Why we need pixel data encryption (perceptual encryption), not the complete file encryption (conventional encryption)?

I am working on securing medical images, but I could not find the answer to this question. Is the complete encryption of the file (image pixel data+ metadata(header file)) better than encrypting only pixel data (distorted image) which preserve the header information (codec information) of the encrypted image?

In the literature, I found the following: Security mechanisms developed in the existing digital medical image systems are mostly based on the conventional encryption techniques such as DES, AES, IDEA, RSA. These mechanisms have been widely recognized as the standard multimedia encryption mechanisms. However, conventional mechanisms have obvious limitations on the multimedia data such as images shown in the following papers:

  1. Socek D, Magliveras S, C’ulibrk D, Marques O, Kalva H, Furht B. Digital video encryption algorithms based on correlation-preserving permutations. EURASIP J Inform Security 2007.
  2. Yoon, Ji Won, and Hyoungshick Kim. "An image encryption scheme with a pseudorandom permutation based on chaotic maps." Communications in Nonlinear Science and Numerical Simulation 15.12 (2010): 3998-4006.
  3. Pareek, Narendra K., and Vinod Patidar. "Medical image protection using genetic algorithm operations." Soft Computing 20.2 (2016): 763-772.
  • 1
    $\begingroup$ Well, have you read those papers? $\endgroup$ – fkraiem Dec 22 '16 at 13:58
  • 1
    $\begingroup$ @fkraiem Better not read the last two - they may leave unsuspecting readers - eh - confused. They use histograms and randomness testing to perform security analysis.The first paper at least clearly states the user requirements. I'd say: ignore these papers unless you have one of the requirements listed in that paper. Yacoub, there is an insurmountable mountain of crap papers; better assume they are for academic purposes (only). $\endgroup$ – Maarten Bodewes Dec 22 '16 at 14:05
  • 1
    $\begingroup$ Just from the title, I'd classify papers 2 and 3 bullshit, it is telling that they were not published in cryptography related venues. The first one looks very dubious as well, since "correlation-preserving" sounds like "leaks information about the plaintext" to me, though it might justify that with some specialized requirements. $\endgroup$ – CodesInChaos Dec 22 '16 at 16:01

The one "obvious limitation" of conventional encryption applied to multimedia data is that once encrypted, multimedia data no longer is compressible, or recognized as multimedia data (with metadata) by non-decrypting devices. This is solved by compressing before encryption, and accordingly decrypting before decompression and rendering. Confidentiality of multimedia data is a problem fully solved by conventional encryption.

There however might be practical reasons for perceptual encryption, including

  • We want the encrypted version to pass thru devices designed exclusively for multimedia data, including compressing perhaps with perceptually acceptable loss, or/and lacking error-correction.
  • We want a subset of the true data (e.g. metadata, sound but not image, a low-definition version..) to remain accessible by non-decrypting devices (which goes straight against the normal objective of encryption: that an adversary without key can not learn anything).
  • As an interesting specialization of the previous point, we might want that the non-encrypted low-definition version and the encrypted high-definition version share as much bandwidth as possible, which implies some perceptual model.

Pictures have EXIF data describing them. This data is in a text format. You could write a program that encrypts data with an ascii value greater than 127 while leaving text characters alone. Loop through each byte in the image file. If the byte x is greater than 127, replace it with 382-x.

But it might be easier/more secure to encrypt the entire image and store the details about the image in a text file or a database.

  • 2
    $\begingroup$ Are you suggesting to leave half the bytes of the file unchanged and use a simple replacement scheme on the other half? That is not encryption. $\endgroup$ – bmm6o Dec 23 '16 at 20:02
  • $\begingroup$ @bmm6o Yes, you are right, this is my point. I found that the people are doing what Russell Hankins, My question is why we need to do such thing is it a hardware requirement for PACs? $\endgroup$ – Yacoub Dec 26 '16 at 20:17
  • $\begingroup$ Which is better, complete/strong encryption of the entire file or partial/weak encryption of just the image data? Since this is medical, encrypt the file according to this standard, your choices are to either encrypt the file with strong encryption or the entire hard drive: sookasa.com/resources/HIPAA-encryption $\endgroup$ – Russell Hankins Dec 30 '16 at 4:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.