[Catalog-sig] Mandatory Reset of PyPI Passwords

[Donald Stufft]

On Wednesday, February 13, 2013 at 3:36 PM, Antoine Pitrou wrote:
> Donald Stufft <donald.stufft <at> gmail.com (http://gmail.com)> writes:
> > 
> > Why is it worse? SHA1 isn't terribly broken AFAIK.
> > 
> > Because you lower the available entropy, "birthday paradox". 
> 
> How so? Collisions are highly unlikely on a non-broken 160-bit hash function.
> I don't understand how the birthday paradox is a practical problem.
> 
> Regards
> 
> Antoine.
Sorry I was wrong about why. I asked the Security Researcher at work (I'm not
an expert, I just implement solutions the experts come up with ;) )

bcrypt(sha1(plaintext)) is bad because sha1 shouldn't be used because it's been
"broken". bcrypt(sha256(plaintext)) is better than just plain bcrypt(plaintext) because
because only considers a maximum number of characters (I believe it's in the 50's).

So basically bcrypt of a hash is secure as long as the hash is secure, but
sha1 shouldn't be considered secure anymore.

However Passlib doesn't have a bcrypt + hash backend and I would be loathe
to suggest PyPI permanently switch to a custom untested/not widely used
backend. 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/catalog-sig/attachments/20130213/c0289d4e/attachment.html>