4 ms·

Is there any (valid) reason why someone would want to use their own hash function? Secure hash functions, as far as I am aware, are pretty established and widel

by weavie 9y ago

Is there any (valid) reason why someone would want to use their own hash function? Secure hash functions, as far as I am aware, are pretty established and widely available and really easy to use..

once_inc 9y ago
No. There is no valid reason to do so.
- DyslexicAtheist 9y ago
  if you understand the Math and agree that whatever you do needs to go through a proper peer review then why not. Saying "nobody may research the subject because you'll fail" is bad science. If you work in this space, know your Math, and don't rely on your invention to be put into production this year then why not study the problem space and innovate by following the established best practices!? Problems arise when someone thinks they should now turn this invention into a money-cow or label it "proprietary magic" (under the pretext of protecting IP/copyright).
  - akvadrako 9y ago
    > if you understand the Math and agree that whatever you do needs to go through a proper peer review then why not. Because crypto is really hard and you can't prove your new algorithm is secure. Only after being a popular high value target for years can you have much trust in your new hash. Even SHA2 might have flaws nobody knows about - that's the main purpose behind having SHA3 standardised - as a backup.¹ If you are not the designer of the algorithm it's even worse, because an exploit could be put in intentionally by proper choice of constants which are in some cases undetectable. [1] https://security.stackexchange.com/questions/152360/should-we-be-using-sha3-2017 https://security.stackexchange.com/questions/152360/should-w...
bb88 9y ago
Yes. But it would more likely be the case where they are encrypting the data before hashing it to reduce the chance for a hash collision.
DyslexicAtheist 9y ago
that knowledge while widely available within the engineering community isn't a given outside that circle. We're kind of arguing in a bubble. From a marketers pov they can sell it as "revolutionary proprietary PQ-proof, AI resistant and other made up bunch of words", and the masses buy it and don't care about it as long you tell them it's safe. The market today isn't controlled by engineering or common sense but by whoever shouts the loudest. And once people/companies commit to a certain technology (ideologically and/or financially) it'll be even harder to convince them of them betting on the wrong horse. Why bother with any solid (expensive) engineering practices when you can just raise money based on a whitepaper, entice the first investors by pointing to a hoard of shills all backing your warez, then use the noise generated by ICO promises to ward off any critics. So all you need to do is make your tech political and you can get away with anything thanks to the noise. Rolling their own broken crypto wasn't the only problem that made infosec community get outraged. Bypassing peer-review processes and then threatening with litigation and insulting researchers ("Hi Neha, are you drunk?") was all part of why nobody wants to touch them with a 10ft pole now. The worst of it is that none of this will stop IOTA from continuing on this road and they'll just play the same game that Trump plays with his hardcore supporter base. Forget about arguing with technical arguments because they'll just wear you down as the discussion with Matthew Green has shown. When their lead engineer thinks that a hash function doesn't have to be collision resistant then there is simply no point in wasting your time trying to make them see the problem (they won't)

by weavie 9y ago

Is there any (valid) reason why someone would want to use their own hash function? Secure hash functions, as far as I am aware, are pretty established and widely available and really easy to use..

once_inc 9y ago

No. There is no valid reason to do so.

DyslexicAtheist 9y ago
if you understand the Math and agree that whatever you do needs to go through a proper peer review then why not. Saying "nobody may research the subject because you'll fail" is bad science. If you work in this space, know your Math, and don't rely on your invention to be put into production this year then why not study the problem space and innovate by following the established best practices!? Problems arise when someone thinks they should now turn this invention into a money-cow or label it "proprietary magic" (under the pretext of protecting IP/copyright).
- akvadrako 9y ago
  > if you understand the Math and agree that whatever you do needs to go through a proper peer review then why not. Because crypto is really hard and you can't prove your new algorithm is secure. Only after being a popular high value target for years can you have much trust in your new hash. Even SHA2 might have flaws nobody knows about - that's the main purpose behind having SHA3 standardised - as a backup.¹ If you are not the designer of the algorithm it's even worse, because an exploit could be put in intentionally by proper choice of constants which are in some cases undetectable. [1] https://security.stackexchange.com/questions/152360/should-we-be-using-sha3-2017 https://security.stackexchange.com/questions/152360/should-w...

bb88 9y ago

Yes. But it would more likely be the case where they are encrypting the data before hashing it to reduce the chance for a hash collision.

DyslexicAtheist 9y ago

that knowledge while widely available within the engineering community isn't a given outside that circle. We're kind of arguing in a bubble. From a marketers pov they can sell it as "revolutionary proprietary PQ-proof, AI resistant and other made up bunch of words", and the masses buy it and don't care about it as long you tell them it's safe. The market today isn't controlled by engineering or common sense but by whoever shouts the loudest. And once people/companies commit to a certain technology (ideologically and/or financially) it'll be even harder to convince them of them betting on the wrong horse. Why bother with any solid (expensive) engineering practices when you can just raise money based on a whitepaper, entice the first investors by pointing to a hoard of shills all backing your warez, then use the noise generated by ICO promises to ward off any critics. So all you need to do is make your tech political and you can get away with anything thanks to the noise. Rolling their own broken crypto wasn't the only problem that made infosec community get outraged. Bypassing peer-review processes and then threatening with litigation and insulting researchers ("Hi Neha, are you drunk?") was all part of why nobody wants to touch them with a 10ft pole now. The worst of it is that none of this will stop IOTA from continuing on this road and they'll just play the same game that Trump plays with his hardcore supporter base. Forget about arguing with technical arguments because they'll just wear you down as the discussion with Matthew Green has shown. When their lead engineer thinks that a hash function doesn't have to be collision resistant then there is simply no point in wasting your time trying to make them see the problem (they won't)