Now that we have the production Google Endpoint Encryption Key, we can
update the tests to use the correct GEEK cert chain where applicable.
Test: VtsHalRemotelyProvisionedComponentTargetTest
Test: VtsAidlKeyMintTargetTest
Bug: 191301285
Change-Id: I84b557c6bad34741ffe6671fc941d9e266b73241
Merged-In: I84b557c6bad34741ffe6671fc941d9e266b73241
We will use the 'Attestation IDs State' field in DeviceInfo to
determine whether a device is still provisionable or not. Once a
production device has left the factory, certain attestated device ids
should be fixed, and 'Attestation IDs State' should reflect this
by reporting "locked".
Remove stale, duplicated DeviceInfo description from ProtectedData.aidl
Test: None, just a doc change
Bug: 192017485
Change-Id: I4e0a840a8f415b3b410801805a158c46be30ec6a
Merged-In: I4e0a840a8f415b3b410801805a158c46be30ec6a
Fix the device-unique attestation chain specification: The chain should
have two or three certificates.
In case of two certificates, the device-unique key should be used for
the self-signed root.
In case of three certificates, the device-unique key should be certified
by another key (ideally shared by all StrongBox instances from the same
manufacturer, to ease validation).
Adjust the device-unique attestation tests to accept two or three
certificates in the chain.
Additionally, the current StrongBox KeyMint implementation can not yet
generate fully-valid chains (with matching subjects and issuers), so
relax that check.
Bug: 191361618
Test: m VtsAidlKeyMintTargetTest
Change-Id: I6e6bca33ebb4af67cac8e41a39e9c305d0f1345f
There are two tags that cannot be currently removed but should be
removed in KeyMint V2. Mark them as deprecated and point to the bug
for deletion.
Bug: 183737811
Test: That it compiles.
Change-Id: I80ccaedeb777fdb249a8cb021db6628da32d6029
Get two test BCCs, then ensure that no repeated keys are found.
Bug: 192687735
Test: VtsHalRemotelyProvisionedComponentTargetTest
Change-Id: I48f86e7dfa9ab4bc6303a8d1b64ac7ca6ac76bbf
Merged-In: I48f86e7dfa9ab4bc6303a8d1b64ac7ca6ac76bbf
We need both the build fingerprint as well as the CSR when uploading
data to the APFE provisioning server. Add a utility function to format
the output as a JSON blob so that it may be easily collected in the
factory in a serialized data format, then later uploaded.
Test: libkeymint_remote_prov_support_test
Test: VtsAidlKeyMintTargetTest
Test: VtsHalRemotelyProvisionedComponentTargetTest
Bug: 191301285
Change-Id: I751c5461876d83251869539f1a395ba13cb5cf84
- Make clear that CERTIFICATE_NOT_{BEFORE,AFTER} must be specified for
generating/importing asymmetric keys.
- Fix enforcement level of Tag::UNLOCKED_DEVICE_REQUIRED.
- Fix reference to exportKey() for Tag::STORAGE_KEY to mention
convertStorageKeyToEphemeral instead.
- Mark Tag::CONFIRMATION_TOKEN as deprecated.
Test: none, comment change
Bug: 188672564
Change-Id: I68727b024f6b6743403941763aefca64e3eb091a
Include a unit test to verify the GEEK cert chain is valid.
Test: libkeymint_remote_prov_support_test
Bug: 191301285
Change-Id: Icf9cfa165fbccb24b36b03ff3ce729a7e9c44cfd
Merged-In: Icf9cfa165fbccb24b36b03ff3ce729a7e9c44cfd
This functionality will be used for the factory tooling, so we should
test it. Additionally, some new functionality will soon be added, and
it also needs to be tested.
Test: libkeymint_remote_prov_support_test
Bug: 191301285
Change-Id: I6a8798fc4b09fff1e829185a4b9e471921e5d2a9
Merged-In: I6a8798fc4b09fff1e829185a4b9e471921e5d2a9
It's possible that corrupted ciphertext decrypts just fine. e.g. the
output ends with "0x01".
However, the chances of this happening are relatively low
(roughly 1/256). Corrupt the ciphertext up to 8 times, ensuring that
the likelihood of multiple successful decryptions is so miniscule that
it's effectively impossible.
Test: Ran *PaddingCorrupted tests 50000 times
Change-Id: If40ecd7817819921c020ea9b86ada18c4c77ea55
This flag is never used anywhere, so just remove it. When used, it would
bypass signature checks. This is something we generally don't want to
do, even in testing. So remove the flag so there's no temptation to use
it.
