Microsoft

mbj@microsoft.com http://self-issued.info/

Nomura Research Institute

n-sakimura@nri.co.jp http://nat.sakimura.org/

Security JOSE Working Group JavaScript Object Notation JSON JSON Web Key JWK Thumbprint Fingerprint Digest This specification defines a method for computing a hash value over a JSON Web Key (JWK). It defines which fields in a JWK are used in the hash computation, the method of creating a canonical form for those fields, and how to convert the resulting Unicode string into a byte sequence to be hashed. The resulting hash value can be used for identifying or selecting the key represented by the JWK that is the subject of the thumbprint.

This specification defines a method for computing a hash value (a.k.a. digest) over a JSON Web Key (JWK) . It defines which fields in a JWK are used in the hash computation, the method of creating a canonical form for those fields, and how to convert the resulting Unicode string into a byte sequence to be hashed. The resulting hash value can be used for identifying or selecting the key represented by the JWK that is the subject of the thumbprint, for instance, by using the base64url-encoded JWK Thumbprint value as a kid (key ID) value.

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 "Key words for use in RFCs to Indicate Requirement Levels" . The interpretation should only be applied when the terms appear in all capital letters.

This specification uses the same terminology as the "JSON Web Key (JWK)" , "JSON Web Signature (JWS)" , and "JSON Web Algorithms (JWA)" specifications. This term is defined by this specification: The digest value for a JWK.

The thumbprint of a JSON Web Key (JWK) is computed as follows: Construct a JSON object containing only the required members of a JWK representing the key and with no whitespace or line breaks before or after any syntactic elements and with the required members ordered lexicographically by the Unicode code points of the member names. (This JSON object is itself a legal JWK representation of the key.) Hash the octets of the UTF-8 representation of this JSON object with a cryptographic hash function H. For example, SHA-256 might be used as H. See for a discussion on the choice of hash function. The resulting value is the JWK Thumbprint with H of the JWK. The details of this computation are further described in subsequent sections.

This section demonstrates the JWK Thumbprint computation for the JWK below (with the long line broken for display purposes only):

As defined in "JSON Web Key (JWK)" and "JSON Web Algorithms (JWA)" , the required members for an RSA public key are: kty n e Therefore, these are the members used in the thumbprint computation. Their lexicographic order, per , is: e kty n Therefore, the JSON object constructed as an intermediate step in the computation is as follows (with the line broken for display purposes only):

