Internet-Draft ACVP ML-DSA April 2024
Celi Expires 1 November 2024 [Page]
Workgroup:
Network Working Group
Internet-Draft:
draft-celi-acvp-ml-dsa-01
:
Published:
Intended Status:
Informational
Expires:
Author:
C. Celi, Ed.

ACVP ML-DSA JSON Specification

Status of This Memo

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 1 November 2024.

▲

Table of Contents

1. Acknowledgements

There are no acknowledgements.

2. Abstract

This document defines the JSON schema for testing Module Lattice-based Digital Signature Algorithm (ML-DSA) implementations with the ACVP specification.

3. Introduction

The Automated Crypto Validation Protocol (ACVP) defines a mechanism to automatically verify the cryptographic implementation of a software or hardware crypto module. The ACVP specification defines how a crypto module communicates with an ACVP server, including crypto capabilities negotiation, session management, authentication, vector processing and more. The ACVP specification does not define algorithm specific JSON constructs for performing the crypto validation. A series of ACVP sub-specifications define the constructs for testing individual crypto algorithms. Each sub-specification addresses a specific class of crypto algorithms. This sub-specification defines the JSON constructs for testing Module Lattice-based Digital Signature Algorithm (ML-DSA) implementations using ACVP.

4. Conventions

4.1. Notation conventions

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 of [RFC2119] and [RFC8174] when, and only when, they appear in all capitals, as shown here.

4.2. Terms and Definitions

4.2.1. Prompt

JSON sent from the server to the client describing the tests the client performs

4.2.2. Registration

The initial request from the client to the server describing the capabilities of one or several algorithm, mode and revision combinations

4.2.3. Response

JSON sent from the client to the server in response to the prompt

4.2.4. Test Case

An individual unit of work within a prompt or response

4.2.5. Test Group

A collection of test cases that share similar properties within a prompt or response

4.2.6. Test Vector Set

A collection of test groups under a specific algorithm, mode, and revision

4.2.7. Validation

JSON sent from the server to the client that specifies the correctness of the response

5. Supported ML-DSA Algorithms

The following ML-DSA algorithms MAY be advertised by the ACVP compliant cryptographic module. The list is in the form "algorithm / mode / revision".

6. Test Types and Test Coverage

6.1. Test Types

The ACVP server performs a set of tests on the specified ML-DSA algorithm in order to assess the correctness and robustness of the implementation. A typical ACVP validation session SHALL require multiple tests to be performed for every supported permutation of ML-DSA capabilities. This section describes the design of the tests used to validate implementations of the ML-DSA algorithms.

  • ML-DSA / keyGen / * "AFT" - Algorithm Functional Test. The IUT is REQUIRED for each test case provided, to generate a key pair from a provided seed. The key pair is communicated to the ACVP server and validated.
  • ML-DSA / sigGen / * "AFT" - Algorithm Functional Test. This testing mode expects the IUT to generate valid signatures based on the ACVP provided message. The signature is then compared to the known result by the ACVP server given the IUT's communicated parameter set, public key, and signature.
  • ML-DSA / sigGen / * "GDT" - Generated Data Test. This testing mode expects the IUT to generate a valid signature and key pair for the ACVP provided message. The client SHALL communicate the public key and signature back to the server. The signature is then validated by the ACVP server using the defined signature verification routine.
  • ML-DSA / sigVer / * "AFT" - Algorithm Functional Test. The ACVP server generates a series of signatures to communicate to the IUT. The IUT is REQUIRED to determine the validity of the signature given the parameter set, key, and message.

6.2. Test Coverage

The tests described in this document have the intention of ensuring an implementation is conformant to [FIPS204].

6.2.1. Requirements Covered

  • The tests will ensure conformity and correctness of an implementation of the algorithms supported.

6.2.2. Requirements Not Covered

  • FIPS 204 Section 3.5. Additional Requirements. Requirements outlined in this section are not testable by an ACVP server. An ACVP server will not test the zeroization of intermediate values, security strength of the deterministic random bit generators (DRBGs), or incorrect length signatures or public keys.

