On 05/28/10 02:07 PM, Jeffrey Hutzelman wrote:
--On Tuesday, May 25, 2010 01:33:45 PM -0700 Bart Smaalders
<[email protected]> wrote:
Note that package manifests contain an FMRI, which includes the
timestamp
of the date of publication. This is part of the hash text
for the signature, and thus cannot be modified w/o invalidating
all the signatures.
You mean, the timestamp the person creating the signature wishes you
to _believe_ was the date of publication. The two are not necessarily
the same thing. Checking that the key used to create a signature was
valid on some date contained in the signed content is the same as not
checking the expiration at all, because _the signer can lie about the
timestamp_.
Yes, and if you don't trust the publisher, who is the signer, not to
fake the timestamp, don't add them as a publisher to your image because
clearly they shouldn't be trusted to deliver content to your system
either. Note that a third party cannot change the timestamp without
invalidating the publishers original timestamp.
To be clear, the act of trust is adding a publisher. The purpose of a
signature is to verify that the package has arrived from the publisher
unmodified.
Expired certs need to be used to check old packages; as long as the
package was signed when the cert was valid we will accept the cert
for that package version.
And how do you know the package is old, as opposed to merely claiming
to be old. How do you know the package was not created yesterday by
an attacker using an expired key which he has compromised?
If a cert's key has been compromised, that certificate should be
revoked, which is different than expiring due to time.
Brock
It needs to be possible to install Solaris 11 many years after FCS from
original media w/o having to ignore signatures.
Yes, it does. But the solution to that problem is not to simply
ignore expiration entirely.
-- Jeffrey T. Hutzelman (N3NHS) <[email protected]>
Carnegie Mellon University - Pittsburgh, PA
_______________________________________________
security-discuss mailing list
[email protected]
_______________________________________________
security-discuss mailing list
[email protected]