The octets of the UTF-8 representation of this JSON object are: [123, 34, 101, 34, 58, 34, 65, 81, 65, 66, 34, 44, 34, 107, 116, 121, 34, 58, 34, 82, 83, 65, 34, 44, 34, 110, 34, 58, 34, 48, 118, 120, 55, 97, 103, 111, 101, 98, 71, 99, 81, 83, 117, 117, 80, 105, 76, 74, 88, 90, 112, 116, 78, 57, 110, 110, 100, 114, 81, 109, 98, 88, 69, 112, 115, 50, 97, 105, 65, 70, 98, 87, 104, 77, 55, 56, 76, 104, 87, 120, 52, 99, 98, 98, 102, 65, 65, 116, 86, 84, 56, 54, 122, 119, 117, 49, 82, 75, 55, 97, 80, 70, 70, 120, 117, 104, 68, 82, 49, 76, 54, 116, 83, 111, 99, 95, 66, 74, 69, 67, 80, 101, 98, 87, 75, 82, 88, 106, 66, 90, 67, 105, 70, 86, 52, 110, 51, 111, 107, 110, 106, 104, 77, 115, 116, 110, 54, 52, 116, 90, 95, 50, 87, 45, 53, 74, 115, 71, 89, 52, 72, 99, 53, 110, 57, 121, 66, 88, 65, 114, 119, 108, 57, 51, 108, 113, 116, 55, 95, 82, 78, 53, 119, 54, 67, 102, 48, 104, 52, 81, 121, 81, 53, 118, 45, 54, 53, 89, 71, 106, 81, 82, 48, 95, 70, 68, 87, 50, 81, 118, 122, 113, 89, 51, 54, 56, 81, 81, 77, 105, 99, 65, 116, 97, 83, 113, 122, 115, 56, 75, 74, 90, 103, 110, 89, 98, 57, 99, 55, 100, 48, 122, 103, 100, 65, 90, 72, 122, 117, 54, 113, 77, 81, 118, 82, 76, 53, 104, 97, 106, 114, 110, 49, 110, 57, 49, 67, 98, 79, 112, 98, 73, 83, 68, 48, 56, 113, 78, 76, 121, 114, 100, 107, 116, 45, 98, 70, 84, 87, 104, 65, 73, 52, 118, 77, 81, 70, 104, 54, 87, 101, 90, 117, 48, 102, 77, 52, 108, 70, 100, 50, 78, 99, 82, 119, 114, 51, 88, 80, 107, 115, 73, 78, 72, 97, 81, 45, 71, 95, 120, 66, 110, 105, 73, 113, 98, 119, 48, 76, 115, 49, 106, 70, 52, 52, 45, 99, 115, 70, 67, 117, 114, 45, 107, 69, 103, 85, 56, 97, 119, 97, 112, 74, 122, 75, 110, 113, 68, 75, 103, 119, 34, 125] Using SHA-256 as the hash function H, the JWK SHA-256 Thumbprint value is the SHA-256 hash of these octets, specifically: [55, 54, 203, 177, 120, 124, 184, 48, 156, 119, 238, 140, 55, 5, 197, 225, 111, 251, 158, 133, 151, 21, 144, 31, 30, 76, 89, 177, 17, 130, 245, 123] The base64url encoding of this JWK SHA-256 Thumbprint value (which might, for instance, be used as a kid (key ID) value) is:

Only the required members of a key's representation are used when computing its JWK Thumbprint value. As defined in "JSON Web Key (JWK)" and "JSON Web Algorithms (JWA)" , the required members for an elliptic curve public key for the curves specified in Section 6.2.1.1 of RFC 7518 , in lexicographic order, are: crv kty x y The required members for an RSA public key, in lexicographic order, are: e kty n The required members for a symmetric key, in lexicographic order, are: k kty As other kty (key type) values are defined, the specifications defining them should be similarly consulted to determine which members, in addition to kty, are required.

The JWK Thumbprint of a JWK representing a private key is computed as the JWK Thumbprint of a JWK representing the corresponding public key. This has the intentional benefit that the same JWK Thumbprint value can be computed both by parties using either the public or private key. The JWK Thumbprint can then be used to refer to both keys of the key pair. Application context can be used to determine if the public or private key is the one being referred to by the JWK Thumbprint. This specification defines the method of computing JWK Thumbprints of JWKs representing private keys for interoperability reasons -- so that different implementations computing JWK Thumbprints of private keys will produce the same result.

Optional members of JWKs are intentionally not included in the JWK Thumbprint computation so that their absence or presence in the JWK does not alter the resulting value. The JWK Thumbprint value is a digest of the members required to represent the key as a JWK -- not of additional data that may also accompany the key. Optional members are not included so that the JWK Thumbprint refers to a key -- not a key with an associated set of key attributes. Different application contexts might or might not include different subsets of optional attributes about the key in the JWK. If these were included in the calculation of the JWK thumbprint, the values would be different for those JWKs, even though the keys are the same. The benefit of including only the JWK required members is that the JWK Thumbprint of any JWK representing the key remains the same, regardless of any other attributes that are present. Different kinds of thumbprints could be defined by other specifications that might include some or all additional JWK members, if use cases arise where such different kinds of thumbprints would be useful. See Section 9.1 of RFC 7517 for notes on some ways to cryptographically bind attributes to a key.

