[Cryptography] But it's encrypted so it must be OK

[Ray Dillinger]

On 10/26/23 22:28, Peter Gutmann wrote:
> For those who can't read AlgorithmIdentifiers, that's 40-bit RC2, circa 1987.
> Being used today, in financial EDI.
>
> I suggested they just ignore the RSA part and brute-force the 40-bit key on
> each message.


Yet another example of putting the "Backward" in backward 
compatibility.  >:-(

"It was secure forty years ago, so it must still be working" is a normal 
thing to think.  That's how mature security technologies like steel 
locks work.  But cryptography isn't a mature technology.  It's closer to 
being a mature technology than it was in 1987, but we still value 
immediate performance more than we value security that will last the way 
a steel lock lasts. So computer capabilities continue to advance and 
invalidate the stuff we worked on five or ten or twenty years ago, and 
continues to blindside normal people who WILL ALWAYS think cryptography 
works like a normal security product - that if it's secure then it stays 
secure.

The "continues to work" assumption is the way normal people think. 
Cryptography doesn't actually fail in a way that alerts them to the fact 
that it's not working any more, unless someone has already stolen all 
their money.

This is the awe-inspiring responsibility of being on any standards 
committee, folks.  Whether you consider it a reasonable thing to expect 
or not, anything you standardize will be used and expected to be secure 
for the rest of your natural life and most of your children's, because 
that's the way normal people think about normal security products.  If 
that expectation is false, you can't explain the limitations in a way 
that anyone who's not a pro will ever remember to even think about. 
Anything with limitations that would need to be explained, shouldn't be 
standardized.

Bear

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.metzdowd.com/pipermail/cryptography/attachments/20231027/9560daa8/attachment.htm>