7. Capabilities Registration

ACVP requires crypto modules to register their capabilities. This allows the crypto module to advertise support for specific algorithms, notifying the ACVP server which algorithms need test vectors generated for the validation process. This section describes the constructs for advertising support of ML-DSA algorithms to the ACVP server.

The algorithm capabilities MUST be advertised as JSON objects within the 'algorithms' value of the ACVP registration message. The 'algorithms' value is an array, where each array element is an individual JSON object defined in this section. The 'algorithms' value is part of the 'capability_exchange' element of the ACVP JSON registration message. See the ACVP specification [ACVP] for more details on the registration message.

7.1. Prerequisites

Each algorithm implementation MAY rely on other cryptographic primitives. For example, RSA Signature algorithms depend on an underlying hash function. Each of these underlying algorithm primitives must be validated, either separately or as part of the same submission. ACVP provides a mechanism for specifying the required prerequisites:

Prerequisites, if applicable, MUST be submitted in the registration as the prereqVals JSON property array inside each element of the algorithms array. Each element in the prereqVals array MUST contain the following properties

Table 1: Prerequisite Properties
JSON Property Description JSON Type
algorithm a prerequisite algorithm string
valValue algorithm validation number string

A "valValue" of "same" SHALL be used to indicate that the prerequisite is being met by a different algorithm in the capability exchange in the same registration.

An example description of prerequisites within a single algorithm capability exchange looks like this

"prereqVals":
[
  {
    "algorithm": "Alg1",
    "valValue": "Val-1234"
  },
  {
    "algorithm": "Alg2",
    "valValue": "same"
  }
]

7.2. Required Prerequisite Algorithms for ML-DSA Validations

Each ML-DSA implementation relies on other cryptographic primitives. For example, ML-DSA keyGen uses an underlying SHA algorithm. Each of these underlying algorithm primitives must be validated, either separately or as part of the same submission. ACVP provides a mechanism for specifying the required prerequisites:

Table 2: Required ML-DSA Prerequisite Algorithms JSON Values
JSON Value Description JSON Type Valid Values
algorithm a prerequisite algorithm string SHA, or DRBG
valValue algorithm validation number string Actual number or "same"
prereqAlgVal prerequisite algorithm validation object with algorithm and valValue properties See above

7.3. ML-DSA Algorithm Registration Properties

Each ML-DSA algorithm capability advertised is a self-contained JSON object using the following values.

Table 3: ML-DSA Algorithm Capabilities JSON Values
JSON Value Description JSON Type Valid Values
algorithm The ML-DSA algorithm to be validated string See Section 5
mode The ML-DSA mode to be validated string See Section 5
revision The algorithm testing revision to use string See Section 5
prereqVals Prerequisite algorithm validations array of prereqAlgVal objects See Section 7.2
parameterSets The ML-DSA parameter sets supported array of strings "ML-DSA-44", "ML-DSA-65", "ML-DSA-87"
deterministic The ML-DSA signature generation modes supported array of booleans true, false

Note: The deterministic property is only applicable to ML-DSA / sigGen / FIPS204.

7.3.1. ML-DSA keyGen Mode Capabilities Example

Below is an example of the registration for ML-DSA / keyGen / FIPS204

{
    "algorithm": "ML-DSA",
    "mode": "keyGen",
    "revision": "FIPS204",
    "prereqVals": [
        {
            "algorithm": "SHA",
            "valValue": "123456"
        }
    ],
    "parameterSets": ["ML-DSA-44", "ML-DSA-65", "ML-DSA-87"]
}

7.3.2. ML-DSA sigGen Mode Capabilities Example

Below is an example of the registration for ML-DSA / sigGen / FIPS204

{
    "algorithm": "ML-DSA",
    "mode": "sigGen",
    "revision": "FIPS204",
    "prereqVals": [
        {
            "algorithm": "SHA",
            "valValue": "123456"
        }
    ],
    "parameterSets": ["ML-DSA-44", "ML-DSA-65", "ML-DSA-87"],
    "deterministic": [true, false]
}

