-
Notifications
You must be signed in to change notification settings - Fork 13
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
WIP: Added vts and provisioning plugins for AWS Nitro enclave attestation … #58
base: main
Are you sure you want to change the base?
Conversation
…documents Signed-off-by: Derek D. Miller <[email protected]>
Signed-off-by: Derek D. Miller <[email protected]>
51eba16
to
e61cd60
Compare
Signed-off-by: Derek D. Miller <[email protected]>
Signed-off-by: Derek D. Miller <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
hi @dreemkiller, I'm leaving an initial set of comments about the provisioning path. If you want we can have a chat to go through the high level design aspects.
## Trust Anchor | ||
|
||
```json | ||
{ | ||
"scheme": "AWS_NITRO", | ||
"type": "VERIFICATION_KEY", | ||
"attributes": { | ||
"psa.hw-model": "RoadRunner", | ||
"psa.hw-vendor": "ACME", | ||
"psa.iak-pub": "MFkwEwYHKoZIzj0CAQYIKoZIzj0DAQcDQgAE6Vwqe7hy3O8Ypa+BUETLUjBNU3rEXVUyt9XHR7HJWLG7XTKQd9i1kVRXeBPDLFnfYru1/euxRnJM7H9UoFDLdA==", | ||
"psa.impl-id": "IllXTnRaUzFwYlhCc1pXMWxiblJoZEdsdmJpMXBaQzB3TURBd01EQXdNREU9Ig==", | ||
"psa.inst-id": "AUyj5PUL8kjDl4cCDWj/0FyIdndRvyZFypI/V6mL7NKW" | ||
} | ||
} | ||
``` |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
there is no need to for an explicit trust-anchor endorsement, I think. The CN=aws.nitro-enclaves, C=US, O=Amazon, OU=AWS
cert can be embedded in the package that provides the attestation document verification.
{ | ||
"scheme": "AWS_NITRO", | ||
"type": "REFERENCE_VALUE", | ||
"attributes": { | ||
"psa.hw-model": "RoadRunner", | ||
"psa.hw-vendor": "ACME", | ||
"psa.impl-id": "IllXTnRaUzFwYlhCc1pXMWxiblJoZEdsdmJpMXBaQzB3TURBd01EQXdNREU9Ig==", | ||
"psa.measurement-desc": 1, | ||
"psa.measurement-type": "BL", | ||
"psa.measurement-value": "h0KPxSKAPTEGXnvOPPA/5HUJZjHl4Hu9eg/eYMTPJcc=", | ||
"psa.signer-id": "rLsRx+TaIXIFUjzkzhokWuGiOa48a/2eeHH35di66Gs=", | ||
"psa.version": "2.1.0" | ||
} | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this isn't what you want. (looks like a leftover?)
what you really want is to define a way to use CoMID to transport the expected PCR values and some kind of enclave identifier.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Definitely a leftover.
I am currently not transporting any expected PCR values. That's not how Veracruz attestation works (they are embedded in the generated certificate and the client authenticates them). I've thought a bit how it might be done for other systems, but yes, we need to figure out how to do that for more general use-cases.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we need to figure out how to do that for more general use-cases.
👍 definitely an important item for discussion.
To support both use cases we need to create a non-ambiguous signal that the verifier plugin can use to distinguish between "don't bother checking PCRs for this workload" and "no matching PCRs for this workload"
Vendor string | ||
Model string |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
is this information available in the attestation document?
here you need something that can uniquely correlate the incoming evidence (i.e., the attestation document) with the expected reference values. would module_id
work for this purpose?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Having thought a bit more about precise enclave identification, it seems like PCR0 would do the job?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
if so, we'd need a new instance-id type in comid since we don't currently have a place for a sha-384 string.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
PCRs 0-2 are a function of the input image (not the platform). They certainly could be included as part of a COSWID, but I'm not sure of the use-case.
Putting PCR8 in a COSWID would effectively tie the attestation to a specific AWS Customer, which may be valuable.
PCR3 ties you to a specific IAM role. Perhaps interesting.
