5 ms·

Don't roll your own crypto. If that was much better encryption, everyone would be using it.

taylorfinley 8y ago
I think we all agree rolling your own crypto is dangerous, but what siliconunit described is just a one time pad. Assuming your key data is truly random and unknown to your attacker, isn't this kind of the gold standard for uncrackable cyphered communication?
- jpatokal 8y ago
  In theory, but getting that stream of "real random" and above all distributing that one time pad securely to the recipient (chicken, meet egg) so they can read the message make this highly impractical.
  - vectorEQ 8y ago
    dont need the random really. just need an offline method for exchanging the pad :D or just use an unexpected channel like radio or so to send it
- Eliezer 8y ago
  If interpreted that way, storing the noise and memorizing the offset, it amounts to having a privately stored one-time pad to be used as key... with a passphrase with as many bits in it as the offset. That’s probably not a lot of bits. You are better off storing the data, encrypted with a real pass phrase, wherever you would have stored the random noise stream that is needed to read the data in any case. Don’t roll your own crypto.
  - A2017U1 8y ago
    Using a one time pad isn't "rolling your own crypto". It is also provably secure unlike every other cryptosystem.
    - function_seven 8y ago
      Sure, except now you have an equal amount of data that needs to be stored somewhere. The pad is as long as the data. Either you have to encrypt that using some other means, or you do the weak "store offset" method.
penagwin 8y ago
I agree with not rolling your own crypto for your primary mechanism, layering it (in the correct order!) is guaranteed to be at least as secure as your strongest crypto. That's what google did with CECPQ1, they use "new hope" which is a quantum resistant algorithm with traditional methods (X25519). That way if new hope is cracked, they are still using an industry standard you'd have to crack as well.

I think we all agree rolling your own crypto is dangerous, but what siliconunit described is just a one time pad. Assuming your key data is truly random and unknown to your attacker, isn't this kind of the gold standard for uncrackable cyphered communication?

jpatokal 8y ago
In theory, but getting that stream of "real random" and above all distributing that one time pad securely to the recipient (chicken, meet egg) so they can read the message make this highly impractical.
- vectorEQ 8y ago
  dont need the random really. just need an offline method for exchanging the pad :D or just use an unexpected channel like radio or so to send it
Eliezer 8y ago
If interpreted that way, storing the noise and memorizing the offset, it amounts to having a privately stored one-time pad to be used as key... with a passphrase with as many bits in it as the offset. That’s probably not a lot of bits. You are better off storing the data, encrypted with a real pass phrase, wherever you would have stored the random noise stream that is needed to read the data in any case. Don’t roll your own crypto.
- A2017U1 8y ago
  Using a one time pad isn't "rolling your own crypto". It is also provably secure unlike every other cryptosystem.
  - function_seven 8y ago
    Sure, except now you have an equal amount of data that needs to be stored somewhere. The pad is as long as the data. Either you have to encrypt that using some other means, or you do the weak "store offset" method.

penagwin 8y ago

I agree with not rolling your own crypto for your primary mechanism, layering it (in the correct order!) is guaranteed to be at least as secure as your strongest crypto. That's what google did with CECPQ1, they use "new hope" which is a quantum resistant algorithm with traditional methods (X25519). That way if new hope is cracked, they are still using an industry standard you'd have to crack as well.