7.3.3. ML-DSA sigVer Mode Capabilities Example

Below is an example of the registration for ML-DSA / sigVer / FIPS204

{
    "algorithm": "ML-DSA",
    "mode": "sigVer",
    "revision": "FIPS204",
    "prereqVals": [
        {
            "algorithm": "SHA",
            "valValue": "123456"
        }
    ],
    "parameterSets": ["ML-DSA-44", "ML-DSA-65", "ML-DSA-87"]
}

8. Test Vectors

The ACVP server provides test vectors to the ACVP client, which are then processed and returned to the ACVP server for validation. A typical ACVP validation test session would require multiple test vector sets to be downloaded and processed by the ACVP client. Each test vector set represents an individual algorithm defined during the capability exchange. This section describes the JSON schema for a test vector set used with Module Lattice-based Digital Signature Algorithm (ML-DSA) algorithms.

The test vector set JSON schema is a multi-level hierarchy that contains meta data for the entire vector set as well as individual test vectors to be processed by the ACVP client. The following table describes the JSON elements at the top level of the hierarchy.

Table 4: Top Level Test Vector JSON Elements
JSON Values Description JSON Type
acvVersion Protocol version identifier string
vsId Unique numeric vector set identifier integer
algorithm Algorithm defined in the capability exchange string
mode Mode defined in the capability exchange string
revision Protocol test revision selected string
testGroups Array of test groups containing test data, see Section 8, Paragraph 6 array

An example of this would look like this

[
  {
    "acvVersion": <version>
  },
  {
    "vsId": 1,
    "algorithm": "Alg1",
    "mode": "Mode1",
    "revision": "Revision1.0",
    "testGroups": [ ... ]
  }
]

==

8.1. ML-DSA keyGen Test Vectors

8.1.1. ML-DSA keyGen Test Groups JSON Schema

The testGroups element at the top level in the test vector JSON object is an array of test groups. Test vectors are grouped into similar test cases to reduce the amount of data transmitted in the vector set. For instance, all test vectors that use the same key size would be grouped together. The Test Group JSON object contains meta data that applies to all test vectors within the group. The following table describes the secure hash JSON elements of the Test Group JSON object.

The test group for ML-DSA / keyGen / FIPS204 is as follows:

Table 5: ML-DSA keyGen Test Group JSON Object
JSON Value Description JSON type
testType The test operation performed string
parameterSet The ML-DSA parameter set used string
tests Array of individual test vector JSON objects, which are defined in Section 8.1.2 array

8.1.2. ML-DSA keyGen Test Case JSON Schema

Each test group contains an array of one or more test cases. Each test case is a JSON object that represents a single test vector to be processed by the ACVP client. The following table describes the JSON elements for each ML-DSA test vector.

Table 6: ML-DSA keyGen Test Case JSON Object
JSON Value Description JSON type
tcId Numeric identifier for the test case, unique across the entire vector set integer
seed The seed used to generate the key pair hex

The following is an example JSON object sent from the server to the client for ML-DSA / keyGen / FIPS204.

[
    {
        "acvVersion": <acvp-version>
    },
    {
        "vsId": 1564,
        "algorithm": "ML-DSA",
        "mode": "keyGen",
        "revision": "FIPS204",
        "testGroups": [
            {
                "tgId": 1,
                "testType": "AFT",
                "parameterSet": "ML-DSA-44",
                "tests": [
                    {
                        "tcId": 1,
                        "seed": "C105DC2..."
                    }
                ]
            }
        ]
    }
]

8.2. ML-DSA sigGen Test Vectors

8.2.1. ML-DSA sigGen Test Groups JSON Schema

The testGroups element at the top level in the test vector JSON object is an array of test groups. Test vectors are grouped into similar test cases to reduce the amount of data transmitted in the vector set. For instance, all test vectors that use the same key size would be grouped together. The Test Group JSON object contains meta data that applies to all test vectors within the group. The following table describes the secure hash JSON elements of the Test Group JSON object.

The test group for ML-DSA / sigGen / FIPS204 is as follows:

Table 7: ML-DSA sigGen Test Group JSON Object
JSON Value Description JSON type
testType The test operation performed string
parameterSet The ML-DSA parameter set used string
deterministic Whether the signatures should be generated using the deterministic or non-deterministic routine boolean
tests Array of individual test vector JSON objects, which are defined in Section 8.2.2 array

8.2.2. ML-DSA sigGen Test Groups JSON Schema

Each test group contains an array of one or more test cases. Each test case is a JSON object that represents a single test vector to be processed by the ACVP client. The following table describes the JSON elements for each ML-DSA test vector.

