| Internet Engineering Task Force | S. Farrell |
| Internet-Draft | Trinity College Dublin |
| Intended status: Standards Track | D. Kutscher |
| Expires: October 01, 2012 | NEC |
| C. Dannewitz | |
| University of Paderborn | |
| B. Ohlman | |
| A. Keranen | |
| Ericsson | |
| P. M. Hallam-Baker | |
| Comodo Group Inc. | |
| April 2012 |
Naming Things with Hashes
draft-farrell-decade-ni-04
This document defines a set of ways to identify a thing using the output from a hash function, specifying URI, URL, binary and human "speakable" formats for these names. The various formats are designed to support, but not require, a strong link to the referenced object such that the referenced object may be authenticated to the same degree as the reference to it.
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 http://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 October 01, 2012.
Copyright (c) 2012 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 (http://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.
Names or identifiers are used in various protocols for identifying resources. In many scenarios those names or identifiers contain values that are hash function outputs. However, different deployments have chosen various different ways to include hash function outputs in such names or identifiers. This document specifies standard ways to do that to aid interoperability.
Hash function outputs can be used to ensure uniqueness in terms of mapping URIs [RFC3986] to a specific resource, or to make URIs hard to guess for security reasons. Since, there is no standard way to interpret those strings, today in general only the creator of the URI knows how to use the hash function output. Other protocols, such as application layer protocols for accessing "smart objects" in constrained environments also require more compact (e.g., binary) forms of such identifiers, while in other situations people may have to input such values or talk about them, e.g., in a voice call.
As another example, protocols for accessing in-network storage servers need a way to identify stored resources uniquely and in a location-independent way so that replicas on different servers can be accessed by the same name. Also, such applications may require verifying that a resource representation that has been obtained actually corresponds to the name that was used to request the resource, i.e., verifying the integrity of the name-data binding.
Similarly, in the context of information-centric networking [ref.netinf-design] [ref.ccn] and elsewhere there is value in being able to compare a presented resource against the URI that was dereferenced in order to access that resource. If a cryptographically-strong comparison function can be used then this allows for many forms of in-network storage, without requiring as much trust in the infrastructure used to present the resource. The outputs of hash functions can be used in this manner, if presented in a standard way.
Additional applications might include creating references from web pages delivered over HTTP/TLS; DNS resource records signed using DNSSEC or data values embedded in certificates, Certificate Revocation Lists (CRLs), or other signed data objects.
The new URI scheme defined here allows for the use of a query-string, similar to how query-strings are used in HTTP URLs. A companion specification [niexts] describes specific values that can be used in such query strings for various purposes and other extensions to this basic format specification.
The "ni" URI scheme defined here is very similar to the "magnet link" informally defined in various other protocols. [magnet]
In addition to the URI form we also define a ".well-known" URL equivalent, and a way to include a hash as a segment of an HTTP URL, as well as a binary format for use in protocols that require more compact names and a human-speakable text form that could be used, e.g. for reading out (parts of) the name over a voice connection.
Not all uses of these names require use of the full hash output - truncated hashes can be safely used in some environments. For this reason, we define a new IANA registry for hash functions to be used with this specification so as not to mix strong and weak (truncated) hash algorithms in other protocol registries.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].
Syntax definitions in this memo are specified according to ABNF [RFC5234].
This section contains basic considerations common to all formats.
When verifying whether two names refer to same object, an implementation MUST only consider the digest algorithm identifier and the digest value, i.e., it MUST NOT consider the authority field from a URI or any parameters and MUST consider two hashes identical regardless of encoding, if they encode the same binary value, and are the same length.
The sha-256 algorithm as specified in [RFC4055] is mandatory to implement, that is, implementations MUST be able to generate/send and to accept/process names based on a sha-256 hash. However implementations MAY support additional hash algorithms and MAY use those for specific names, for example in a constrained environment where sha-256 is non-optimal or where truncated names are needed to fit into corresponding protocols (when a higher collision probability can be tolerated).
Truncated hashes MAY be supported if needed. When a hash value is truncated the name MUST indicate this. Therefore we use different hash algorithm strings for these, such as sha-256-32 for a 32-bit truncation of a sha-256 output. (Note that a 32-bit truncated hash is essentially useless for security but might be useful for naming.)
