Re: [keyassure] Opening issue #21: "Need to specify which crypto

[Martin Rex <[e-mail]>]

Phillip Hallam-Baker wrote:
> 
> Plenty is not the point.
> 
> SHA1 is no longer regarded as acceptably secure. SHA2 is based on the
> same construction and is thus similarly suspect.
> 
> We don't use SHA1 in legacy browsers because we like it. We use SHA1
> in legacy browsers because we have not worked out how to transition
> away from it. The transition from MD5 to SHA1 was painless only
> because all browsers were required to support both standards.

Transition from MD5 to SHA1?  I seem to have missed that entirely.

MD5 was removed from the PRF in the SSLv3->TLSv1.0 protocol revision,
but even TLSv1.1 uses a combination of MD5+SHA1 for digital signatures.

And there are still a number of TrustAnchors in every Browser that
carry an md5WithRsaEncryption signature.

But it's actually worse than that -- VeriSign still has a CA cert
in service with an md2WithRsaEncryption -- and is still using it
productively (check https://www.hp.com/).


Considering the transition SHA-1 --> SHA-256.  That was goofed
in TLSv1.2, by defining the default hash algorithm for the digital
signature structs to default to SHA-1 (only) -- which is actually
weaker than SSLv3/TLSv1.0/TLSv1.1 ...


> 
> This particular topic is one on which the Security ADs and the IETF
> chair have very very specific opinions on. And given their role in
> trying to effect an industry wide transition to stronger algorithms, I
> think that they are quite right to insist on them.

DANEs security is vitally dependent on DNSSEC, and for DNSSEC, SHA-1
seems to be the mandatory-to-implement algorithm for signatures.


Inuitively, why don't we tie DANE's hash algorithm to DNSSECs algorithm?


Making DANE an early adopter of SHA-3 while DNSSEC still uses SHA-1
amounts to silicon snake oil -- i.e. needless complexity.


-Martin