Re: [Geopriv] draft-ietf-geopriv-rfc3825bis

From: James M. Polk ^lt;jmpolk@cisco.com>
Date: Tue Aug 17 2010 - 11:52:41 EDT

At 02:46 AM 8/17/2010, Tschofenig, Hannes (NSN - FI/Espoo) wrote:

> > At 08:09 AM 8/10/2010, Tschofenig, Hannes (NSN - FI/Espoo) wrote:
> > >Think about a regular hotel network.
> >
> > many/most wired hotel networks are DSL-like based, so this doesn't
> > apply as directly as it may seem.
>
>Whether the hotel network provider uses DSL or Cable does not matter for
>this purpose.

huh? Cable uses BPI, and DSL systems are individually run. How are
they the same, and why is this an issue?

> >
> > What you are suggesting (that the server MUST NOT hand out location
> > in shared mediums) is requiring a DHCP server to have physical
> > topological awareness before implementing option 123. The same DHCP
> > server can be unaware of ever network topology between it and the
> > client - therefore I believe you are placing an excessive burden or
> > limitation on DHCP as an LCI delivery means.
>
>The GEOPRIV working group is about describing what privacy properties
>are being provided for the solutions we develop. You have always been
>quite keen on documenting everything in detail -- particularly when
>other people proposed something.
>
>Describing what we assume are reasonable operational considerations that
>are applicable is a good thing, I believe.

and yet your text could prevent a valid design choice(s). So how is that good?

>We cannot have all these chats about how important privacy is for us but
>then when it actually has an operational impact then we back-off.
>
>In this specific case, I am not sure I understand what you mean with
>additional requirements. A DHCP server better has an idea about the
>network topology since otherwise how does it hand out the IP addresses
>and other configuration parameters.

for one, RAIO means it doesn't matter what topology is between the
server and the RAIO, but your suggested text doesn't account for that.

another point, in a built wiremap database, as long as the final link
is switched, the entire topology is immaterial - which isn't
accounted for in your text either.

and still another point, if the final hop is within the house, and
the house has a shared topology, does that mean the location provided
by the home gateway (acting as a DHCP server to the home) is an
invalid topology? You text says this home gateway MUST NOT hand out
any location, which is just plain wrong.

> >
> > I'm concerned about the SMB or residential gateway impact of adding
> > this context - which can easily lead to some vendor(s) reading what
> > you are proposing and consider DHCP not appropriate for homes or
> > small businesses inadvertently, where it otherwise would be logical
> > to use. Many gateways of this sort have DHCP as a client to the WAN,
> > and as a server to the LAN.
> >
> > We need to be careful with how we word any changes.
>
>There are two parts here:
>
>1) The first part is to describe what the privacy risks are with the
>technology

the doc already has a warning about DHCP communications not being
secure, and that already passed IESG review in RFC 3825. What changed
in this version that made things so much more insecure to warrant
modifying the text about that warning?

>2) The second part is to make some recommendations for anticipated
>envionments.

Then I suggest you offer text about every type of topology and see if
you can get "strong" consensus, which is required to change the text
of a "bis" document.

>With SMBs and the usage of this document you are touch on item #2. Your
>recommendation would be that the privacy risks are not so dramatic in
>such an environment and I believe it is OK to say that.
>
>You still want to talk about item #1 in the document.

it is already there, as I and others have stated.

