| Internet-Draft | REIM | October 2020 |
| Birkholz, et al. | Expires 26 April 2021 | [Page] |
This document describes interaction models for remote attestation procedures (RATS). Three conveying mechanisms - Challenge/Response, Uni-Directional, and Streaming Remote Attestation - are illustrated and defined. Analogously, a general overview about the information elements typically used by corresponding conveyance protocols are highlighted. Privacy preserving conveyance of Evidence via Direct Anonymous Attestation is elaborated on in the context of the Attester, Endorser, and Verifier role.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 26 April 2021.¶
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶
Remote ATtestation procedureS (RATS, [I-D.ietf-rats-architecture]) are workflows composed of roles and interactions, in which Verifiers create Attestation Results about the trustworthiness of an Attester's system component characteristics. The Verifier's assessment in the form of Attestation Results is created based on Attestation Policies and Evidence - trustable and tamper-evident Claims Sets about an Attester's system component characteristics - generated by an Attester. The roles Attester and Verifier, as well as the Conceptual Messages Evidence and Attestation Results are concepts defined by the RATS Architecture [I-D.ietf-rats-architecture]. This document defines interaction models that can be used in specific RATS-related solution documents. The primary focus of this document is the conveyance of attestation Evidence. The reference models defined can also be applied to the conveyance of other Conceptual Messages in RATS. Specific goals of this document are to:¶
In summary, this document enables the specification and design of trustworthy and privacy preserving conveyance methods for attestation Evidence from an Attester to a Verifier. While the conveyance of other Conceptual Messages is out-of-scope the methods described can also be applied to the conveyance of, for example, Endorsements or Attestation Results.¶
This document uses the following set of terms, roles, and concepts as defined in [I-D.ietf-rats-architecture]:¶
Attester, Verifier, Relying Party, Conceptual Message, Evidence, Endorsement, Attestation Result, Appraisal Policy, Attesting Environment, Target Environment¶
A PKIX Certificate is an X.509v3 format certificate as specified by [RFC5280].¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The term "Remote Attestation" is a common expression and often associated or connoted with certain properties. The term "Remote" in this context does not necessarily refer to a remote entity in the scope of network topologies or the Internet. It rather refers to a decoupled system or entities that exchange the payload of the Conceptual Message type called Evidence [I-D.ietf-rats-architecture]. This conveyance can also be "Local", if the Verifier role is part of the same entity as the Attester role, e.g., separate system components of the same Composite Device (a single RATS entity). Examples of these types of co-located environments include: a Trusted Execution Environment (TEE), Baseboard Management Controllers (BMCs), as well as other physical or logical protected/isolated/shielded Computing Environments (e.g. embedded Secure Elements (eSE) or Trusted Platform Modules (TPM)). Readers of this document should be familiar with the concept of Layered Attestation as described in Section 4.3 Two Types of Environments of an Attester in [I-D.ietf-rats-architecture] and the definition of Attestation as described in [I-D.ietf-rats-tpm-based-network-device-attest].¶
This document focuses on generic interaction models between Attesters and Verifiers in order to convey Evidence. Complementary procedures, functions, or services that are required for a complete semantic binding of the concepts defined in [I-D.ietf-rats-architecture] are out-of-scope of this document. Examples include: identity establishment, key distribution and enrollment, time synchronization, as well as certificate revocation.¶
Furthermore, any processes and duties that go beyond carrying out remote attestation procedures are out-of-scope. For instance, using the results of a remote attestation that are created by the Verifier, e.g., how to triggering remediation actions or recovery processes, as well as such remediation actions and recovery processes themselves, are also out-of-scope.¶
The interaction models illustrated in this document are intended to provide a stable basis and reference for other solutions documents inside or outside the IETF. Solution documents of any kind can reference the interaction models in order to avoid text clones and to avoid the danger of subtle discrepancies. Analogously, deviations from the generic model descriptions in this document can be illustrated in solutions documents to highlight distinct contributions.¶
DAA [DAA] is a signature scheme used in RATS that allows preservation of the privacy of users that are associated with an Attester (e.g. its owner). Essentially, DAA can be seen as a group signature scheme with the feature that given a DAA signature no-one can find out who the signer is, i.e., the anonymity is not revocable. To be able to sign anonymously an Attester has to obtain a credential from a DAA Issuer. The DAA Issuer uses a private/public key pair to generate a credential for an Attester and makes the public key (in the form of a public key certificate) available to the verifier in order to enable them to validate the DAA signature obtained as part of the Evidence.¶
In order to support these DAA signatures, the DAA Issuer MUST associate a single key pair with each group of Attesters and use the same key pair when creating the credentials for all of the Attesters in this group. The DAA Issuer's public key certificate used by a group of Attesters replaces the individual Attester Identity documents during authenticity validation as a part of the appraisal of Evidence conducted by a Verifier. This is in contrast to intuition that there has to be a unique Attester Identity per device.¶
This document extends the duties of the Endorser role as defined by the RATS architecture with respect to the provision of these Attester Identity documents to Attesters. The existing duties of the Endorser role and the duties of a DAA Issuer are quite similar as illustrated in the following subsections.¶
