[P2PSIP] Mirja Kühlewind's No Objection on draft-ietf-p2psip-share-09: (with COMMENT)

Discussion:

Mirja Kuehlewind

2016-10-31 14:06:56 UTC

Mirja Kühlewind has entered the following ballot position for
draft-ietf-p2psip-share-09: No Objection

When responding, please keep the subject line intact and reply to all
email addresses included in the To and CC lines. (Feel free to cut this
introductory paragraph, however.)

Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
for more information about IESG DISCUSS and COMMENT positions.

The document, along with other ballot positions, can be found here:
https://datatracker.ietf.org/doc/draft-ietf-p2psip-share/

----------------------------------------------------------------------
COMMENT:
----------------------------------------------------------------------

Quick questions on sec 6.3. (Validating Write Access through an ACL):
Do I really need to validate the authorization chain in the ACL every
time I give access to a resource? Wouldn't I rather validate the ACL when
it's modified and then simply assume that it is sufficient that I have an
entry in the ACL to provide access?

Thomas C. Schmidt

2016-10-31 14:37:49 UTC

Permalink

Hi Mirja,

you are right in the sense that (a) if all previous evaluations have
been performed without a failure, and (b) if no revocation occurred (or
(c) a previous revocation has cleaned up all further delegation
entries), then the write procedure can rely on the single delegation
entry that matches the current user name of the writer.

However, this includes several "ifs". For instance, if cleanup of the
delegation list has not been completed at the time of granting write
access, errors in the trust chain may occur. This could introduce
unwanted attack surface.

Our rationale behind designing this complete, self-contained procedure
was (a) writing an ACL list is not a frequent operation (so complexity
is not the major concern), and (b) keeping all operations simple,
robust, and of minimal dependence w.r.t. each other.

That's why it's like that.

Cheers,
Thomas

Post by Mirja Kuehlewind
Mirja Kühlewind has entered the following ballot position for
draft-ietf-p2psip-share-09: No Objection
When responding, please keep the subject line intact and reply to all
email addresses included in the To and CC lines. (Feel free to cut this
introductory paragraph, however.)
Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
for more information about IESG DISCUSS and COMMENT positions.
https://datatracker.ietf.org/doc/draft-ietf-p2psip-share/
----------------------------------------------------------------------
----------------------------------------------------------------------
Do I really need to validate the authorization chain in the ACL every
time I give access to a resource? Wouldn't I rather validate the ACL when
it's modified and then simply assume that it is sufficient that I have an
entry in the ACL to provide access?

--
Prof. Dr. Thomas C. Schmidt
° Hamburg University of Applied Sciences Berliner Tor 7 °
° Dept. Informatik, Internet Technologies Group 20099 Hamburg, Germany °
° http://www.haw-hamburg.de/inet Fon: +49-40-42875-8452 °
° http://www.informatik.haw-hamburg.de/~schmidt Fax: +49-40-42875-8409 °

Mirja Kuehlewind (IETF)

2016-11-04 10:20:57 UTC

Permalink

Hi Thomas,

Post by Thomas C. Schmidt
Hi Mirja,
you are right in the sense that (a) if all previous evaluations have been performed without a failure, and (b) if no revocation occurred (or (c) a previous revocation has cleaned up all further delegation entries), then the write procedure can rely on the single delegation entry that matches the current user name of the writer.

This comes down to me to only one ‚if‘ and that is actually point c. And I’d hope that c would always happen.

Post by Thomas C. Schmidt
However, this includes several "ifs". For instance, if cleanup of the delegation list has not been completed at the time of granting write access, errors in the trust chain may occur. This could introduce unwanted attack surface.

Could you document this attack surface in the doc…?

Post by Thomas C. Schmidt
Our rationale behind designing this complete, self-contained procedure was (a) writing an ACL list is not a frequent operation (so complexity is not the major concern), and (b) keeping all operations simple, robust, and of minimal dependence w.r.t. each other.

Don’t you have to do the check every time you check write access for a shared resource? That can be much more often.

Mirja

Post by Thomas C. Schmidt
That's why it's like that.
Cheers,
Thomas

Thomas C. Schmidt

2016-11-04 18:35:29 UTC

Permalink

Hi Mirja,

Post by Mirja Kuehlewind (IETF)

Could you document this attack surface in the doc…?

Mhmm, if complete validation is performed (as requested by the
document), the attack surface is not there. As pointed out in 6.2, valid
revocation is a single operation. Otherwise there would be probably a
lengthy discussion on "how can I disturb a peer" (DoS, malicious
requests, ...)

We would rather be robust and skip this set of options ;)

Post by Mirja Kuehlewind (IETF)

Don’t you have to do the check every time you check write access for a shared resource? That can be much more often.

As of Section 6.5, an accessing peer can cache (the list and the
validation). So it can memorize previous checks and doesn't have to do
them over an over again. However, initially and in the case of actual
writing the shared resource, we want a full verification.

Cheers,
Thomas

Post by Mirja Kuehlewind (IETF)

Post by Thomas C. Schmidt

Mirja Kuehlewind (IETF)

2016-11-04 18:47:41 UTC

Permalink

Hi,

that means in the cached case where you know for sure there have been no modification (because it a non-modifiable copy) you don’t do the check, but only if you are basically not sure if there has been a modification in the mean time. That makes more sense. Probably worth stating this more explicitly in the doc.

Mirja

Post by Thomas C. Schmidt
Hi Mirja,

Post by Mirja Kuehlewind (IETF)

Could you document this attack surface in the doc…?

Mhmm, if complete validation is performed (as requested by the document), the attack surface is not there. As pointed out in 6.2, valid revocation is a single operation. Otherwise there would be probably a lengthy discussion on "how can I disturb a peer" (DoS, malicious requests, ...)
We would rather be robust and skip this set of options ;)

Post by Mirja Kuehlewind (IETF)

Don’t you have to do the check every time you check write access for a shared resource? That can be much more often.

As of Section 6.5, an accessing peer can cache (the list and the validation). So it can memorize previous checks and doesn't have to do them over an over again. However, initially and in the case of actual writing the shared resource, we want a full verification.
Cheers,
Thomas

Post by Mirja Kuehlewind (IETF)

Post by Thomas C. Schmidt