Artikkel om bildesikkerhet

[mad_arab]

Leste denne artikkelen, og jeg synes såpass mye av det var interessant at jeg quoter en hel del av det. Det dreier seg om sikkerhet i bilder, og hvor farlig det faktisk kan være å legge ut bilder hvis man ikke ønsker å bli identifisert. Metadata kan røpe mye om kamera, fotograf og til og med sted. Thumbnail kan være umodifisert på et bilde som du har forandret, og i tillegg går det faktisk an å spore bilde til et _spesifikt_ kamera ved å analysere det, for deretter å identifisere hvor det ble kjøpt og muligens av hvem. Det kunne kanskje vært ideer for en artikkel på akam.no?

 

 

http://netzreport.googlepages.com/hidden_d...jpeg_files.html

 

Among other things, the metadata section of a file can contain information about:

make and model of the digital camera

time and date the picture was taken

distance the camera was focused at

location information (GPS) where the picture was taken

small preview image (thumbnail) of the picture

firmware version, serial numbers, name and version of the image manipulation program, etc. ...

 

[...]

 

Many digital cameras embed a small preview image (thumbnail) of the picture in the header of each JPEG file. This makes it possible to quickly browse the pictures. Not all image manipulation programs update this thumbnail along with the main picture. [...] For example, a supposedly anonymized picture of a person still shows his or her identity in the thumbnail. Another, more embarrassing example is the case of television personality Cat Schwartz (e.g. TechTV). Schwartz had published a photograph of herself on her personal blog. Because the program she had used to edit the picture did not update the thumbnail, the thumbnail revealed more nude facts than originally intended.

 

The following real-life case happened in February 2006: The Washington Post published an interview with a computer hacker: Invasion of the Computer Snatchers. The hacker had agreed to be interviewed only if he was not identified by name or hometown. In addition to the interview, a disguised picture of the hacker was published. Unfortunately, the picture contained IPTC metadata about the city and state where it was taken.

 

[...]

 

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.