Another method of authenticating video images is to show scientifically that an image has not been altered or conversely that alteration has occurred. One such method has been the use of watermarking which modifies the digital video content by the superimposition of a watermark. As Erik Berg noted:[i]
Digital watermarks are generally used for one of three purposes – data monitoring, copyright protection and data authentication. Digital watermarking can be either visible or invisible to the viewer. A visible watermark is primarily designed to display a copyright notice, telling the viewer who owns the image rights. An invisible watermark can also be used to assert a copyright, but it is designed to catch those persons who would try to infringe on the copyright, by using the image without permission of the owner. Watermarks can also be used to authenticate the integrity of an image. When used for this purpose, a software program is used to calculate a unique number, based upon the image data. To establish specific ownership of the image, an encryption key assigned to the owner of the image can be used to encrypt the unique number generated from the image. The encrypted number is then inserted, by various methods, into the image itself.
The United Kingdom House of Lords Select Committee on Science and Technology recognized both the value of watermarking as part of an authentication scheme and some of the limitations:
Watermarks
Watermarks provide an extra level of security to an image in addition to an audit trail if they are added at source as the image is being captured by the camera. With a conventional image the watermark (eg an identifying code or logo) would need to be visible in the scene and may thus obscure other vital information. In a digital image it is possible to hide the watermark within the image data with a form of encryption: although the watermark can be present in all parts of the image (down to pixel scale), the image looks normal and the watermark can be viewed only with the appropriate decryption key. It would also be possible to encrypt the whole image so that it is meaningless to anyone viewing it without the appropriate equipment and decryption key.
Two types of invisible watermark are envisaged and are being developed by companies including CRL IBM, NEC and others (see IBM evidence for detailed description of watermarking technology). The first (a permanent watermark or ‘tattoo’) will help copyright owners monitor the use of their material. The watermark is hidden within the image and will remain with the image (or image segment) if it is copied, modified or otherwise processed (ie the watermark is almost impossible to destroy or remove from the image no matter what is done to it). For evidential use this would allow copyrighted materials to be identified, and it might also be useful for identifying the source of images which have undergone significant modification. The second type of watermark is the fragile watermark which is destroyed by any processing or modification attempt other than just viewing the image. This could have significant potential for authenticating images used in evidence: an image which should have a watermark but does not, or has a damaged one, would have low evidential value without other corroboration.
The main limitation of all watermarks is that ideally they should be applied by the camera and thus the camera manufacturers have to install software which will do this. Only if watermarking becomes the norm for cameras where images might be used for evidential purposes, or if there is overwhelming demand from customers, are manufacturers likely to adopt such technology. There are also problems with the export/import and use of certain levels of encryption technology which might be used to generate watermarks.
One key difficulty with watermarking is that the effect of inserting the mark into the image is to intentionally alter the image, albeit for altruistic reasons. Providing the watermark alteration is adequately explained, this should not pose a problem in court unless the watermark has obscured a key portion of the video image. A further concern is that some watermarking techniques can reportedly introduce frequency distortion or other visible image degradation[ii] and a number of papers have shown that watermarking is not a foolproof method of image authentication as there are techniques available for removing or disabling digital watermarks.[iii]
Indeed the Select Committee on Science and Technology expressed reservations in relying solely on a watermarking authentication scheme. The Committee stated:
3.8 Some witnesses, including the Association of Chief Police Officers in Scotland (p 154) and the Crown Prosecution Service (p 165) were in favour of images incorporating some form of watermark or other authentication (added at the time of image capture) to increase their utility in evidence. However many other witnesses (eg the Home Office and Forensic Science Service (p 128), Mr Castell (p 160), Justice (p 3) and the Chartered Institute of Arbitrators (p 156) held the view that all such technologies were fallible and were thus of limited use. Even if complex watermarking or high levels of encryption are used then, eventually, someone may crack the code and either remove a digital signature or re-instate a fragile watermark after making modifications to the image. Authenticating technologies must therefore keep ahead of possible abusers in order for them to remain of value.
Despite these reservations, there may be a legitimate role for watermarking in a thorough image authentication scheme. Counsel need to be aware of how watermarks are created and how they may properly be used to either prove authentication or confirm image alteration.
[i] Berg, E., The legal edge: digital imaging, a legal primer, The FBI Associate Magazine, 22 at 22-23, January/February 2001, at pages 23 and 40.
[ii] Kutter, M. and Petitcolas, F., A fair benchmark for image watermarking systems, presented at Electronic Imaging ’99, Security and Watermarking of Multimedia Contents, San Jose, CA, January 1999.
[iii] Supra, note 1, at endnote 16.