Unlike the certificate, which can be deleted, revoked, or in some other way
removed from the CA's database and the server, a certificate cannot be
removed from a CT log, meaning it's impossible to delete PII in compliance
with the EU directive. 

-----Original Message-----
From: Trans [mailto:[email protected]] On Behalf Of Ben Laurie
Sent: Thursday, November 3, 2016 11:31 AM
To: Peter Bowen <[email protected]>
Cc: Melinda Shore <[email protected]>; [email protected]; Stephen Farrell
<[email protected]>
Subject: Re: [Trans] Topicality

On 3 November 2016 at 16:36, Peter Bowen <[email protected]> wrote:
> On Thu, Nov 3, 2016 at 8:45 AM, Ben Laurie <[email protected]> wrote:
>> On 1 November 2016 at 01:04, Peter Bowen <[email protected]> wrote:
>>>
>>> There are certificates that have personal information (e.g. given 
>>> name, surname, and physical location) in the subject distinguished 
>>> name or the subject alternative names.  It is very possible that 
>>> there may be a desire to redact this information (in fact it could 
>>> even be required in some jurisdictions as CT could be considered a
database).
>>> We already see this with domains in many countries where the full 
>>> registrant details are not publicly available.
>>>
>>> Additionally, while 6962-bis does focus on certificates used for TLS 
>>> server authentication, there are other types of certificates that 
>>> could be logged.  For example a certificate that is used with email 
>>> and contains both the email address and given/surname.  It might be 
>>> that the owner of the certificate only wishes to disclose this 
>>> binding to people receiving email from them rather than publicly 
>>> disclosing it.  We have also seen requests for certificates that 
>>> cover phone numbers.  A similar situation would apply -- having an
"unlisted"
>>> number should be possible, where the subject details (other than 
>>> phone
>>> number) are only known to a group of people who communicate with the 
>>> number.
>>
>> For these use cases, something CONIKS-like is probably more appropriate.
>
> How does CONIKS doesn't solve the problem of server authentication 
> certificates that identify an individual as the certificate 
> subscriber?

Sorry, I was referring to the latter cases - e.g. S/MIME (which is what you
seem to be describing).

>  Even if we assume for the moment that 6962bis is only for 
> "traditional" PKI  and only for server authentication, these certs are 
> still problematic to log unredacted.

Since the personal info in those cases is not relevant to browsers
validation of the cert, in principle I can't see why we would care if it
were redacted. OTOH, the certificate is public anyway, so not sure why
redacting it in the log is useful.

The real problem is the certificate, not the log entry!

_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

Reply via email to