When a hash value is truncated to N bits the left-most or most significant in network byte order N bits from the binary representation of the hash value MUST be used as the truncated value. An example of a 128-bit hash output truncated to 32 bits is shown in Figure 1.
128-bit hash: 0x265357902fe1b7e2a04b897c6025d7a2
32-bit truncated hash: 0x26535790
When the input to the hash algorithm is a public key value, as may be used by various security protocols, the hash SHOULD be calculated over the public key in an X.509 SubjectPublicKeyInfo structure (Section 4.1 of [RFC5280]). This input has been chosen primarily for compatibility with DANE [I-D.ietf-dane-protocol], but also includes any relevant public key parameters in the hash input, which is sometimes necessary for security reasons. Note also that this does not force use of X.509 or full compliance with [RFC5280] since formatting any public key as a SubjectPublicKeyInfo is relatively straightforward and well supported by libraries.
Any of the formats defined below can be used to represent the resulting name for a public key.
Other than in the above special case where public keys are used, we do not specify the hash function input here. Other specifications are expected to define this.
A Named Information (ni) URI consists of the following components:
It is OPTIONAL for implementations to check the integrity of the URI/resource mapping when sending, receiving or processing "ni" URIs.
The digest value MUST be encoded using base64url [RFC4648] encoding.
The query segment of a URI is NOT hierarchical. Thus escape encoding of slash '/' characters is NOT required. Since application code often attempts to enforce such encoding, decoders MUST recognize the use of URI escape encoding (e.g., '%2f' or '%2F' for the slash character). Section 3.4 of [RFC3986] states that "The characters slash ("/") and question mark ("?") may represent data within the query component."
Consequently no special escaping mechanism is required for the query parameter portion of ni URIs. URI escaping is however frequently imposed automatically by scripting environments. Thus to ensure interoperability, implementations SHOULD NOT generate URIs that employ URI character escaping, and implementations MUST NOT reject any URIs that employ URI character escaping.
The Named Information URI has the following syntax:
niname ="ni://" [ authority ] "/" algval [ "?" query ]
algval = alg ";" val
alg = 1*unreserved
val = 1*unreserved
unreserved = ALPHA / DIGIT / "-" / "." / "_" / "~"
Note that "unreserved" is defined in the URI specification [RFC3986] Section 2.3, in the way shown above. The "authority" and "query" types are also from the URI specification. [RFC3986]
The "val" field MUST contain the output of applying the hash function ("alg") to its defined input, which defaults to the object bytes that are expected to be returned when the URI is dereferenced.
We define a mapping between URIs following the ni URI scheme and HTTP or HTTPS URLs that makes use of the .well-known URI [RFC5785] by defining an "ni" suffix (see Section 9).
The HTTP(S) mapping MAY be used in any context where clients without support for ni URIs are needed without loss of interoperability or functionality.
For an ni name of the form "ni://n-authority/alg;val?query-string" the corresponding HTTP(S) URL produced by this mapping is "http://h-authority/.well-known/ni/alg/val?query-string", where "h-authority" is derived as follows: If the ni name has a specified authority (i.e., the n-authority is non-empty) then the h-authority MUST have the same value. If the ni name has no authority specified (i.e. the n-authority string is empty), a h-authority value MAY be derived from the application context. For example, if the mapping is being done in the context of a web page then the origin [RFC6454] for that web site can be used. Of course, there are in general no guarantees that the object named by the ni URI will be available at the corresponding HTTP(S) URL. But in the case that any data is returned, the retriever can determine whether or not it is content that matches the ni URI.
If an application is presented with a HTTP(S) URL with "/.well-known/ni/" as the start of its pathname component, then the reverse mapping to an ni URI either including or excluding the authority might produce an ni URI that is meaningful, but there is no guarantee that this will be the case.
When mapping from a ni URI to a .well-known URL, an implementation will have to decide between choosing an "http" or "https" URL. If the object referenced does in fact match the hash in the URL, then there is arguably no need for additional data integrity, if the ni URI or .well-known URL was received "securely." However TLS also provides confidentiality, so there may still be reasons to use the "https" URL scheme even in this case. In general however, whether to use "http" or "https" is something that needs to be decided by the application.
