On 09/10/2012 09:30 PM, Paul Lambert wrote:
> 
>> -----Original Message-----
>> From: Rosen, Brian [mailto:[email protected]]
>> Sent: Monday, September 10, 2012 1:14 PM
> ...
>> It would be easy to discover the certs of the DBAs, because there are
>> few of them, and controlled by regulators.  Discovery of the server can
>> include discovery of the certs.  The other way is harder for
>> manufacturers because of the issues you cite, but overall, it's much
>> easier than shared secrets.
> So, we may be in agreement :-)
> 
> The DBs can be discovered, but they need some common trust point
> in the devices per regulatory authority.
> 
>>
>> It is a requirement to authenticate both sides, so we need solutions
>> for both.
> Looking at the security stuff documented by this group, we have
> as a threat:
>     3.2.1.  Impersonation of a master device

Where's that? I only see a "MAY" in [1] for client auth
(which I think is correct).

S.

[1]
http://tools.ietf.org/html/draft-ietf-paws-problem-stmt-usecases-rqmts-08#section-8

> A little late for me to argue this requirement, so yes we
> have created a requirement for master device authentication.
> I don't see why a Rogue device even cares about the DB.  The
> devices are "certified" as compliant to the regulatory rules.
> Extra cryptography will not prevent Rogue devices from simply
> transmitting in whatever channel they want.  
> 
> Given this client authentication requirement - I totally agree
> that client public keys are a little more work, but always 
> better than distributing secret keys.
> 
> Paul
> 
>>
>> Brian
>>
>>
>>
>>  -----Original Message-----
>> From:        Paul Lambert [mailto:[email protected]]
>> Sent:        Monday, September 10, 2012 04:04 PM Eastern Standard Time
>> To:  Rosen, Brian; 'Peter McCann'; 'Peter Stanforth'; 'Stephen
>> Farrell'; '[email protected]'
>> Cc:  '[email protected]'
>> Subject:     RE: [paws] PAWS security
>>
>>> -----Original Message-----
>>> From: Rosen, Brian [mailto:[email protected]]
>> ...
>>> The easiest way to do this is to have the manufacturer put a cert in
>>> the device that it signs.  Then the only thing the database needs is
>>> the public key of the manufacturer,
>> You cannot prevent spoofing of a DB without having a root of some sort
>> that "certifies" real db providers.  This is a small number of
>> certificates installed in the servers.
>>
>> Having a per-device cert signed by a manufacturer and installed in the
>> factory is harder.  It's more certificates and complicates the
>> manufacturing process.  The authentication to a manufacturer is not
>> necessarily useful. The authentication of the manufacturers FCC ID and
>> device Id might be of interest, but the regulations do not mandate
>> (AFAIK) this complexity of assignment of a few bits of per device
>> regulatory information.
>>
>>> Why do we want to make this harder?
>> There are two requirements:
>>  - authentication of multiple DB providers within a regulatory domain
>>  - "secure" transmission of Regulatory ID and Device identification
>>     to DB
>> Proposal below was primarily focused on the first requirement above.
>> Your proposal for device certs is for the second.
>> Both might be required ... device certs have other options also than
>> manufacturer oriented certs and is a longer discussion.
>>
>> Paul
>>
>>> Because some manufacturers think its too much work to do that?
>>>
>>> Brian
>>>
>>>
>>>
>>>  -----Original Message-----
>>> From:       Paul Lambert [mailto:[email protected]]
>>> Sent:       Monday, September 10, 2012 03:15 PM Eastern Standard Time
>>> To: Peter McCann; Peter Stanforth; Stephen Farrell;
>>> [email protected]
>>> Cc: [email protected]
>>> Subject:    Re: [paws] PAWS security
>>>
>>>
