Re: [Rats] Composite Evidence

[Dave Thaler <[e-mail]>]

Agree with Ned that we should keep composite devices separate from time series snapshots of the same component in terms of concepts.

From: Smith, Ned <[e-mail]>
Sent: Wednesday, January 22, 2020 5:54 PM
To: Eric Voit (evoit) <[e-mail]>; Birkholz, Henk <[e-mail]>; Michael Richardson <[e-mail]>; Dave Thaler <[e-mail]>
Cc: [e-mail]
Subject: Re: Composite Evidence

Eric,
I think there are two types of “composite” things that could be considered (i) decompositions of hardware and software such as a motherboard and the sub-components that could be plugged into it and (ii) time series snapshots of the same component.

You seem to be asserting (ii) where a time series snapshot of the same thing could be considered “composite evidence”.

I think both cases are valid, but maybe it makes sense to use different terms for each as they seem to have distinct properties. (e.g. evidence having multiple entries of a component with the same name  under (i) implies multiple instances of the same type of device – such as two NIC cards. Whereas under (ii) multiple entries implies there is one instance of the NIC sampled over a time interval.

-Ned

From: "Eric Voit (evoit)" <[e-mail]<mailto:[e-mail]>>
Date: Wednesday, January 22, 2020 at 4:04 PM
To: Henk Berkholz <[e-mail]<mailto:[e-mail]>>, Michael Richardson <[e-mail]<mailto:[e-mail]>>, Dave Thaler <[e-mail]<mailto:[e-mail]>>, "Smith, Ned" <[e-mail]<mailto:[e-mail]>>
Cc: "[e-mail]<mailto:[e-mail]>" <[e-mail]<mailto:[e-mail]>>
Subject: Composite Evidence

Henk,         Dave,
Michael,    Ned,

I promised you a definition for Composite Evidence.  You can see my proposed definition directly the text below, but I am not willing yet to place it in a Pull Request. I thought an email thread might be helpful first.

Anyway my strawman definition for Composite Evidence is:  Evidence which includes multiple sub-elements of evidence, more than one of which can be computationally verified to have been generated by a specific Attester Subcomponent or Verifier.

I built this definition considering the passport model, which looks like it will often needs to use composite evidence.  As an example of why I believe this, see the use case below.

    .--------------.
    |  Verifier A  |
    '--------------'
        ^     [2]
        |     Verifier A signed Attestation Results @time(x) (
    evidence(  |  determination, hash(TpmQuote@time(x)))
    TpmQuote   |
    @time(x))  |
       [1]     V
     .-------------.                           .---------------.
     |  Attester   |<------nonce @time(y)---[3]|  Verifier B   |
     |    .-----.  |                           |       /       |
     |    | Tpm |  |[4]-composite evidence ( ->| Relying Party |
     |    '-----'  |      TpmQuote@time(y),    '---------------'
     '-------------'      TpmQuote@time(x),
                          Verifier A signed Attestation Results @time(x) )


In the example above, evidence at time x is generated and signed within a TPM.  This would *not* be composite evidence.   This evidence would be evaluated by Verifier A, signed, and returned as Attestation Results to the Attester.   A subsequent request from a Relying Party at time y could pull three independently signed elements of evidence from the Attester.  These three would comprise the composite evidence which when taken together would allow Verifier B / Relying Party to evaluate the current trustworthiness of the Attester.

Does this definition meet your needs?

Thanks,
Eric