Via its Attesting Environments, an Attester only generates Evidence about its Target Environments. After being appraised to be trustworthy, a Target Environment may become a new Attesting Environment in charge of generating Evidence for further Target Environments. [I-D.ietf-rats-architecture] explains this as Layered Attestation. Layered Attestation has to start with an initial Attesting Environment. In essence, there cannot be turtles all the way down [turtles]. At this rock bottom of Layered Attestation, the Attesting Environments are always called Roots of Trust (RoT). An Attester cannot generate Evidence about its own RoTs by design. As a consequence, a Verifier requires trustable statements about this subset of Attesting Environments from a different source than the Attester itself. The corresponding trustable statements are called Endorsements and originate from external, trustable entities that take on the role of an Endorser (e.g., supply chain entities).¶
In order to enable the use of DAA, an Endorser role takes on the duties of a DAA Issuer in addition to its already defined duties. DAA Issuers offer zero-knowledge proofs based on public key certificates used for a group of Attesters [DAA]. Effectively, these certificates share the semantics of Endorsements, with the following exceptions:¶
The zero-knowledge proofs required cannot be created by an Attester alone - like the Endorsements of RoTs - and have to be created by a trustable third entity - like an Endorser. Due to that semantic overlap, the Endorser role is augmented via the definition of DAA duties as defined below. This augmentation enables the Endorser to convey trustable third party statements both to Verifier roles and Attester roles.¶
In order to ensure an appropriate conveyance of Evidence via interaction models in general, the following set of prerequisites MUST be in place to support the implementation of interaction models:¶
This section defines the information elements that are vital to all kinds interaction models. Varying from solution to solution, generic information elements can be either included in the scope of protocol messages (instantiating Conceptual Messages) or can be included in additional protocol parameters or payload. Ultimately, the following information elements are required by any kind of scalable remote attestation procedure using one or more of the interaction models provided.¶
The following subsections introduce and illustrate the interaction models:¶
Each section starts with a sequence diagram illustrating the interactions between Attester and Verifier. While the interaction models presented focus on the conveyance of Evidence, the intention of this document is in support of future work that applies the presented models to the conveyance of other Conceptual Messages, namely Attestation Results, Endorsements, Reference Values, or Appraisal Policies.¶
All interaction models have a strong focus on the use of a handle to incorporate a type of proof of freshness. The ways these handles are processed is the most prominent difference between the three interaction models.¶
.----------. .----------.
| Attester | | Verifier |
'----------' '----------'
| |
| |
valueGeneration(targetEnvironment) |
| => claims |
| |
| <------requestEvidence(handle, authSecIDs, claimSelection) |
| |
claimsCollection(claimSelection) |
| => collectedClaims |
| |
evidenceGeneration(handle, authSecIDs, collectedClaims) |
| => evidence |
| |
| returnEvidence-------------------------------------------> |
| returnEventLog-------------------------------------------> |
| |
| evidenceAppraisal(evidence, eventLog, refClaims)
| attestationResult <= |
| |
¶
This Challenge/Response Remote Attestation procedure is initiated by the Verifier, by sending a remote attestation request to the Attester. A request includes a Handle, a list of Authentication Secret IDs, and a Claim Selection.¶
In the Challenge/Response model, the handle is composed of qualifying data in the form of a cryptographically strongly randomly generated, and therefore unpredictable, nonce. The Verifier-generated nonce is intended to guarantee Evidence freshness.¶
The list of Authentication Secret IDs selects the attestation keys with which the Attester is requested to sign the Attestation Evidence. Each selected key is uniquely associated with an Attesting Environment of the Attester. As a result, a single Authentication Secret ID identifies a single Attesting Environment.¶
Analogously, a particular set of Evidence originating from a particular Attesting Environments in a composite device can be requested via multiple Authentication Secret IDs. Methods to acquire Authentication Secret IDs or mappings between Attesting Environments to Authentication Secret IDs are out-of-scope of this document.¶
The Claim Selection narrows down the set of Claims collected and used to create Evidence to those that the Verifier requires. If the Claim Selection is omitted, then by default all Claims that are known and available on the Attester MUST be used to create corresponding Evidence. For example when performing a boot integrity evaluation, a Verifier may only be requesting a particular subset of claims about the Attester, such as Evidence about BIOS and firmware the Attester booted up, and not include information about all currently running software.¶
While it is crucial that Claims, the Handle, as well as the Attester Identity information MUST be cryptographically bound to the signature of Evidence, they may be presented in an encrypted form.¶
Cryptographic blinding MAY be used at this point. For further reference see section Section 11.¶
As soon as the Verifier receives signed Evidence, it validates the signature, the Attester Identity, as well as the Nonce, and appraises the Claims. Appraisal procedures are application-specific and can be conducted via comparison of the Claims with corresponding Reference Claims, such as Reference Integrity Measurements. The final output of the Verifier are Attestation Results. Attestation Results constitute new Claims Sets about an Attester's properties and characteristics that enables Relying Parties, for example, to assess an Attester's trustworthiness.¶
.----------. .----------.
