Tweet
Fingerprint of Digital Cameras Many users may also not know that digital cameras leave an individual fingerprint in each picture. This allows to reliably link pictures to the camera with which they were taken -- in much the same way that forensic examiners can link bullets to the gun that fired them.
Professor Jessica Fridrich and two members of her Binghamton University research team exploit the fact that every digital camera produces tiny imperfections (noise) within a picture. Each camera has a characteristic way of producing noise (even cameras of the same make and model) due to inevitable irregularities during the manufacturing process of the camera and its sensors. Although the digital noise is largely invisible to the human eye, the team around Fridrich have developed algorithms to analyze the noise and thus to determine the individual fingerprint. According to Fridrich, the technique is accurate 99.99 percent of the time. A limitation is that it requires multiple pictures taken by the same camera to determine the fingerprint; a single picture is not sufficient.
With the help of the fingerprint, it is possible to tell if a picture was taken by a certain camera. It is even possible to detect image tampering. While unchanged regions of a picture keep their digital fingerprint, regions that have been tampered with lose their characteristic noise. Even if a picture has been compressed to a smaller file size (e.g. to send it by email), the fingerprint remains detectable.
Whereas Fridrich needs multiple pictures for her analysis, a technique developed by Nasir Memon of Polytechnic University in Brooklyn requires only a single picture. Memon's technique relies on the fact that different digital camera manufacturers use different interpolation algorithms. An interpolation algorithm is used by digital cameras to give each pixel of a digital photograph the correct color. As these algorithms leave telltale traces in the pictures and vary from company to company, Memon can match a picture to a camera brand with an accuracy of 90 percent.
Software tools that are capable of removing digital fingerprints do not seem to exist.
Professor Jessica Fridrich and two members of her Binghamton University research team exploit the fact that every digital camera produces tiny imperfections (noise) within a picture. Each camera has a characteristic way of producing noise (even cameras of the same make and model) due to inevitable irregularities during the manufacturing process of the camera and its sensors. Although the digital noise is largely invisible to the human eye, the team around Fridrich have developed algorithms to analyze the noise and thus to determine the individual fingerprint. According to Fridrich, the technique is accurate 99.99 percent of the time. A limitation is that it requires multiple pictures taken by the same camera to determine the fingerprint; a single picture is not sufficient.
With the help of the fingerprint, it is possible to tell if a picture was taken by a certain camera. It is even possible to detect image tampering. While unchanged regions of a picture keep their digital fingerprint, regions that have been tampered with lose their characteristic noise. Even if a picture has been compressed to a smaller file size (e.g. to send it by email), the fingerprint remains detectable.
Whereas Fridrich needs multiple pictures for her analysis, a technique developed by Nasir Memon of Polytechnic University in Brooklyn requires only a single picture. Memon's technique relies on the fact that different digital camera manufacturers use different interpolation algorithms. An interpolation algorithm is used by digital cameras to give each pixel of a digital photograph the correct color. As these algorithms leave telltale traces in the pictures and vary from company to company, Memon can match a picture to a camera brand with an accuracy of 90 percent.
Software tools that are capable of removing digital fingerprints do not seem to exist.
Comment