>>>> This restriction may hold today for the FCC, but I have a hard time
>>>> believing that a pre-existing manufacturer-database relationship
>>>> will be mandatory in every regulatory domain for the forseeable
>>>> future.  I think we should design to minimize the friction in
>>>> changing database providers.
>>>
>>> The least friction would be to hardwire devices with a PK "root" for
>>> each regulatory domain that the device supports.
>>>
>>> Valid DBs would have credentials signed by the key holder of roots
>> for
>>> each regulatory domain
>>>
>>> Paul
>>>
>>>
>>>> -Pete
>>>>
>>>> Peter Stanforth wrote:
>>>>> I completely agree with your statement below. If we are good any
>>>> radio
>>>>> will have the "Technical" capability to work with any database.
>>>>> BUT, and this is a big BUT, the regulation may not allow it - The
>>>>> FCC does not today. And as a DBA I only want to work with devices
>>>>> where I have a predefined relationship. Not least of which,
>>>>> because as I am authorized by a Regulator I want to make sure I
>> am
>>>>> doing everything I can to avoid their wrath.
>>>>> This has come up before. PAWS is defining an API/Protocol, not
>> the
>>>>> regulation nor the business cases at this time. So I still say
>>>>> that
>>>> we
>>>>> can comply with your statement without considering a "User".
>>>>>
>>>>> On MonSep/10/12 Mon Sep 10, 2:18 PM, "Peter McCann"
>>>>> <[email protected]> wrote:
>>>>>
>>>>>> This is a key question we need to answer.
>>>>>>
>>>>>> I thought we were defining a protocol that would allow a radio
>>> from
>>>>>> any manufacturer to interoperate with a database of any database
>>>> vendor.
>>>>>>
>>>>>> -Pete
>>>>>>
>>>>>> Peter Stanforth wrote:
>>>>>>> Pete, I am not sure I believe in a use case where "users" move
>>>>>>> from one DB to another. The relationship is between the radio
>>>>>>> vendor and the DBA. The radio vendor may have multiple
>>>>>>> relationships and allow an end user  to choose one, but it will
>>> be
>>>>>>> from the predefined
>>>> list.
>>>>>>> As a DBA I  don't think I would want, and in the USA we are not
>>>>>>> permitted, to establish a relationship with a user without a
>>>>>>> preexisting relationship with the radio vendor. So security
>>> should
>>>>>>> be primarily a radio vendor-DBA issue. Wearing my DBA hat I
>>>>>>> don't think I would want to service a radio that I do not have
>> a
>>>>>>> preexisting relationship with it's manufacturer.  I am not sure
>>>> that
>>>>>>> user relationships are even within the scope of what PAWS
>> should
>>>>>>> be
>>>> defining. Peter S.
>>>>>>>
>>>>>>> On MonSep/10/12 Mon Sep 10, 1:58 PM, "Peter McCann"
>>>>>>> <[email protected]> wrote:
>>>>>>>
>>>>>>>> Personally, I think that manually provisioned symmetric keys
>>> will
>>>>>>>> severely limit the ability for users to flexibly move from one
>>>>>>>> database provider to another.  It will require some secure
>>>>>>>> out-of-band channel on which to transmit the secret key from
>>>>>>>> one party to the other, and a mechanism for the user to push
>>>>>>>> the
>>>> secret
>>>>>>>> key down into the master device.
>>>>>>>>
>>>>>>>> Provisioning of certificates does not require the secret
>>> channel,
>>>>>>>> only one that is integrity protected.  I think this is much
>>>>>>>> easier to achieve in practice.
>>>>>>>>
>>>>>>>> -Pete
>>>>>>>>
>>>>>>>> Stephen Farrell wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 09/07/2012 10:01 PM, [email protected] wrote:> At the
>>>>>>>>> Vancouver F2F we had extensive discussions on the security
>>> model
>>>>>>>>> for PAWS. Draft-das proposes to use shared secrets pre-
>>>> provisioned
>>>>>>>>> in the master devices for authentication, while draft-lei
>>>> proposes
>>>>>>>>> to mandate client certificates into master devices.
>>>>>>>>>>