| Attester | | Verifier |
'----------' '----------'
| |
valueGeneration(targetEnvironment) |
| => claims |
| |
| .--------------------. |
| <----------handle | | |
| | Handle Distributor | |
| | | handle----------> |
| '--------------------' |
| |
evidenceGeneration(handle, authSecIDs, collectedClaims) |
| => evidence |
| |
| pushEventLog---------------------------------------------> |
| pushEvidence---------------------------------------------> |
| |
| appraiseEvidence(evidence, eventLog, refClaims)
| attestationResult <= |
~ ~
| |
valueGeneration(targetEnvironment) |
| => claimsDelta |
| |
evidenceGeneration(handle, authSecIDs, collectedClaims) |
| => evidence |
| |
| pushEventLogDelta----------------------------------------> |
| pushEvidence---------------------------------------------> |
| |
| appraiseEvidence(evidence, eventLogDelta, refClaims)
| attestationResult <= |
| |
¶
Uni-Directional Remote Attestation procedures can be initiated both by the Attester and by the Verifier. Initiation by the Attester can result in unsolicited pushes of Evidence to the Verifier. Initiation by the Verifier always results in solicited pushes to the Verifier. The Uni-Directional model uses the same information elements as the Challenge/Response model. In the sequence diagram above, the Attester initiates the conveyance of Evidence (comparable with a RESTful POST operation or the emission of a beacon). While a request of evidence from the Verifier would result in a sequence diagram more similar to the Challenge/Response model (comparable with a RESTful GET operation), the specific manner how handles are created and used always remains as the distinguishing quality of this model. In the Uni-Directional model, handles are composed of trustable signed timestamps as shown in [I-D.birkholz-rats-tuda], potentially including other qualifying data. The handles are created by an external 3rd entity - the Handle Distributor - that includes a trustworthy source of time and takes on the role of a Time Stamping Authority (TSA, as initially defined in [RFC3161]). Timstamps created from local clocks (absolute clocks using a global timescale, as well as relative clocks, such as tick-counters) of Attesters and Verifiers MUST be cryptographically bound to fresh Handles received from the Handle Distributor. This binding provides a proof of synchronization that MUST be included in every evidence created. Correspondingly, evidence created for conveyance via this model provides a proof that it was fresh at a certain point in time. Effectively, this allows for series of evidence to be pushed to multiple Verifiers, simultaniously. Methods to detect excessive time drift that would mandate a fresh Handle to be received by the Handle Distributor, as well as timing of handle distribution are out-of-scope of this document.¶
.----------. .----------.
| Attester | | Verifier |
'----------' '----------'
| |
valueGeneration(targetEnvironment) |
| => claims |
| |
| <----subscribeEvidence(handle, authSecIDs, claimSelection) |
| subscriptionResult --------------------------------------> |
| |
evidenceGeneration(handle, authSecIDs, collectedClaims) |
| => evidence |
| |
| pushEventLog---------------------------------------------> |
| pushEvidence---------------------------------------------> |
| |
| evidenceAppraisal(evidence, eventLog, refClaims)
| attestationResult <= |
~ ~
| |
valueGeneration(targetEnvironment) |
| => claimsDelta |
| |
evidenceGeneration(handle, authSecIDs, collectedClaims) |
| => evidence |
| |
| pushEventLogDelta----------------------------------------> |
| pushEvidence---------------------------------------------> |
| |
| evidenceAppraisal(evidence, eventLogDelta, refClaims)
| attestationResult <= |
| |
¶
Streaming Remote Attestation procedures require the setup of subscription state. Setting up subscription state between a Verifier and an Attester is conducted via a subscribe operation. This subscribe operation is used to convey the handles required for Evidence generation. Effectively, this allows for series of evidence to be pushed to a Verifier similar to the Uni-Directional model. While a Handle Distributor is not required in this model, it is also limited to bi-lateral subscription relationships, in which each Verifier has to create and provide its individual handle. Handles provided by a specific subscribing Verifier MUST be used in Evidence generation for that specific Verifier. The Streaming model uses the same information elements as the Challenge/Response and the Uni-Directional model. Methods to detect excessive time drift that would mandate a refreshed Handle to be conveyed via another subscribe operation are out-of-scope of this document.¶
Depending on the use cases covered, there can be additional requirements. An exemplary subset is illustrated in this section.¶
