0

Can machine learning be used to improve encryption algorithms e.g.image encryption algorithms that use ECC ?

polkaadotskjm227
  • 11

1 Answers1

2

No. There is large consensus that:

  • There is nothing special about image encryption algorithm, they are just encryption algorithms.
  • ECC is useful for such algorithms inasmuch as they need to be made public-key. That's done by way of hybrid cryptography. Directly encrypting images with ECC cryptography would have terrible performance, thus is not worth serious consideration.
  • There's no clear need to improve any of these algorithms in a way where machine learning seems to be necessary or even helpful.

In a quite different direction: Machine learning also has a poor track record at attacking modern cryptographic algorithms, but that idea may not be worthless. Perhaps machines one day will learn cryptanalysis.

fgrieu
  • 140,762
  • 12
  • 307
  • 587
  • 1
    https://eprint.iacr.org/2020/957 seems a decent example of ML in attacking PUFs, though those aren't what I'd generally classify as cryptography. And https://eprint.iacr.org/2020/1235 tries to use it finding differentials in block ciphers, though it's not terribly good at it. – SAI Peregrinus Oct 15 '20 at 15:00
  • 1
    With regards to ML-aided attacks, this thread and the references cited therein seem worth mentioning: https://crypto.stackexchange.com/questions/72175/using-ai-to-perform-cryptanalysis .

    Automated tools of any kind that can do interesting cryptanalysis could also potentially be very helpful in pushing the design of e.g. ultra lightweight ciphers further. I would therefore not be quite as pessimistic as you about machines maybe some day learning to design ciphers.

     – Polytropos Oct 15 '20 at 22:01