Re: [EAT] [EXTERNAL] Re: [Rats] Attestation BoF charter updates?

Henk Birkholz <> Wed, 17 October 2018 12:20 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 29371130DD3; Wed, 17 Oct 2018 05:20:11 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -6.9
X-Spam-Status: No, score=-6.9 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, RCVD_IN_DNSWL_HI=-5, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id obx4oi1okH3H; Wed, 17 Oct 2018 05:20:05 -0700 (PDT)
Received: from ( []) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by (Postfix) with ESMTPS id 955DC1277C8; Wed, 17 Oct 2018 05:20:04 -0700 (PDT)
Received: from ( []) by (8.14.4/8.14.4/Debian-4.1ubuntu1) with ESMTP id w9HCJupQ012389 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Wed, 17 Oct 2018 14:19:57 +0200
Received: from [] ( by ( with Microsoft SMTP Server (TLS) id 14.3.408.0; Wed, 17 Oct 2018 14:19:50 +0200
To: "" <>, Laurence Lundblade <>, Guy Fedorkow <>
CC: Hannes Tschofenig <>, "Smith, Ned" <>, "Jeremy O'Donoghue" <>, "" <>, Denis <>
References: <> <> <> <> <> <> <> <> <> <> <> <> <>
From: Henk Birkholz <>
Message-ID: <>
Date: Wed, 17 Oct 2018 14:19:44 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <>
Content-Type: text/plain; charset="utf-8"; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
X-Originating-IP: []
Archived-At: <>
Subject: Re: [EAT] [EXTERNAL] Re: [Rats] Attestation BoF charter updates?
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: EAT - Entity Attestation Token <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Wed, 17 Oct 2018 12:20:11 -0000

Hi Laurence,

maybe I understand now where the general confusion is originating from.

I dug through some parts of the EAT draft again and could it be that 
UIED are identifier and not an identities? They literally are claims 
(aka assertions) that can be signed as content of a CWT as illustrated in


1.4.2. talks about identity documents by mentioning "X.509 certificate 
chains or something similar but different". But as you highlighted 
anything with regard to that is explicitly intended _not_ to go into the 
EAT claim set as content, but is stored on the thing (or maybe on a 
thing that signs "on behalf of" the thing the EAT is about?) and 
consequently is only used to potentially sign the EAT claim set (via 
COSE). Is that correct?

As a result, an EAT would never be able to act as an identity document 
itself, but the key material that comes along with an x.509 certificate 
(as one example of many highlighted) could be used to sign the EAT.

RATS are able to use the "X.509 certificate chains or something similar 
but different" and the corresponding key material comes along with a 
certificate (as one example amongst others) to establish Attestation 

