[saag] Re: [nasr] Re: Re: NASR BOF Follow-Up

On Thu, Apr 3, 2025 at 6:09 AM Richard Barnes <[email protected]> wrote:

>
> On Q18, I think Meiling's summary accurately captures what was said at the
> microphone, but I think that response is deeply misleading.  This solution
> will result in detailed network state information going outside of operator
> networks, or it cannot work.  Let me try to state it clearly:
>
> - The basic premise here is that the Network wants to attest to a Relying
> Party some properties about the overall state of the network
> - In order to evaluate whether the properties are true, a Verifier has to
> to actually look at the detailed network state
> - The Verifier has to be either the Relying Party or someone the Relying
> Party trusts
> - The Verifier **cannot** be the Network, because then the Network is just
> self-asserting and you don't need any technology at all
>
> Yes, the technical implementation might involve hashes.  But
> fundamentally, the information has to leave the Network.  So another thing
> the proponents need to demonstrate is that there's any plausible scenario
> where a network operator is going to be willing to expose this information.
>

I think the key point here is that attestation just acts to
demonstrate that a given endpoint has a given configuration, but what
relying parties are actually interested in is whether the endpoint
conforms to a specific policy.

In some cases, there is a relatively straight line between
configuration and policy. For instance, in the case of DRM, the
content provider wants to know that the endpoint won't exfiltrate the
raw content, and it can assure itself of this by determining that the
endpoint running one of a small set of media players/CDMs, etc. that
it knows prevent -- or at least resist -- exfiltration. This is
also consistent with just knowing a hash of the code because there
are only a small number of valid configurations.

However, it's not clear to me that that's true in this case, because
unlike media players, network devices are highly configurable and a
large number of the configuration directives might impact the relevant
security claims. Thus, determining whether an element is policy
conformant is a matter of knowing not just what code it is running
but the state of every relevant configuration directive. One could
imagine this working at least three ways:

- Having the element attest to the state of the code and the
  relevant configuration directives.

- Having the element attest to the code and then somehow assert
  that the state of the configuration directives guarantees
  some set of properties (e.g., "all data of class X follows
  path Y").

In the former case, it is quite likely that there will be a large
number of valid states, because each directive may have multiple
acceptable values, and so you end up with combinatoric explosion
issues if you just have a list of hashes [0]. In the latter case we
need some language for talking about the policy claims; I suppose
that's possible, but it seems like it could get complicated quickly,
and I don't see anything like that in these drafts, though perhaps
I'm missing it.

Either approach requires studying the impact of every existing
configuration directive for each device type to know what the
impact will be on the relevant policy claims. This seems challenging
at best.

-Ekr


[0] Of course you could hash the individual configuration values,
but this is much the same as publishing the configuration.