The required members in the input to the hash function are ordered lexicographically by the Unicode code points of the member names. Characters in member names and member values MUST be represented without being escaped. This means that thumbprints of JWKs that require such characters are not defined by this specification. (This is not expected to limit the applicability of this specification, in practice, as the members of JWK representations are not expected to use any of these characters.) The characters specified as requiring escaping by Section 7 of are quotation mark, reverse solidus (a.k.a. backslash), and the control characters U+0000 through U+001F. If the JWK key type uses members whose values are themselves JSON objects, then the members of those objects MUST likewise be lexicographically ordered. (As of the time of this writing, none are defined that do.) If the JWK key type uses members whose values are JSON numbers, and if those numbers are integers, then they MUST be represented as a JSON number as defined in Section 6 of without including a fraction part or exponent part. For instance, the value 1.024e3 MUST be represented as 1024. This means that thumbprints of JWKs using numbers that are not integers are not defined by this specification. Also, as noted in "The I-JSON Message Format" , implementations cannot expect an integer whose absolute value is greater than 9007199254740991 (i.e., that is outside the range [-(2**53)+1, (2**53)-1]) to be treated as an exact value. (As of the time of this writing, none are defined that use JSON numbers.) See for a discussion of further practical considerations pertaining to the representation of the hash input.

A specific hash function must be chosen by an application to compute the hash value of the hash input. For example, SHA-256 might be used as the hash function by the application. While SHA-256 is a good default choice at the time of this writing, the hash function of choice can be expected to change over time as the cryptographic landscape evolves. Note that in many cases, only the party that creates a key will need to know the hash function used. A typical usage is for the producer of the key to use the base64url-encoded JWK Thumbprint value as a kid (key ID) value. In this case, the consumer of the kid treats it as an opaque value that it uses to select the key. However, in some cases, multiple parties will be reproducing the JWK Thumbprint calculation and comparing the results. In these cases, the parties will need to know which hash function was used and use the same one.

Note that a key need not be in JWK format to create a JWK Thumbprint of it. The only prerequisites are that the JWK representation of the key be defined and the party creating the JWK Thumbprint be in possession of the necessary key material. These are sufficient to create the hash input from the JWK representation of the key, as described in .

Implementations will almost certainly use functionality provided by the platform's JSON support when parsing the JWK and emitting the JSON object used as the hash input. As a practical consideration, future JWK member names and values should be avoided for which different platforms or libraries might emit different representations. As of the time of this writing, all defined JWK member names and values use only printable ASCII characters, which should not exhibit this problem. Note however, that JSON.stringify() cannot be counted on to lexicographically sort the members of JSON objects, so while it could be used to emit some kinds of member values, different code is likely to be needed to perform the sorting. In particular, while the operation of lexicographically ordering member names by their Unicode code points is well defined, different platform sort functions may produce different results for non-ASCII characters, in ways that may not be obvious to developers. If writers of future specifications defining new JWK key type values choose to restrict themselves to printable ASCII member names and values (which are for machine and not human consumption anyway), some future interoperability problems might be avoided. However, if new JWK members are defined that use non-ASCII member names or values, their definitions should specify the exact Unicode code point sequences used to represent them. This is particularly important in cases in which Unicode normalization could result in the transformation of one set of code points into another under any circumstances. Use of escaped characters in JWKs for which JWK Thumbprints will be computed should be avoided. Use of escaped characters in the hash input JWKs derived from these original JWKs is prohibited. There is a natural representation to use for numeric values that are integers. However, this specification does not attempt to define a standard representation for numbers that are not integers or that contain an exponent component. This is not expected to be a problem in practice, as the required members of JWK representations are expected to use only numbers that are integers. Use of number representations containing fraction or exponent parts in JWKs for which JWK Thumbprints will be computed should be avoided. All of these practical considerations are really an instance of Jon Postel's principle: "Be liberal in what you accept, and conservative in what you send."

JWK Thumbprint values are computed on the JWK members required to represent a key, rather than all members of a JWK that the key is represented in. Thus, they are more analogous to applications that use digests of X.509 Subject Public Key Info (SPKI) values, which are defined in Section 4.1.2.7 of , than to applications that use digests of complete certificate values, as the x5t (X.509 certificate SHA-1 thumbprint) value defined for X.509 certificate objects does. While logically equivalent to a digest of the SPKI representation of the key, a JWK Thumbprint is computed over a JSON representation of that key, rather than over an ASN.1 representation of it.

