area is identified as a forged area.
Figure 1. A general framework of NDHBDIA techniques