Viele Grüße,


On 10/17/2018 01:19 PM, Laurence Lundblade wrote:
> Wasn’t saying there was ambiguity in the TCG case. Not being a TCGer, 
> was just was trying to understand what the TCG definition is.
> Your mentioning 802.1AR DevID helps a lot as I have some knowledge of it.
> I see now that my speculation about SW / config version was wrong. 
> That’s fine. I only wrote it to try to get to clarity.
> Thanks for help clarifying.
> In EAT, we defined a UEID for what TCGer’s mean by “identity":
>     UEID's identify individual manufactured entities / devices such as a
>     mobile phone, a water meter, a Bluetooth speaker or a networked
>     security camera.  It may identify the entire device or a submodule or
>     subsystem.  It does not identify types, models or classes of devices.
>     It is akin to a serial number, though it does not have to be
>     sequential.
> See here <> 
> for more on UEID.
> In EAT there is a strong desire to separate the signing key material 
> from UEID to have flexibility in the signing scheme. For example FIDO 
> and Android both have privacy-preserving schemes were a signing key is 
> shared by 100,000 devices. (Some use cases need privacy, some don’t, 
> perhaps because of legal agreements with customers, and correlation 
> between use cases and privacy schemes can’t be counted on for 
> high-volume devices like mobile phones).
> I don’t know what the right answer(s) is(are) in the design yet...
> LL
>> On Oct 17, 2018, at 4:26 PM, Guy Fedorkow < 
>> <>> wrote:
>> Hi Laurence,
>>  I don't think there's ambiguity in the TCG case...  current work in 
>> attestation focuses on using the 802.1AR DevID spec, which corresponds 
>> exactly to one of your cases below -- it's a vendor, model and serial 
>> number.
>>  It's true that TCG technology can be used to validate software 
>> integrity, but that's hardly the same as identity.
>>  Considerations of user privacy commonly accepted in TCG would 
>> preclude the inclusion of software state as part of identity, 
>> practicalities aside.  In general, privacy considerations lead TCG to 
>> carefully avoid unauthorized disclosure of either software or hardware 
>> based assertions.
>>  Let me know if I'm missing the point...
>> Thanks
>> /guy fedorkow, juniper networks
>> -----Original Message-----
>> From: Laurence Lundblade < 
>> <>>
>> Sent: Wednesday, October 17, 2018 12:31 AM
>> To: Henk Birkholz < 
>> <>>
>> Cc: Smith, Ned < <>>; 
>> Hannes Tschofenig < 
>> <>>; 
>> <>; Jeremy O'Donoghue < 
>> <>>; Denis < 
>> <>>; <>
>> Subject: Re: [EAT] [EXTERNAL] Re: [Rats] Attestation BoF charter updates?
>> I think the TCG has a certain developed notion of identity that is 
>> different from others.  A way to get sharper about what a particular 
>> identity means is to describe the space and how the identifier selects 
>> in the space.  Here’s a few examples:
>> - The space is physical human beings with fingers. A fingerprint 
>> scanner select one of these humans. This is typical biometric auth.
>> - The space is humans that know a password. A password authenticator 
>> selects the humans that know the password. Typical authentication.
>> - The space is humans as legal entities / citizens with criminal 
>> records, passports and such. A national ID selects.
>> - The space is physical manufactured devices from one maker.  A serial 
>> number narrows it down to one physical device.
>> - The space is physical devices from one maker. A model number or 
>> version, narrows down to a subset of these devices.
>> My understanding of the TCG world is that identities goes beyond any 
>> of the above because they include things like the version of the SW 
>> and configuration of the device. The identity is largely a hash of an 
>> identifier of the physical device, various SW components and various 
>> configuration states.  The point being that a change in any of these 
>> is a change in the security characteristics of the device and if it 
>> was trusted with a particular key or particular data in the past, that 
>> trust should be revoked until re established.
>> - The space is physical devices from any maker
>>  + Measurement of the SW (the SW versions)
>>     + Measurements of some configuration
>> Changing any of these change the identity of the device.
>> I know I hardly need to ask, but TCG folks, did I describe this correctly?
>> So that said, many, perhaps most, use cases for EAT do not want this 
>> kind of device identity. It is not a prerequisite for trusting  claims 
>> coming from the device. In some cases this level of identity seems 
>> extremely disruptive as small SW upgrades could result in re 
>> enrollment of biometrics. We already are annoyed when we have to 
>> answer security questions when start using a service from a new device.
>> I think this WG needs to address both this very tight TCG-style device 
>> identity and the more heuristic EAT-style claims-for-risk-engines. I 