PCR4 ties you to a specific AWS EC2 instance, which I have no idea why you would want to do this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
PCRs 0-2 are a function of the input image (not the platform).
sure, but what we need is a unique identifier for the workload itself. The assurance that it's run on genuine Nitro is derived from the cryptographic verification, isn't it?
type Extractor struct{} | ||
|
||
func (o Extractor) SwCompExtractor(rv comid.ReferenceValue) ([]*proto.Endorsement, error) { | ||
return nil, fmt.Errorf("Not implemented, not needed?") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this is where you'd read the expected PCRs from the CoMID, so it looks like it's needed
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think needed for some use-cases, but not all (see my above comments about Veracruz's needs).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
👍 this needs designing
// return structpb.NewStruct(swAttrs) | ||
} | ||
|
||
func (o Extractor) TaExtractor(avk comid.AttestVerifKey) (*proto.Endorsement, error) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this instead should not be needed: the trust anchor is fixed and can be embedded in the package that provides the verification logics
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The trust anchor is a long-lived CA Certificate. The way it currently works is that certificate is passed into the AuthenticateDocument
function call in go-nitro-enclave-attestation-document
: https://github.com/veracruz-project/go-nitro-enclave-attestation-document/blob/0893a6c144113b9a29f3111101f64b4d6cf1c999/lib.go#L37
I did not want to embed a certificate, no matter how long-lived, in the logic of the module. Thus, it is being provided as part of the provisioning process.
} | ||
|
||
func (o *NitroSwCompAttributes) FromMeasurement(m comid.Measurement) error { | ||
return fmt.Errorf("Not implemented, not needed?") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ditto, this is needed, but first you need to define your CoMID profile to carry the PCRs
"type": "uuid", | ||
"value": "51d1a77a-9908-49e0-92df-e9c872a9b0bd" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
what does this uuid represent?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just randomly generated when I was getting started. I do not use it for anything.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ok. the environment identifier is used (indirectly) as a lookup key for the endorsements. so it should be somewhat bound to the incoming evidence.
}, | ||
"verification-keys": [ | ||
{ | ||
"key": "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" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
as I said above, this does not need to be an external endorsement: it's effectively an immutable root that is common to any possible nitro-based enclave, as such it belongs to the verification package
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I disagree. While it's long-lived, it's possible that AWS may update it or even revoke it. That sort of thing does not belong in the package.
Additionally, considering your suggestion that we use an alternate CA cert for testing (and generate "fake" AWS Nitro attestation documents for this testing), embedding it in the package seems like it would complicate things.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I disagree. While it's long-lived, it's possible that AWS may update it or even revoke it. That sort of thing does not belong in the package.
there is no way to revoke a root CA cert (it's the classic chicken & egg). You can only remove it from the "root store". and one can quite naturally consider the Go package the "root store" for the nitro attestation PKI.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Additionally, considering your suggestion that we use an alternate CA cert for testing (and generate "fake" AWS Nitro attestation documents for this testing), embedding it in the package seems like it would complicate things.
that sounds like an implementation detail, e.g. one could do:
// verify.go
func verifyWithRoot(evidence []bytes, rootCA x509.Certificate) error {
// ...
}
func Verify(evidence []bytes) error {
return verifyWithRoot(evidence, AWSRootCA)
}
and
// verify_test.go
func TestVerify_withAltRoot_fail(t *testing.T) {
// ...
expectedErr := `cannot anchor path onto a trusted root`
assert.EqualError(t, verifyWithRoot(testEvidence, altRootCA), expectedErr)
}
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, it's not revokable, but that doesn't mean it will not be revoked. It would not be too difficult for them to send a notice to all AWS users that have used NItro enclaves. Certainly messy, but possible.
Additionally, AWS may in the future want to use different root CA certs for different situations (consider their US government cloud, which I imagine already does, if it supports Nitro at all).
I still feel that the CA Cert does not belong in the authentication library, it should be provisioned into the Veraison infra.
If the CA cert was in the authentication library, Veracruz has zero use for Veraison in the nitro use-case, because we are basically only relying on Veraison to handle provisioning and maintenance of that CA Cert.
"key": "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" | ||
}, | ||
{ | ||
"key": "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" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
curiosity: what is this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The first one is the CA key. The 2nd one is garbage to enable a "Dual Key" test. I don't think there's an actual use-case here.
Signed-off-by: Derek D. Miller <[email protected]>
Signed-off-by: Derek D. Miller <[email protected]>
Signed-off-by: Derek D. Miller <[email protected]>
Can we declare this OBE and close it? |
…documents