Table 8: ML-DSA sigGen Test Case JSON Object
JSON Value Description JSON type
tcId Numeric identifier for the test case, unique across the entire vector set integer
message The message used to generate the signature hex
sk When the test group property "testType": "AFT", the secret key that should be used to generate a signature hex
rnd When the test group properties "testType": "AFT" and "deterministic": false, the random value used to generate the signature hex

The following is an example JSON object sent from the server to the client for ML-DSA / sigGen / FIPS204.

[
    {
        "acvVersion": <acvp-version>
    },
    {
        "vsId": 0,
        "algorithm": "ML-DSA",
        "mode": "sigGen",
        "revision": "FIPS204",
        "testGroups": [
            {
                "tgId": 1,
                "testType": "GDT",
                "parameterSet": "ML-DSA-44",
                "deterministic": true,
                "tests": [
                    {
                        "tcId": 1,
                        "message": "0D873AEFD..."
                    }
                ]
            },
            {
                "tgId": 2,
                "testType": "AFT",
                "parameterSet": "ML-DSA-44",
                "deterministic": true,
                "tests": [
                    {
                        "tcId": 61,
                        "sk": "E94EB...",
                        "message": "00E959E..."
                    }
                ]
            },
            {
                "tgId": 3,
                "testType": "AFT",
                "parameterSet": "ML-DSA-44",
                "deterministic": false,
                "tests": [
                    {
                        "tcId": 71,
                        "sk": "D4E36A5...",
                        "message": "34F4B...",
                        "rnd": "7AFF2F22A..."
                    }
                ]
            }
        ]
    }
]

8.3. ML-DSA sigVer Test Vectors

8.3.1. ML-DSA sigVer Test Groups JSON Schema

The testGroups element at the top level in the test vector JSON object is an array of test groups. Test vectors are grouped into similar test cases to reduce the amount of data transmitted in the vector set. For instance, all test vectors that use the same key size would be grouped together. The Test Group JSON object contains meta data that applies to all test vectors within the group. The following table describes the secure hash JSON elements of the Test Group JSON object.

The test group for ML-DSA / sigVer / FIPS204 is as follows:

Table 9: ML-DSA sigVer Test Group JSON Object
JSON Value Description JSON type
testType The test operation performed string
parameterSet The ML-DSA parameter set used string
pk The public key used to generate the signature hex
tests Array of individual test vector JSON objects, which are defined in Section 8.3.2 array

8.3.2. ML-DSA sigVer Test Groups JSON Schema

Each test group contains an array of one or more test cases. Each test case is a JSON object that represents a single test vector to be processed by the ACVP client. The following table describes the JSON elements for each ML-DSA test vector.

Table 10: ML-DSA sigVer Test Case JSON Object
JSON Value Description JSON type
tcId Numeric identifier for the test case, unique across the entire vector set integer
message The message used to verify with the signature hex
signature The signature to verify hex

The following is an example JSON object sent from the server to the client for ML-DSA / sigVer / FIPS204.

[
    {
        "acvVersion": <acvp-version>
    },
    {
        "vsId": 0,
        "algorithm": "ML-DSA",
        "mode": "sigVer",
        "revision": "FIPS204",
        "isSample": false,
        "testGroups": [
            {
                "tgId": 1,
                "testType": "AFT",
                "parameterSet": "ML-DSA-44",
                "publicKey": "3FE65294...",
                "tests": [
                    {
                        "tcId": 1,
                        "message": "4F0D7...",
                        "signature": "C29A11B6C..."
                    },
                    {
                        "tcId": 2,
                        "message": "84793...",
                        "signature": "6A99215FC2..."
                    }
                ]
            }
        ]
    }
]

9. Test Vector Responses

After the ACVP client downloads and processes a vector set, it must send the response vectors back to the ACVP server. The following table describes the JSON object that represents a vector set response.