Confidentiality of exchanged attestation information may be desirable. This requirement usually is present when communication takes place over insecure channels, such as the public Internet. In such cases, TLS may be uses as a suitable communication protocol that preserves confidentiality. In private networks, such as carrier management networks, it must be evaluated whether or not the transport medium is considered confidential.¶
In particular use cases mutual authentication may be desirable in such a way that a Verifier also needs to prove its identity to the Attester, instead of only the Attester proving its identity to the Verifier.¶
Depending on given usage scenarios, hardware support for secure storage of cryptographic keys, crypto accelerators, as well as protected or isolated execution environments can be mandatory requirements. Well-known technologies in support of these requirements are roots of trusts, such as Hardware Security Modules (HSM), Physically Unclonable Functions (PUFs), Shielded Secrets, or Trusted Executions Environments (TEEs).¶
Note to RFC Editor: Please remove this section as well as references to [BCP205] before AUTH48.¶
This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in [BCP205]. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist.¶
According to [BCP205], "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".¶
The open-source implementation was initiated and is maintained by the Fraunhofer Institute for Secure Information Technology - SIT.¶
The open-source implementation is named "CHAllenge-Response based Remote Attestation" or in short: CHARRA.¶
The open-source implementation project resource can be located via: https://github.com/Fraunhofer-SIT/charra¶
The current version ('1bcb469') implements a challenge/response interaction model and is aligned with the exemplary specification of the CoAP FETCH bodies defined in Section Appendix A of this document.¶
The CHARRA project and all corresponding code and data maintained on github are provided under the BSD 3-Clause "New" or "Revised" license.¶
The implementation requires the use of the official Trusted Computing Group (TCG) open-source Trusted Software Stack (TSS) for the Trusted Platform Module (TPM) 2.0. The corresponding code and data is also maintained on github and the project resources can be located via: https://github.com/tpm2-software/tpm2-tss/¶
The implementation uses the Constrained Application Protocol [RFC7252] (http://coap.technology/) and the Concise Binary Object Representation [RFC7049] (https://cbor.io/).¶
In a remote attestation procedure the Verifier or the Attester MAY want to cryptographically blind several attributes. For instance, information can be part of the signature after applying a one-way function (e. g. a hash function).¶
There is also a possibility to scramble the Nonce or Attester Identity with other information that is known to both the Verifier and Attester. A prominent example is the IP address of the Attester that usually is known by the Attester itself as well as the Verifier. This extra information can be used to scramble the Nonce in order to counter certain types of relay attacks.¶
Olaf Bergmann, Michael Richardson, and Ned Smith¶
Changes from version 00 to version 01:¶
Changes from version 01 to version 02:¶
Changes from version 02 to version 00 RATS related document¶
Changes from version 00 to version 01¶
Changes from version 01 to version 02¶
Changes from version 02 to version 03¶
The following CDDL specification is an exemplary proof-of-concept to illustrate a potential implementation of the Challenge/Response Interaction Model. The transfer protocol used is CoAP using the FETCH operation. The actual resource operated on can be empty. Both the Challenge Message and the Response Message are exchanged via the FETCH operation and corresponding FETCH Request and FETCH Response body.¶
In this example, evidence is created via the root-of-trust for reporting primitive operation "quote" that is provided by a TPM 2.0.¶
RAIM-Bodies = CoAP-FETCH-Body / CoAP-FETCH-Response-Body
CoAP-FETCH-Body = [ hello: bool, ; if true, the AK-Cert is conveyed
nonce: bytes,
pcr-selection: [ + [ tcg-hash-alg-id: uint .size 2, ; TPM2_ALG_ID
[ + pcr: uint .size 1 ],
]
],
]
CoAP-FETCH-Response-Body = [ attestation-evidence: TPMS_ATTEST-quote,
tpm-native-signature: bytes,
? ak-cert: bytes, ; attestation key certificate
]
TPMS_ATTEST-quote = [ qualifiediSigner: uint .size 2, ;TPM2B_NAME
TPMS_CLOCK_INFO,
firmwareVersion: uint .size 8
quote-responses: [ * [ pcr: uint .size 1,
+ [ pcr-value: bytes,
? hash-alg-id: uint .size 2,
],
],
? pcr-digest: bytes,
],
]
TPMS_CLOCK_INFO = [ clock: uint .size 8,
resetCounter: uint .size 4,
restartCounter: uint .size 4,
save: bool,
]
¶