> It was not raised in the BoF, but I would like to point out a logical
> inconsistency at the core of the proposition here: On the one hand, the
> proponents assert that end-to-end cryptographic protections are not
> sufficient to keep endpoints safe.  On the other hand, the premise of the
> proposed solution is ... more cryptography.  Attestations, MACSEC, etc.  So
> the argument here cannot rest on a hand-wavy "cryptography isn't enough";
> you need to articulate what specific risks you're addressing.
>
> Honestly, when I read the charter for this work, I did not initially have
> as negative a reaction.  I initially presumed it was like verified SFC, and
> that you would do things like making a firewall rule a little more
> permissive if the flow had been verified upstream.  It's not implausible
> that there are some use cases of that shape.  But instead, the
> presentations in the BoF framed an impossible problem and an unworkable
> solution, which clearly I could not support.
>
> Hope that helps,
> --Richard
>
>
> On Thu, Apr 3, 2025 at 12:05 AM Meiling Chen <[email protected]>
> wrote:
>
>> Hi all,
>>
>> We will continue to track and actively address the issues raised on NASR
>> BoF, for convenience, I have brought the Q&A record to the email list.
>> Welcome to continue discussing with those who are concerned about these
>> issues.
>>
>> Q1: Why is not end-to-end protection sufficient and why should we care
>> about where the traffic goes through?
>> Q2:  Why this is important to assure at the network layer?
>> Q3: List actual concrete examples of existing regulations.
>> Q4: List the technical requirements in detail.
>> Q5: NASR being a possible way to address pervasive traffic monitoring
>> threats?
>>
>> Q6: Proof of transit does not imply that the packet didn’t exit the path.
>> ---->PoNT is out of scope.
>>
>> Q7: There is work about proof of transit that has been done in SFC, what
>> is needed to highlight what is not there (in routing, RATS, SFC PoT) which
>> NASR will provide.-->
>> NASR is the joint use of attestation techncology and proof of transit.
>>
>> Q8: What about lawful interceptions?---->
>> Trustworthiness as defined in RATS can attest that a router does what is
>> supposed to do, including lawful interception if it is what it is supposed
>> to do.
>>
>> Q9: There is work from RATS, like RFC 9684, that can be used in NASR. Why
>> an independent effort? Why not putting this work in an existing WG?---->
>> RATS is the building block to create path attestation, whose usage is
>> then verified with proof of transit. These two things together look out of
>> the scope of RATS. NASR does not want to replace RFC 9684, but rather
>> leverage it.
>>
>> Q10: What is exactly the definition of a path in NASR? BGP session?
>> Tunnel? physical links?---->
>> A document in PANRG has a good definition of what a path is.
>>
>> Q11: Trustworthiness and traffic engineering are already addressed in
>> different places, so what is new in NASR? ---->
>> The possibility for the customers to verify that the traffic has been
>> engineered in the way they asked. How to do it is the gap NASR wants to
>> fill.
>>
>> Q12: Candidate solutions from problems exist in other venues (Routing
>> security / SFC)---->
>> True, but they do not address the combination (forwarding).
>>
>> Q13: Proposed work may result in fragmentation of the Internet, including
>> and excluding poeople?---->
>> No. That is the wrong terminology to use. NASR is about path complaint to
>> a set of claims. Is a specific service in a limited domain not a general
>> service for the Internet.
>>
>> Q14: Clarification of requirements whether the requirement is to detect
>> nodes along the path that do not support NASR?---->
>> No. Only have proof that traffic went through a set of nodes that support
>> NASR.
>>
>> Q15: Bringing the work to RATS would not work. RATS is already bloated
>> and bringing all this work there is not possible.
>>
>> Q16: PoT has cryptographic cost.--->
>> Implementation on SRv6 shows that the cost is very limited.
>>
>> Q17: Remark about the scope of implementation: Internet or limited
>> domain?--->
>> limited domain implemented at the operator level.
>>
>> Q18: Need to verify every configuration (binary / configuration files /
>> routers) in order to determine the properties of the path?--->This is
>> not how remote attestation works. There is no configuration shared, the
>> configuration is attested through trusted third party. You share hash
>> values that attest the the property.
>>
>> Q19: Why not doing this in the SFC WG?---->Piece can be done in other
>> WG, but from a global perspective wold be better to have a WG.
>>
>> Q20: For reliability, what about having multiple paths? Can it be a
>> scalability issue? ----Yes, the attribute of multiple path is included.
>>
>> Best,
>> Meiling
>>
>>
>> *From:* Liuchunchi\(Peter\) <[email protected]>
>> *Date:* 2025-03-29 19:19
>> *To:* [email protected]; IETF SAAG <[email protected]>
>> *CC:* [email protected]
>> *Subject:* [saag] NASR BOF Follow-Up
>>
>> Hi NASR, SAAG,
>>
>>
>>
>> Thanks everyone for participating and contributing valuable feedbacks to
>> the NASR BOF last week! Thank you Nancy and Luigi for mediating a
>> productive discussion, and thanks Deb for supervising the process.
>>
>>
>>
>> During the process, we had extensive discussions that showed both major
>> interests in the problem and differing understanding of the objective. For
>> example, NASR represents new use case requirements that aim to extend
>> Remote Attestation benefits to the network scale, providing more
>> comprehensive security visibility and behavioral assurance across network
>> deployments to the customers. But NASR is not intended, in any capacity, to
>> serve as a substitute for end-to-end encryption technology, etc.
>>
>>
>>
>> The poll result shows a clear consensus that this is an IETF problem, but
>> a rougher (at best) or no consensus that the problem is well understood.
>>
>>
>>
>> To help on that, next steps we will work on a dedicated document that
>> collects all valuable feedbacks from the BOF and resolve them one-by-one.
>> Again, we sincerely thank everyone who contributed valuable feedbacks and
>> cares about bringing more security and transparency to networks!
>>
>>
>>
>> Best regards,
>>
>> Peter (Chunchi) Liu
>>
>>
>>
>> _______________________________________________
>> saag mailing list -- [email protected]
>> To unsubscribe send an email to [email protected]
>>
> --
> nasr mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
>

_______________________________________________
saag mailing list -- [email protected]
To unsubscribe send an email to [email protected]