Table 11: Response JSON Object
JSON Property Description JSON Type
acvVersion The ACVP version used string
vsId The vector set identifier integer
testGroups The test group objects in the response, see Table 12 array

An example of this is the following

{
    "acvVersion": "version",
    "vsId": 1,
    "testGroups": [ ... ]
}

The 'testGroups' section is used to organize the ACVP client response in a similar manner to how it distributes vectors.

Table 12: Response Group Objects
JSON Property Description JSON Type
tgId The test group identifier integer
tests The test case objects in the response, depending on the algorithm see Table 13, Table 15, or Table 16 array

An example of this is the following

{
    "tgId": 1,
    "tests": [ ... ]
}

9.1. ML-DSA keyGen Test Vector Responses

Each test group contains an array of one or more test cases. Each test case is a JSON object that represents a single test vector to be processed by the ACVP client. The following table describes the JSON elements for each ML-DSA / keyGen / FIPS204 test vector.

Table 13: ML-DSA keyGen Test Case Response JSON Object
JSON Value Description JSON type
tcId The test case identifier integer
pk The computed public key hex
sk The computed secret key hex

The following is an example JSON test vector response object for ML-DSA / keyGen / FIPS204.

[
    {
        "acvVersion": <acvp-version>
    },
    {
        "vsId": 0,
        "testGroups": [
            {
                "tgId": 1,
                "tests": [
                    {
                        "tcId": 1,
                        "pk": "1012798...",
                        "sk": "1012798..."
                    }
                ]
            }
        ]
    }
]

9.2. ML-DSA sigGen Test Vector Responses

The test groups for ML-DSA / sigGen / FIPS204 contain public key properties. The groups can be described using the following table.

Table 14: ML-DSA sigGen Test Group Response JSON Object
JSON Value Description JSON type
tgId The test group identifier integer
pk The public key the IUT used when the prompt test group property "testType": "GDT" hex
tests The individual test cases for the group array

Each test group contains an array of one or more test cases. Each test case is a JSON object that represents a single test vector to be processed by the ACVP client. The following table describes the JSON elements for each ML-DSA / sigGen / FIPS204 test vector.

Table 15: ML-DSA sigGen Test Case Response JSON Object
JSON Value Description JSON type
tcId The test case identifier integer
signature The generated signature hex

The following is an example JSON test vector response object for ML-DSA / sigGen / FIPS204.

[
    {
        "acvVersion": <acvp-version>
    },
    {
        "vsId": 1564,
        "testGroups": [
            {
                "tgId": 1,
                "pk": "A9ADC1E82...",
                "tests": [
                    {
                        "tcId": 1,
                        "signature": "D7D4275..."
                    }
                ]
            },
            {
                "tgId": 2,
                "tests": [
                    {
                        "tcId": 2,
                        "signature": "75300EBBB6C..."
                    }
                ]
            }
        ]
    }
]

9.3. ML-DSA sigVer Test Vector Responses

Each test group contains an array of one or more test cases. Each test case is a JSON object that represents a single test vector to be processed by the ACVP client. The following table describes the JSON elements for each ML-DSA / sigVer / FIPS204 test vector.

Table 16: ML-DSA sigVer Test Case Response JSON Object
JSON Value Description JSON type
tcId The test case identifier integer
testPassed Whether or not the signature verified boolean

The following is an example JSON test vector response object for ML-DSA / sigVer / FIPS204.

[
    {
        "acvVersion": <acvp-version>
    },
    {
        "vsId": 0,
        "testGroups": [
            {
                "tgId": 1,
                "tests": [
                    {
                        "tcId": 1,
                        "testPassed": false
                    }
                ]
            }
        ]
    }
]

10. Security Considerations

There are no additional security considerations outside of those outlined in the ACVP document.

11. IANA Considerations

This document does not require any action by IANA.

12. Normative References

[RFC2119]
Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", RFC 2119, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/info/rfc2119>.
[RFC8174]
Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", RFC 8174, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/info/rfc8174>.
[FIPS204]
"", NIST FIPS 204.
[ACVP]
Fussell, B., Vassilev, A., and H. Booth, "Automatic Cryptographic Validation Protocol", ACVP, .

Author's Address

Christopher Celi (editor)