-
Notifications
You must be signed in to change notification settings - Fork 546
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
Define a new sigstore Bundle type to include additional verification data and materials #2422
Conversation
Signed-off-by: Hector Fernandez <[email protected]>
Signed-off-by: Hector Fernandez <[email protected]>
// RekorPayload holds metadata about recording a Signature's ephemeral key to | ||
// a Rekor transparency log. | ||
// Use Payload instead of TransparencyLogEntry to keep backwards compatibility. | ||
Payload RekorPayload |
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 kept the Payload
field instead of renaming it to TransparencyLogEntry
, as specified in the protobuf spec, to keep backwards compatibility.
@@ -87,7 +87,7 @@ func TestOptions(t *testing.T) { | |||
LayerMediaType: ctypes.SimpleSigningMediaType, | |||
ConfigMediaType: types.OCIConfigJSON, | |||
Annotations: map[string]string{ | |||
BundleAnnotationKey: "{\"SignedEntryTimestamp\":null,\"Payload\":{\"body\":null,\"integratedTime\":0,\"logIndex\":0,\"logID\":\"\"}}", | |||
BundleAnnotationKey: "{\"Payload\":{\"body\":null,\"integratedTime\":0,\"logIndex\":0,\"logID\":\"\"},\"SignedEntryTimestamp\":null,\"EntryTimestampAuthority\":null,\"CertBytes\":null,\"PublicKeyIdentifier\":\"\"}", |
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 the main difference with the new bundle type. Whenever we start using a TSA the bundle will look more like:
"Bundle":{"Payload":{"body":null,"integratedTime":0,"logIndex":0,"logID":""},"SignedEntryTimestamp":null,"EntryTimestampAuthority":"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","CertBytes":null, "PublicKeyIdentifier": ""}
// a Rekor transparency log. | ||
// Use Payload instead of TransparencyLogEntry to keep backwards compatibility. | ||
Payload RekorPayload | ||
|
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 am awaiting feedback before adding other fields such as the messageSignature
included in the spec.
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.
Those fields overlap with LocalSignedPayload
too. The top level fields in that would ideally be removed, or else it would open mismatch issues.
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.
Good point! That could easily happen.
@asraa @priyawadhwa @haydentherapper I'd like to hear your thoughts about this proposal 👀 . I decided to adapt the current implementation to the fields in the protobuf spec. Also, I included a new field EntryTimestampAuthority that will contains the response as received from the TSA server. |
If we are beginning to adopt the Sigstore bundle format, I would prefer we use exactly what has been discussed and approved. Otherwise we now have a third format effectively. If we want to deviate, we should discuss the change in the protobuf specs repo first imo |
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 looks really good! just one question.
) | ||
|
||
type Bundle struct { | ||
VerificationData |
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's the difference between VerificationData
and VerificationMaterial
? I wonder if we could come up with more descriptive names so it's more clear :)
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.
It might be better to port that comment to the upstream bundle, but these are terms being used by those protos.
Data is extra auxiliary data for verification (timestamps, rekor entries), but verification material is the direct material (key hint, certificates)
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 am gonna port these and other questions to the upstream bundle. Thanks.
PublicKeyIdentifier string | ||
} | ||
|
||
// TimestampVerificationData contains various timestamped data following RFC3161. |
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 we discussed in the protobuf-specs repo, SETs are not just timestamp verification data, hence why we went with this nested format - https://github.com/sigstore/protobuf-specs/blob/main/protos/sigstore_bundle.proto#L53-L59
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 created one issue here sigstore/protobuf-specs#13. Hopefully we can get more ppl involved on this discussion.
SignedEntryTimestamp []byte | ||
|
||
// EntryTimestampAuthority contains the recorded entry from timestamp authority server. | ||
EntryTimestampAuthority []byte |
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 would prefer a name that includes RFC3161 rather than Timestamp Authority, as the former makes it clear what the format of the time record is
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 makes sense. I like it.
EntryTimestampAuthority []byte | ||
} | ||
|
||
func EntryToBundle(tLogEntry *models.LogEntryAnon, signedEntryTimestamp, entryTimestampAuthority, certBytes []byte, pubKeyID string) *Bundle { |
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.
Tests?
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, we need tests but I wanted to firstly get some feedback, as I expected things might change.
