Re: [websec] HPKP & different encodings of the same public key

[Alice Wonder <[e-mail]>]

On 05/15/2016 04:04 PM, Jeffrey Walton wrote:
>>>> can petition to get fixed), rather than spec bug (that we all have to
>>>> workaround).
>>>
>>> It depends on the issuing policies.
>>>
>>> The IETF has no way to specify that a certificate was created or
>>> issued under PKIX, so its a moot point. (It creates a vaccum like the
>>> EV mess, except for standard certificates rather than EV
>>> certificates).
>>
>>
>> HPKP is specified in terms of RFC 5280, so we can assume only PKIX
>> certificates are used for HPKP....
>
> In that case, the IETF provides a document on path building and
> validation (RFC 4158), but not certificate validation (modulo RFC
> 6125). As far as I can tell, its still the wild, wild west with no
> guidance on end-entity certificate validation.
>
> Jeff

DANE is better anyway.

HPKP is trust on first use, and even worse, trust on first use without
   user interaction.
DANE is validate on every use

HPKP only works with https
DANE works with anything tcp or udp

HPKP can be used for a DOS attack. Simple DNS cache poison and send the
   header with a bogus set of keypins and long TTL
DANE validates via DNSSEC avoiding that issue.

I do not understand browser resistance to DANE validation.