>> could see multi-decade evolution where risk engine use cases want more 
>> and more of the tight TCG-style device identity as criminals become 
>> more adept at forging EAT-style claims. (Today risk engines work off 
>> of unsecured browser parameters).
>> The Introduction in the charter seems entirely focused on the 
>> TCG-based identity use case. I think text should be added for 
>> EAT-style attestation. I will write some after some comments on this 
>> message.
>> LL
>>> On Oct 16, 2018, at 8:24 PM, Henk Birkholz 
>>> < 
>>> <>> wrote:
>>> To quote Carsten, there is "the usual terminology problem about 
>>> identity".
>>> Identifying vs. Identity brings back memories about an extensive 
>>> discussion in SACM that in the end resulted in using Profiles...
>>> Profiles are the "things" that you can aggregate identity claims in 
>>> (in SACM these claims are "Target Endpoint Characteristics", which 
>>> are pretty much the same thing as Device Characteristics) - and you 
>>> use the Profiles to (re-)identify target endpoints.
>>> In the scope of remote attestation procedures, I think that "just not 
>>> calling" claim sets that are composed (amongst others) of identifying 
>>> claims "identity documents" might just end up in making up another 
>>> name for something that is effectively composing identity documents 
>>> anyways. If calling the same thing something else helps in some way, 
>>> I am all for it. But I am really not sure.
>>> That said, Ned's elaboration on the relationship of identity and 
>>> attestation claims and corresponding claim semantics / trust 
>>> semantics is spot on, I think.
>>> Viele Grüße,
>>> Henk
>>> On 10/16/2018 03:51 PM, Smith, Ned wrote:
>>>> Agree with Hannes that identity discussion can lead to confusion as 
>>>> identity is often highly dependent on some sort of context (domain 
>>>> of interpretation).
>>>> In order to have productive conversation about attestation and 
>>>> identity, it may be important to understand that attestation 
>>>> semantics sets the context for identity.
>>>> * Attestation attributes have trust semantics (and trust semantics 
>>>> are interpreted through the eyes of the verifier). Identity 
>>>> attributes may also have trust semantics (in addition to identity 
>>>> semantics) hence can be regarded as playing a role in both identity 
>>>> and attestation. Since RATS should be interested primarily in trust 
>>>> semantics, it makes sense to understand, discuss and consider 
>>>> identity in terms of its trust semantics. Hence identity attributes 
>>>> are only interesting to RATS if they are useful for evaluating trust 
>>>> semantics.
>>>> * In a closed ecosystem, the trust characteristics and properties 
>>>> may be implied because there often exists out-of-band processes that 
>>>> ensure the characteristics are applied (for example, if an IT 
>>>> organization sets a policy that all asymmetric keys are to be 
>>>> hardened using a smartcard, they can apply a deployment process 
>>>> where the IT professional verifies the smartcard is used during 
>>>> enrollment.) But in a non-closed ecosystem, there isn't an 
>>>> assumption that out-of-band processes exist. Attestation attempts to 
>>>> make explicit what otherwise is implicit and therefore subject to 
>>>> misinterpretation if the verifier, relying party and attester are 
>>>> not part of the same closed ecosystem. Implicit attestation 
>>>> therefore can lead to semantic confusion if we're not careful. 
>>>> Therefore, the rigor of treating implicit attestation claims 
>>>> (assertions assumed to exist but not explicitly named) should 
>>>> nevertheless be defined explicitly. Then, if there is a need for 
>>>> implicit attestation, the specifications that describe them can 
>>>> describe them in terms of an explicit definition.
>>>> * In the context of asserting trust characteristics, the verifier 
>>>> needs to know that a set of characteristics applies to the instance 
>>>> of the endpoint being evaluated (aka the attester). (Otherwise a 
>>>> MITM could supply characteristics that may be valid, but don’t apply 
>>>> to the specific instance). Consequently, it makes sense to consider 
>>>> that *attestation* keys are needed to authenticate the endpoint 
>>>> being evaluated. This doesn't imply an identity key couldn't double 
>>>> as an attestation key. But if it does, we care about its relevance 
>>>> to the attester endpoint.
>>>> -Ned
>>>> On 10/15/18, 6:49 PM, "Hannes Tschofenig" 
>>>> < <>> wrote:
>>>>    Here are my 5 cents on the term 'identity'.
>>>>         I have been working on identity management for many years 
>>>> (partially because of my OAuth co-chair role) and I have noticed 
>>>> that the use of the term "identity" tends to confuse more than it 
>>>> helps (even more so when the term is applied to devices and software 
>>>> rather than humans).