This specification adds to the instructions for the Designated Experts of the following IANA registries, all of which are in the "JSON Object Signing and Encryption (JOSE)" registry : JSON Web Key Types JSON Web Key Elliptic Curve JSON Web Key Parameters IANA has added a link to this specification in the Reference sections of these registries. For these registries, because of the practical JSON and Unicode considerations described in , the Designated Experts must either: (a) require that JWK member names and values being registered use only printable ASCII characters excluding double quote ('"') and backslash ('\') (the Unicode characters with code points U+0021, U+0023 through U+005B, and U+005D through U+007E), or (b) if new JWK members or values are defined that use other code points, require that their definitions specify the exact Unicode code point sequences used to represent them. Furthermore, proposed registrations that use Unicode code points that can only be represented in JSON strings as escaped characters must not be accepted.

The JSON Security Considerations and Unicode Comparison Security Considerations described in Sections 10.12 and 10.13 of "JSON Web Signature (JWS)" also apply to this specification. Also, as described in , some implementations may produce incorrect results if esoteric or escaped characters are used in the member names. The security implications of this appear to be limited for JWK Thumbprints of public keys, because while it may result in implementations failing to identify the intended key, it should not leak information. The information in a public key is already public in nature, by definition. A hash of a symmetric key has the potential to leak information about the key value. Thus, the JWK Thumbprint of a symmetric key should typically be concealed from parties not in possession of the symmetric key, unless in the application context, the cryptographic hash used, such as SHA-256, is known to provide sufficient protection against disclosure of the key value. A JWK Thumbprint will only uniquely identify a particular key if a single unambiguous JWK representation for that key is defined and used when computing the JWK Thumbprint. (Such representations are defined for all the key types defined in "JSON Web Algorithms (JWA)" .) For example, if an RSA key were to use "e":"AAEAAQ" (representing [0, 1, 0, 1]) rather than the specified correct representation of "e":"AQAB" (representing [1, 0, 1]), then a different thumbprint value would be produced for what could be effectively the same key, at least for implementations that are lax in validating the JWK values that they accept. Thus, JWK Thumbprint values can only be relied upon to be unique for a given key if the implementation also validates that the correct representation of the key is used. Even more insidious is that an attacker may supply a key that is a transformation of a legal key in order to have it appear to be a different key. For instance, if a legitimate RSA key uses a modulus value N and an attacker supplies a key with modulus 3*N, the modified key would still work about 1/3 of the time, but would appear to be a different key. Thus, while thumbprint values are valuable for identifying legitimate keys, comparing thumbprint values is not a reliable means of excluding (blacklisting) the use of particular keys (or transformations thereof).

IANA This specification registers cryptographic algorithms and identifiers to be used with the JSON Web Signature (JWS), JSON Web Encryption (JWE), and JSON Web Key (JWK) specifications. It defines several IANA registries for these identifiers. A JSON Web Key (JWK) is a JavaScript Object Notation (JSON) data structure that represents a cryptographic key. This specification also defines a JWK Set JSON data structure that represents a set of JWKs. Cryptographic algorithms and identifiers for use with this specification are described in the separate JSON Web Algorithms (JWA) specification and IANA registries established by that specification. JSON Web Signature (JWS) represents content secured with digital signatures or Message Authentication Codes (MACs) using JSON-based data structures. Cryptographic algorithms and identifiers for use with this specification are described in the separate JSON Web Algorithms (JWA) specification and an IANA registry defined by that specification. Related encryption capabilities are described in the separate JSON Web Encryption (JWE) specification. National Institute of Standards and Technology The Unicode Consortium

James Manger and John Bradley participated in discussions that led to the creation of this specification. Thanks also to Joel Halpern, Barry Leiba, Adam Montville, Kathleen Moriarty, and Jim Schaad for their reviews of this specification.