Bug: 190942528
Test: VtsHalRemotelyProvisionedComponentTargetTest
Change-Id: I0433c1eedc08e9a5a5ad71347154867dba61689e
Merged-In: I0433c1eedc08e9a5a5ad71347154867dba61689e
The KeyMint AIDL spec requires that "Tag::EC_CURVE must be provided to
generate an ECDSA key". Move the VTS tests to always create ECDSA keys
by curve not key size.
Bug: 188672564
Test: VtsAidlKeyMintTargetTest
Change-Id: I33036387c243b21ab0ecd49221b7e7757598913e
Remove TODOs from the KeyMint specification that were not concrete
enough or did not have enough context to act upon.
Bug: 183737811
Test: That it compiles.
Change-Id: I01899be5e65e9943053aa796a2ab23f1a783a1aa
Makes the AIDL (and its dependencies) available to allow client code
to build against it.
Fixes: 190995136
Test: Client code (in progress) builds.
Change-Id: I06e7486463bca93ed25377c0dca30484a6bbf656
Try all tags in attestion extension one by one
Test: VtsAidlKeyMintTargetTest on CF
Bug: 186735514
Change-Id: I63ca8d298d2d16f707f2437ab48aaa69c1d7563d
Check that the various ATTESTATION_ID_* tags are included if they
have the correct value, and that keygen fails if they have an invalid
value.
Also update attestation tags to include vendor/boot patchlevel if
they're available. (They always should be, but fixing that is a
separate task.)
Bug: 190757200
Test: VtsAidlKeyMintTargetTest
Merged-In: Ibaed7364c6d08c0982e2a9fb6cb864ae42cf39fe
Change-Id: Ibaed7364c6d08c0982e2a9fb6cb864ae42cf39fe
Now that the aidl compiler supports it, use constants from TagType to
indicate the type of each tag, rather than duplicating the values of
the constants.
Test: atest VtsAidlKeyMintTargetTest
Bug: 183737811
Change-Id: Ie8af1f00d04fa05c59cfc72692caecbcf2fae483
The vendor patchlevel is YYYYMMDD not YYYYMM
Bug: 188672564
Bug: 186735514
Test: VtsAidlKeyMintTargetTest
Change-Id: Ia641f8eef84a85aec8f2a0551c192b6874301126
Get description of ASN.1 schema in HAL and the keymint support library
in sync with each other. Change code to always list tags in the same
order (by numeric tag).
Bug: 188672564
Bug: 186735514
Test: VtsAidlKeyMintTargetTest
Change-Id: I620f54ba4a265ea69d174f6f44765a8508bfe803
Add a check that the TAG_EARLY_BOOT_ONLY is included in the returned key
characteristics.
Bug: 188672564
Test: VtsAidlKeyMintTargetTest
Change-Id: I200c61f34888c720c47f6289d79cd21d78436b58
The description should note that keystore-enforced tags are not to be
returned. This is done so that the keymint implementation doesn't have
to bother keeping track of tags it's not repsonsible for dealing with.
Bug: 186685601
Test: none (it's just a comment change)
Change-Id: I3ff94201c262a5071d271b150dbbf21888d678aa
Improve the documentation and tests related to device-unique
attestation on StrongBox KeyMint devices:
* Test that the chain produced is exactly of length 2.
* Document how the chain needs to be structured.
* Explain the trust properties of the key used for the
self-signed root.
Test: atest VtsAidlKeyMintTargetTest
Bug: 187803288
Change-Id: I09bb16d6938b567c114485d2df00bde9d3e1ccf9
This primarily updates CDDL to allow for OEMs who wish to use P256
instead of Ed25519 to do so. One structural change of note that affects
all implementors is that SignedMacAad now includes the tag from the
COSE_Mac0 of MacedKeysToSign to prevent a potential vulnerability that
would exist if an attacker compromised the server's EEK private key.
Bug: 189018262
Test: Purely a comment change
Change-Id: I043a19c6aba0f771315d45c04ab5263b610b5de8
Merged-In: I043a19c6aba0f771315d45c04ab5263b610b5de8
This fixes up the tests to go along with the change to the signature
of the MAC key. Primarily, this adds the MAC tag from the MACing
operation over the public key set to be signed into the AAD of the
signature of said MAC key.
Bug: 189018262
Test: atest VtsHalRemotelyProvisionedComponentTargetTest
Change-Id: Ibdcf242e0ae73dee1a08fe98d939130055e4492e
Merged-In: Ibdcf242e0ae73dee1a08fe98d939130055e4492e