>>>>         I believe we could get away with describing what we want 
>>>> without ever using the term.
>>>>         In this context I would argue that we are about identifying 
>>>> different components, such as hardware and software.
>>>>         Ciao
>>>>    Hannes
>>>>         -----Original Message-----
>>>>    From: EAT < <>> On 
>>>> Behalf Of Smith, Ned
>>>>    Sent: Monday, October 15, 2018 9:00 AM
>>>>    To: Henk Birkholz < 
>>>> <>>; 
>>>> <>; Jeremy O'Donoghue < 
>>>> <>>; Denis < 
>>>> <>>
>>>>    Cc: <>
>>>>    Subject: Re: [EAT] [EXTERNAL] Re: [Rats] Attestation BoF charter 
>>>> updates?
>>>>         I agree with Henk's observation that there is opportunity to 
>>>> find common ground on identity as identity can be a collection of 
>>>> attributes, that if signed, becomes a set of verifiable claims 
>>>> (about identity attributes). Attestation has a similar structure in 
>>>> that it consists of a set of attributes, that if signed becomes a 
>>>> set of verifiable claims. The main difference between identity and 
>>>> attestation attributes is whether the attribute semantically relate 
>>>> to trust (vs. identity). For example, claims of security 
>>>> certification compliance resonates as attestation.
>>>>         It may be possible to combine a set of attestation 
>>>> attributes in such a way that the combination uniquely identifies an 
>>>> entity. But that doesn't make it an identity system per se and 
>>>> shouldn't cause a conflict in charter goals with respect to other 
>>>> identity focused efforts in IETF.
>>>>         Also, inclusion of attributes that don't have semantics that 
>>>> reasonably relate to trust should not be of interest to a working 
>>>> group focused on attestation.
>>>>         -Ned
>>>>         On 10/15/18, 3:53 PM, "EAT on behalf of Henk Birkholz" 
>>>> < <> on behalf of 
>>>> <>> wrote:
>>>>             Hi Jeremy & Denis,
>>>>             all these points of view have merit, I think.
>>>>             Please have a look at these quotes:
>>>>> knowing the provenance of the device
>>>>> having the entity creator/manufacturer provide a set of claims 
>>>>> which can be verified
>>>>> to know the characteristics of a device by issuing to the device a 
>>>>> public key certificate which will only be issued to the device if 
>>>>> it fulfills an expected set of functionalities.
>>>>             To me, these descriptions seem to have different 
>>>> intents, scope or
>>>>        audience and require most certainly different work-flows, but
>>>>        semantically and in consequence representation-wise they also 
>>>> seem quite
>>>>        related.
>>>>             If you think of an identity document as a set of claims 
>>>> (potentially
>>>>        including a public key), that is signed and that can match 
>>>> one or more
>>>>        things, maybe all three types highlighted above are actually 
>>>> covered by
>>>>        the same identity document concept?
>>>>             With respect to identity documents, I think it is safe 
>>>> so say that we
>>>>        can find common ground here rather easily. That said, the way 
>>>> how these
>>>>        identity documents come into existence, when and how they are 
>>>> deployed
>>>>        on the thing or associated with multiple things is another 
>>>> question. We
>>>>        most certainly do not want to end up with a "signing fool" 
>>>> that does not
>>>>        add value to the level of confidence (I am quoting Ned Smith 
>>>> here).
>>>>             This has to be fleshed out explicitly for each scenario 
>>>> and the parts
>>>>        that are reused in every scenario should get special 
>>>> attention in the
>>>>        upcoming RATS architecture draft.
>>>>             Identifying claims aggregated in an identity document 
>>>> most certainly can
>>>>        include device characteristics, public key(s), or claims 
>>>> about the
>>>>        intended functions and capabilities of the thing. Claims can even
>>>>        originate from Claimants that are not the thing itself, but 
>>>> external
>>>>        entities providing an assertion (e.g. these things are 
>>>> CC-certified).
>>>>             All different composites of identifying claims have a 
>>>> specific shared
>>>>        intend in this context though, I think: expressing Attestation
>>>>        Provenance (again - one for one or more things of the same 
>>>> kind). And
>>>>        all of them originate from Claimants and therefore use a root 
>>>> of trust
>>>>        of measurement, I think.
>>>>             That said, establishing trustworthy knowledge about 
>>>> attestation
>>>>        provenance is only the first step and provides the foundation 
>>>> for more
>>>>        complex attestation procedures. In any case, this seem to be 
>>>> required by
>>>>        every scenario that was highlighted so far and therefore will 
>>>> most
>>>>        certainly be covered in a deliverable.