>>>>>>>>>> There seemed to be an understanding that the credential
>> types
>>>>>>>>>> in
>>>>>>> use
>>>>>>>>> for authentication are a matter of a business model chosen by
>>>>>>>>> the provider which deploys white space devices, rather than a
>>>> protocol
>>>>>>>>> decision.
>>>>>>>>>>
>>>>>>>>>> Brian mentioned that the iesg may not allow a document to be
>>>>>>>>> published which specifies how shared secrets are used, but
>>>>>>>>> does not have a provisioning mechanism for the shared secrets
>>> defined.
>>>>>>>>> In my opinion, a mechanism for distributing shared secrets
>>>>>>>>> does not necessarily have to be defined, as a shared secret
>>>>>>>>> can be established using current practices to set up shared
>>>>>>>>> secrets with financial institutions (using the browser).
>>>>>>>>>> Thus, my suggestion would be to describe in the document how
>>>>>>>>>> a shared
>>>>>>>>> secret and how a client certificate is used to authenticate a
>>>>>>>>> master device. Then, if we run into issues with iesg, in
>> worst
>>>>>>>>> case we could remove the shared secret part and keep the
>> other
>>>> one.
>>>>>>>>>>
>>>>>>>>>> I'd like to get additional views on this topic.
>>>>>>>>>
>>>>>>>>> I can help a little here (depending on your definition of
>>>>>>>>> help:-)
>>>>>>>>>
>>>>>>>>> BCP 107 [1] is the one to look at when considering how to
>>>> approach
>>>>>>>>> automated key management requirements. That has an IMO very
>>> good
>>>>>>>>> abstract:
>>>>>>>>>
>>>>>>>>>    The question often arises of whether a given security
>> system
>>>>>>>>>    requires some form of automated key management, or whether
>>>> manual
>>>>>>>>>    keying is sufficient.  This memo provides guidelines for
>>>> making
>>>>>>>>>    such decisions. When symmetric cryptographic mechanisms
>> are
>>>> used
>>>>>>>>>    in a protocol, the presumption is that automated key
>>>> management
>>>>>>>>>    is generally but not always needed.  If manual keying is
>>>>>>>>> proposed,
>>>>> the
>>>>>>>>>    burden of proving that automated key management is not
>>>> required
>>>>>>>>>    falls to the
>>>>>>> proposer.
>>>>>>>>> For those that might be less familiar with IETF processes,
>> the
>>>>>>>>> fact that that's a BCP means that its entirely valid for an
>> AD
>>>>>>>>> to strenuously push back against a WG that wants to ignore
>>>>>>>>> what it
>>>> says.
>>>>>>>>>
>>>>>>>>> Cheers,
>>>>>>>>> S.
>>>>>>>>>
>>>>>>>>> PS: I've not read the drafts, and couldn't make the session
>> in
>>>>>>>>> Vancouver, so this might be entirely off topic, but in the
>>>>>>>>> case
>>>> of
>>>>>>>>> PAWS I'd also argue that the regulators' apparent desire to
>>>>>>>>> force devices to authenticate is not necessarily the right
>> one
>>>>>>>>> for the IETF to force upon all users of the PAWS protocol. So
>>>>>>>>> there could for example be a mode of operation where the DB
>>>>>>>>> signs stuff but doesn't care who's asking and that'd be a far
>>>>>>>>> far easier scenario in terms of automated key management.
>>>>>>>>>
>>>>>>>>> [1] http://tools.ietf.org/html/bcp107
>>>>>>>>> _______________________________________________ paws mailing
>>>>>>>>> list [email protected] https://www.ietf.org/mailman/listinfo/paws
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> paws mailing list
>>>>>>>> [email protected]
>>>>>>>> https://www.ietf.org/mailman/listinfo/paws
>>>>>>
>>>>>>
>>>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> paws mailing list
>>>> [email protected]
>>>> https://www.ietf.org/mailman/listinfo/paws
>>> _______________________________________________
>>> paws mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/paws
_______________________________________________
paws mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/paws

Reply via email to