SDTP/doc/IAF

DRAFT - NO STANDARD - Purpose and specification of SDTPs Image area format (IAF) - NO STANDARD - DRAFT

The transmission of media files, such as images, videos or audio files, can be difficult since most used steganographic algorithms used with SDTP only allow for a low number of individual bytes. To overcome this limitation, IAF is transmitting information about obtaining and cropping the image without transmitting the actual image. While this is not useful if the existance, content or other intrinsic properties of the image itself have to remain confidential, it can be used to show someone an image (or a part of it) that is publically availibe, while not disclosing the fact that the image has been shown, who it showed to whom and which area of the image has been shown.

Here are some cases where IAF might turn out to be useful:

1) A secret location for meeting or deposing an object has to be determined. To do so, Alice sents the URL of a well-known image (say, from Wikimedia commons) and an area of the image to Bob, who downloads the images and views the part Alice marked.

2) Bob has to know how someone looks, and is not aware of the fact that this person has a publically available profile picture on Facebook. Alice does show Bob this image.

And some cases where using it is not a good idea:

1) No one exept Bob and Alice must be aware of the fact that Alice owns a certain object, but Bob has to know it's appearance. IAF would disclose the fact that Alice owns the object (or at least has a photography of it), while sending a compressed image directly using SDTP would be plausible deniable by Alice in case Bob tries to accuse her. (Notice, however, that Alice cannot prevent Bob from letting third persons know how the object looks)


A IAF dataset needs the following information:

- A unique identifier of the images location or more precisly, a valid instruction about a way how (almost) anyone can obtain the image file. In reality, this can be a URL to an anonymous FTP or HTTP server, an ISBN (possibly with page number) or another easily encodable ID.
- A description of the area to be cropped and shown. A possible way would be defining the width and height as two bytes (0-255) that are each a 1/256th of the entire width and height.
- Checksum: For digitally transmitted files, the checksum can be as simple as CRC-16 or similar algorithms. The objective is not to guarantee cryptographic integrity, which is done at a lower level. If the image is represented analog (e.g. printed to paper or on chemical film), the usage of a histogram is helpful. Aspect ratio or other easily measurable parameters are likewise useful. (Notice that this part is not techincally required, but it will improve the reliability a lot. It can also be helpful if a certain directory on a public FTP server contains multiple images and it has to be checked which one the actually meant one is)