>>>>             IHTH,
>>>>             Henk
>>>>                  On 10/15/2018 12:22 PM, Jeremy O'Donoghue wrote:
>>>>> On 12 Oct 2018, at 14:33, Denis < 
>>>>> <>
>>>>> <>> wrote:
>>>>>> The "Problem Statement" continues as follows:
>>>>>> (2) Today, there is no common, standard way for Relying Parties to
>>>>>> know the provenance and characteristics of a device (e.g., an end-user
>>>>>> device,
>>>>>> platform or endpoint, servers, IoT devices, device subsystems and
>>>>>> sub-modules) that may be requesting services.
>>>>>> Knowing the provenance of the device is not always necessary. What is
>>>>>> important is to know that a specific set of functions is (securely)
>>>>>> supported by the device.
>>>>>> There already exists a common way to know the characteristics of a
>>>>>> device by issuing to the device a public key certificate which will
>>>>>> only be issued to the device
>>>>>> if it fulfills an expected set of functionalities. In such a case a
>>>>>> syntax for a trusted claim sets**is not needed.
>>>>> This is indeed a possible way to address the problem, but suffers from
>>>>> challenges which come down to:
>>>>> - Who issues such certificates. An option is that there is some form of
>>>>> security certification behind this issuance otherwise it is of very
>>>>> limited value.
>>>>> - In order to capture the wide variety of different devices (I prefer
>>>>> "entity" here as it is more general than "device"), many different 
>>>>> forms
>>>>> of such certificates would be required. This ends up being much the 
>>>>> same
>>>>> thing as a set of claims.
>>>>> Given the multiplicity of entities for which attestation might be
>>>>> useful, it seems to me that having the entity creator/manufacturer
>>>>> provide a set of claims which can be verified is more flexible than
>>>>> requiring a certificate. It seems likely to me that where a formal
>>>>> security certification is needed there will be a certificate issued by
>>>>> the Certifying Body as one of the claims about an entity, but I can
>>>>> think of many cases (supply chain management, to take just one example)
>>>>> where provenance is by far the most useful information.
>>>>> I'm not in favour of changes to the charter in this direction.
>>>>>> The simplest cases should also be part of the framework.
>>>>>> Note: I personally appreciate that privacy is an aspect that
>>>>>> will/should be taken into consideration.
>>>>>> Denis
>>>>>>> As with the earlier draft, there are many references to verify and
>>>>>>> verifier in the lead up to the deliverable but no deliverables 
>>>>>>> describing
>>>>>>> verification rules. I'd like to see rules in one of these 
>>>>>>> documents and
>>>>>>> think it worth citing in the deliverables. Are bindings to 
>>>>>>> certificate
>>>>>>> management protocols out of scope?
>>>>>>> On 10/11/18, 9:49 AM, "RATS on behalf of Henk Birkholz"
>>>>>>> < <> on behalf 
>>>>>>> of 
>>>>>>> <>>  wrote:
>>>>>>>> Hi all,
>>>>>>>> sorry for the delay. Please find an updated charter proposal here:
>>>>>>>> The BoF proponents tried to merge all comments and proposals on
>>>>>>>> keystores, existing solutions, provenance & device 
>>>>>>>> characteristics, etc
>>>>>>>> that were raised on the list - and align them in homogeneous 
>>>>>>>> fashion.
>>>>>>>> This draft is intended to focus the discussion and improve the 
>>>>>>>> wording.
>>>>>>>> Viele Grüße,
>>>>>>>> Henk
>>>>>>>> On 10/06/2018 08:23 AM, Laurence Lundblade wrote:
>>>>>>>>> Hi Henk,
>>>>>>>>> Bangkok IETF is getting close, will you be able to update the 
>>>>>>>>> charter
>>>>>>>>> for the attestation BoF to properly include EAT?  I sent you 
>>>>>>>>> some text
>>>>>>>>> that just needs to be pasted along with some comments. It 
>>>>>>>>> doesn’t look
>>>>>>>>> like there’s been any updates.
>>>>>>>>> LL
>>>>>>>> _______________________________________________
>>>>>>>> RATS mailing list
>>>>>>>> <>
>>>>>>> _______________________________________________
>>>>>>> EAT mailing list
>>>>>>> <>
>>>>>> _______________________________________________
>>>>>> EAT mailing list
>>>>>> <> <>
>>>>> _______________________________________________
>>>>> EAT mailing list
>>>>> <>
>>>>             _______________________________________________
>>>>        EAT mailing list
>>>> <>
>>>>              _______________________________________________
>>>>    EAT mailing list
>>>> <>
>>>>    IMPORTANT NOTICE: The contents of this email and any attachments 
>>>> are confidential and may also be privileged. If you are not the 
>>>> intended recipient, please notify the sender immediately and do not 
>>>> disclose the contents to any other person, use it for any purpose, 
>>>> or store or copy the information in any medium. Thank you.
>>> _______________________________________________
>>> EAT mailing list
>>> <>
>> _______________________________________________
>> EAT mailing list
>> <>