@haydentherapper this isn't a third format, it's just an addition to the current format. Definitely agree we should also plan for it in the protobuf work, but IIUC that work isn't ready to be used yet (@kommendorkapten please correct me if I'm wrong!). I think it's ok to amend the current format so that we can start testing out this feature & then switch over to the protobufs whenever they're ready. |
@@ -62,7 +62,7 @@ func uploadToTlog(ctx context.Context, sv *sign.SignerVerifier, rekorURL string, | |||
return nil, err | |||
} | |||
fmt.Fprintln(os.Stderr, "tlog entry created with index:", *entry.LogIndex) | |||
return cbundle.EntryToBundle(entry), nil | |||
return cbundle.EntryToBundle(entry, []byte{}, []byte{}, []byte{}, ""), nil |
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.
nit: use nil instead of []byte{}?
) | ||
|
||
type Bundle struct { | ||
VerificationData |
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.
It might be better to port that comment to the upstream bundle, but these are terms being used by those protos.
Data is extra auxiliary data for verification (timestamps, rekor entries), but verification material is the direct material (key hint, certificates)
// a Rekor transparency log. | ||
// Use Payload instead of TransparencyLogEntry to keep backwards compatibility. | ||
Payload RekorPayload | ||
|
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.
Those fields overlap with LocalSignedPayload
too. The top level fields in that would ideally be removed, or else it would open mismatch issues.
I think we can begin to adopt the format, lemme know otherwise @kommendorkapten. I'm worried about introducing an internal If we don't want to adopt the new bundle format yet, I'd prefer we make a struct that makes it clear that it's not another bundle format, but still reuse as much as possible from the bundle format wrt to timestamp fields |
b := &Bundle{} | ||
// If none of the verification data is configured then return nil | ||
if (tLogEntry == nil || tLogEntry.Verification == nil) && len(entryTimestampAuthority) == 0 { | ||
return nil |
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.
why return nil over returning an empty bundle? id be worried about accidental dereferences later
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 agree with you but tLogEntry == nil || tLogEntry.Verification == nil
is reusing the old behavior that also returned a nil bundle.
I broadly agree. That said, I don't expect client implementations to keep the protobuf representations throughout the program; rather, I'd expect them to "parse" the protobuf out into language-idiomatic objects that are a little bit smarter than the generated proto code. This helps us reject invalid representations that are valid in protobuf (e.g., invalid DER). Here's what I'd propose:
|
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 a very-much breaking change to the Golang API. Can we make it official that this API isn't supported by any versioning guarantees?
See also #2365 (reply in thread)
// signatures or any additional timestamped verifications. | ||
type VerificationMaterial struct { | ||
// A chain of X.509 certificates. | ||
CertBytes []byte |
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'd expect this to be a crypto/x509.Certificate
I guess really...is this struct a data format or an object? My preference is always to move parsing as early as possible so we don't have to handle parse errors in the middle of the program logic.
The Sigstore bundle from https://github.com/sigstore/protobuf-specs is in a decent shape I would say. Is it worthy to cut a release in the current stage? Probably no, but I don't think we are miles away. For the npm build provenance we are working on two different implementations of the bundle, generation and verification. One in Go, and one in JavaScript. So we are getting real world testing of the usability now. My feeling is that we will see some small changes in the imminent future, like sigstore/protobuf-specs#10, and cut a release within a few weeks. All the work on the bundle we are doing is expected to be part of sigstore-js and sigstore-go. Beyond the first release, there are a few expected breaking changes, but the horizon for them is a bit unclear, as we are waiting for e.g. Rekor SET V2. |
Thanks everyone for the review and suggestions. I'll keep this PR open so I can take over on these changes whenever there aren't major breaking changes expected from the protobuf spec. In the meantime, I'll try to move on with #2331, and open a different PR with a smaller change that just renames the struct name from |
If it's of any help, we are ready to merge this PR sigstore/protobuf-specs#12 which adds the generated Go code to the repository, so you don't need to rewrite any structs -- you can import them directly from that repository. |
Closing this for now! I'll re-open it when starting this work again. |
Summary
As part of this effort #2331, I'm proposing renaming the original RekorBundle type to a more generic sigstore Bundle type that includes additional data. The new fields added to the sigstore Bundle are inspired from the protobuf specification here: https://github.com/sigstore/protobuf-specs/blob/main/protos/sigstore_bundle.proto.
The next actions include adding support to use a Timestamp Authority server (sigstore/timestamp-authority) to guarantee a signature happened during the cert's validity period.
Release Note
Documentation