James

>Ciao
>Hannes
>
> >
> > James
> >
> >
> > > > -----Original Message-----
> > > > From: ext Marc Linsner [mailto:mlinsner@cisco.com]
> > > > Sent: Tuesday, August 10, 2010 3:59 PM
> > > > To: Tschofenig, Hannes (NSN - FI/Espoo); geopriv@ietf.org
> > > > Subject: Re: [Geopriv] draft-ietf-geopriv-rfc3825bis
> > > >
> > > > Hannes,
> > > >
> > > > What specific network type(s) are you worried about?
> > > >
> > > > -Marc-
> > > >
> > > >
> > > > On 8/10/10 8:25 AM, "Tschofenig, Hannes (NSN - FI/Espoo)"
> > > > <hannes.tschofenig@nsn.com> wrote:
> > > >
> > > > > But the conclusion is missing: if you are on a shared link
> > > > then you must
> > > > > not share location at the level of the individual
> > hosts. I fear that
> > > > > those who implement and deploy would not get the point and would
> > > > > nevertheless reveal information and put the user at risk.
> > > > >
> > > > >> -----Original Message-----
> > > > >> From: ext Marc Linsner [mailto:mlinsner@cisco.com]
> > > > >> Sent: Tuesday, August 10, 2010 3:23 PM
> > > > >> To: Tschofenig, Hannes (NSN - FI/Espoo); geopriv@ietf.org
> > > > >> Subject: Re: [Geopriv] draft-ietf-geopriv-rfc3825bis
> > > > >>
> > > > >> Hannes,
> > > > >>
> > > > >>
> > > > >> On 8/10/10 3:33 AM, "Tschofenig, Hannes (NSN - FI/Espoo)"
> > > > >> <hannes.tschofenig@nsn.com> wrote:
> > > > >>
> > > > >>> Hi all,
> > > > >>>
> > > > >>> during the GEOPRIV meeting I mentioned missing text in
> > > > >>> draft-ietf-geopriv-rfc3825bis regarding security.
> > > > >>>
> > > > >>> DHCP does not provide confidentiality protection as a
> > > > >> built-in feature.
> > > > >>> As Marc mentioned in response to issue#23 (see
> > > > >>> http://trac.tools.ietf.org/wg/geopriv/trac/ticket/23) every
> > > > >> target would
> > > > >>> be given the exact same location information on a
> > shared medium.
> > > > >>>
> > > > >>> Unfortunately, the security consideration section does not
> > > > >> mention this
> > > > >>> aspect with a single word.
> > > > >>
> > > > >> Not true, currently in the security consideration section of
> > > > >> the draft:
> > > > >>
> > > > >> " Since there is no privacy protection for DHCP messages, an
> > > > >> eavesdropper who can monitor the link between the DHCP
> > > > server and
> > > > >> requesting client can discover this LCI."
> > > > >>
> > > > >> I don't believe more text is needed.
> > > > >>
> > > > >> -Marc-
> > > > >>
> > > > >>
> > > > >>
> > > > >>
> > > > >>
> > > > >>
> > > > >> Hence, I suggest to add:
> > > > >>>
> > > > >>> "
> > > > >>> Since there is no confidentiality protection for DHCP
> > > > >> messages, an
> > > > >>> eavesdropper who can monitor the link between the DHCP
> > > > server and
> > > > >>> requesting client can discover this LCI. In cases
> > > > where multiple
> > > > >>> hosts share the same link and can therefore see each
> > > > others DHCP
> > > > >>> messages the DHCP MUST NOT hand out location for
> > > > individual hosts
> > > > >>> but MUST rather provide location of the DHCP relay,
> > > > DHCP server,
> > > > >>> or a similar device instead. This ensures that
> > none of the end
> > > > >>> devices are able to learn exact information of the
> > other hosts
> > > > >>> on the same network.
> > > > >>> "
> > > > >>>
> > > > >>> Ciao
> > > > >>> Hannes
> > > > >>>
> > > > >>> _______________________________________________
> > > > >>> Geopriv mailing list
> > > > >>> Geopriv@ietf.org
> > > > >>> https://www.ietf.org/mailman/listinfo/geopriv
> > > > >>
> > > > >>
> > > > >>
> > > >
> > > >
> > > >
> > >_______________________________________________
> > >Geopriv mailing list
> > >Geopriv@ietf.org
> > >https://www.ietf.org/mailman/listinfo/geopriv
> >
> >

_______________________________________________
Geopriv mailing list
Geopriv@ietf.org
https://www.ietf.org/mailman/listinfo/geopriv
Received on Tue, 17 Aug 2010 10:52:41 -0500

This archive was generated by hypermail 2.1.8 : Tue Aug 17 2010 - 11:53:05 EDT