0

I'm building a website on which I want to make it very easy for users to create pages anonymously and later be able to come back and edit those pages. One way to do so is to give out confidential tickets which are derived from the page ID to those users when they create the page. Later I would be able to derive the page ID from the ticket and allow a user in possession of the ticket to edit that page.

Of course this needs to be secure, so: 1) third parties shouldn't be able to derive the page ID from a given ticket, and even more importantly: 2) third parties should not be able to predict the ticket for a given page ID.

I could of course easily do this by randomly generating the tickets and storing them in a database, however I would like to try and do this without having to store the tickets.

In other words I'm looking for something like this:

  1. Let A be a unique public page ID.
  2. Let S be some kind of unchanging master key or other secret I possess and don't give out.
  3. Using S, transform A to B, such that I can later transform B back to A, but someone not in possession of S cannot.
  4. It should not be possible, or be very hard, to derive S even if you have a large collection of A-B pairs.

Does such a process exist? One thing I thought of is a simple XOR, however that would be easy to crack with just a few examples of A-B pairs. I can't use a hash, because they are intentionally irreversible; I can't derive A from B if B is a hash of A.

|improve this question
  • Is keeping ID secret from the holder of a ticket a requirement? That is, if I create a page and get a ticket, do you want the ID to be hidden from me or is it not a problem if I know it (as long as I can't create a valid ticket, of course)? – Neil Smithline Jun 10 '15 at 16:25
  • @NeilSmithline No, the page ID is public. – Pepijn Schmitz Jun 10 '15 at 16:29
3

This sounds like the perfect use of an HMAC. You create a secure random secret S. Then the ticket for each page will be ID + HMAC(S, ID). When you get a ticket you extract the ID, redo the HMAC and then compare your result with that in the ticket.

While this is likely be simpler and faster than an encryption solution, it will only work if you don't mind ID being in clear-text in the ticket.

|improve this answer
  • If the ticket has the ID encrypted symmetrically (or just uses an AEAD scheme to get authentication and ID together), that would mitigate the cleartext. – cpast Jun 10 '15 at 16:17
  • Yes @cpast. Symmetric encryption does hide the ID. But it may not be needed. I think the question is a bit unclear on this point. – Neil Smithline Jun 10 '15 at 16:23
  • Ah yes of course, I should have thought of that. Thanks! – Pepijn Schmitz Jun 10 '15 at 16:26
  • I don't think I need the page ID to be hidden (I can't think why, it wouldn't prevent any attack I can see; the page ID would be in the URL so it is public anyway), so I don't think the ID being in cleartext is a problem. – Pepijn Schmitz Jun 10 '15 at 16:28
  • 2
    @PepijnSchmitz: You did say in your question that third parties shouldn't be able to derive the page ID from a given ticket - you should remove this requirement from your question if it does not apply. Also consider adding an expiry date or some other method of revoking a ticket should one get exposed accidentally (e.g. page version number). The HMAC should also be calculated over this value. – SilverlightFox Jun 11 '15 at 10:20
1

Looks to me like standard public-key crypto should work. For example, using RSA (with public key N=p*q and e, private key d):

  • S would be your private key d.
  • Transforming: B = A^e mod N.
  • Transform back: B^d = B mod N.

Your requirements seem to then naturally fit the security assumptions of RSA. The one wrinkle here is that anyone can compute B from a given A (which violates your original statement, but not the formal requirements below that). I think this can be fixed by keeping e private as well (and sufficiently large and random), but that would require more careful analysis.

|improve this answer
  • Asymmetric encryption seems unnecessary here. Why is it better than symmetric encryption to solve the poster's problem? – Neil Smithline Jun 10 '15 at 16:21
  • It sounds like an HMAC would be a better fit. Unless you think this might perform much better? – Pepijn Schmitz Jun 10 '15 at 16:25
  • 1
    I guess it depends on the encryption you use in the HMAC @PepijnSchmitz, but symmetric encryption is usually orders of magnitude faster than asymmetric encryption. – Neil Smithline Jun 10 '15 at 16:38
  • Yeah ... if you can do it with symmetric encryption, you're better off. – Ari Trachtenberg Jun 10 '15 at 18:23

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.