April 23, 2007 HelpLine
Q) An interesting problem came up recently with regards to evidence of bylaw contraventions in Pretoria, City of Tshwane, South Africa. The City Councils officials have switched to digital cameras for much of their work. However, it turns out that in court proceedings, the magistrates/judges aren't prepared to accept digital photographs as evidence because of the possibility of "fabricating" evidence. This is a valid point, and they want photographic film evidence, either positives or negatives. The question then is whether there's any photographic processing done in-camera, which leaves a watermark (maybe we should call that a "digimark"), which will indicate when pixels in the original have been removed or added, as opposed to colour, gamma and highlight or shadow adjustment to improve the original photo WITHOUT altering any evidence that might be required in court. If not available yet, this is certainly worthwhile exploring.
Prof. Duncan Baker, Pretoria, South Africa
A) Of course, mandating the use only of film is making an assumption that film can't be used to fabricate evidence. Anyone who has played with double exposures can break that assumption pretty easily. But while your City Councils aren't alone in requiring the use of film, there are also many locales that use digital photography for evidence gathering.
Some digital SLR manufacturers have implemented systems to help ensure images are secure. There's technology that exists to help ensure images are secure. For example, Canon's system pairs sophisticated data security with image-verification technology. This method locks up images so they can be displayed only with secure software, and it can verify if the image has been modified. Special image-verification data is embedded into the image when the digital SLR captures the image. More information can be found at:
Lexar has its LockTight system that works with Nikon equipment. It doesn't have the image-verification system, but it does use security to minimize access to images on a card. Details on that system can be found at:
On the software side of things, I know there are people who are currently investigating algorithms that will attempt to determine if a photographic image has been altered. By evaluating pixels and comparing organic vs. artificial pixel alignment, it might be possible to detect "artificial" pixels without knowing what the original data looked like. It's a long way off, and there are great difficulties with it.