Some applications may benefit from using hashes in existing HTTP URLs or other URLs. To do this one simply uses the "algval" production from the ni name scheme ABNF which may be included in the pathname component of HTTP URLs. In such cases there is nothing present in the URL that ensures that a client can depend on compliance with this specification, so clients MUST NOT assume that any URL with a pathname component that matches the "algval" production was in fact produced as a result of this specification. That URL might or might not be related to this specification, only the context will tell.
When a more space-efficient version of the name is needed, we can use a binary format. The binary format name consists of two fields: a header and the hash value. The header field defines how the identifier has been created and the hash value contains a (possibly truncated) result of a one-way hash over whatever is being identified by the hash value. The format of the binary representation of a name is shown in Figure 3.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Res| Suite ID | Hash Value /
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
/ ... /
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
/ ... |
+-+-+-+-+-+-+-+-+
The Res field is a reserved 2-bit field for future use and MUST be set to zero for this specification.
The hash algorithm and truncation length are specified by the Suite ID. For maintaining efficient encoding for the binary presentation, only a few hash algorithms and truncation lengths are supported. See Section 9.4 for details.
Note that a hash value that is truncated to 120 bits will result in the overall name being a 128-bit value which may be useful with certain use-cases.
Sometimes the name may need to be used in a format that is easy for humans to read and possibly communicate, for example, over the phone. For this purpose, the following more verbose but less ambiguous (when spoken) URI format is defined with scheme name "nih", standing for "Named Information for Humans." (Or possibly "Not Invented Here," which is clearly false, and therefore worth including :-)
As with the ni URI format, nih URI fields are separated by a semi-colon (;) character. The first field is a hash algorithm string, as in the ni URI format. Then the hash value is encoded using ASCII hex characters and lower-case alphabet.
The hash value is OPTIONALLY followed by a checksum. The checksum MUST be calculated as a crc16 over the following parts (encoded as UTF-8 [RFC3629]): the URI scheme and separator ("nih:"), the algorithm string, the first delimiter, (";") the hash value, and the second delimiter (also ";"). The 16-bit result of the crc16 is encoded using network byte order and, like the hash value, with lower-case ASCII hex characters.
The crc16 MUST use the CRC-CCITT polynomial: x^16 + x^12 + x^5 + 1.
humanname = "nih:" algval [ ";" checksum ]
algval = alg ";" val
alg = 1*unreserved
val = 1*unreserved
checksum = 1*unreserved
For algorithms that have a Suite ID reserved (see Figure 7), the alg field MAY contain the ID value as a UTF-8 encoded decimal number instead of the hash name string (for example, "3" instead of "sha-256-120"). Implementations MUST be able to match the decimal ID values for the algorithms and hash lengths that they support even if they do not support the binary presentation. Note that using the decimal presentation instead of the hash name string results in a different checksum for the same name.
The following ni URI references the text "Hello World!" (without the quotes, being 12 characters), using the sha-256 algorithm and no authority field:
ni:///sha-256;A7ogTlDRJuRnTABeBNguhMITZngK8fQ71Uo3gWtqs0A
And the same example shown with an authority would be:
ni://example.com/sha-256;A7ogTlDRJuRnTABeBNguhMITZngK8fQ71Uo3gWtqs0A
The following HTTP URL represents a mapping from the previous ni name based on the algorithm outlined above.
http://example.com/.well-known/ni/sha-256/A7ogTlDRJuRnTABeBNguhMITZngK8fQ71Uo3gWtqs0A
Given the SubjectPublicKeyInfo in Figure 5 we derive the names shown in Figure 6 for this value.
0000000 82 01 0a 02 82 01 01 00 a2 5f 83 da 9b d9 f1 7a 0000020 3a 36 67 ba fd 5a 94 0e cf 16 d5 5a 55 3a 5e d4 0000040 03 b1 65 8e 6d cf a3 b7 db a4 e7 cc 0f 52 c6 7d 0000060 35 1d c4 68 c2 bd 7b 9d db e4 0a d7 10 cd f9 53 0000100 20 ee 0d d7 56 6e 5b 7a ae 2c 5f 83 0a 19 3c 72 0000120 58 96 d6 86 e8 0e e6 94 eb 5c f2 90 3e f3 a8 8a 0000140 88 56 b6 cd 36 38 76 22 97 b1 6b 3c 9c 07 f3 4f 0000160 97 08 a1 bc 29 38 9b 81 06 2b 74 60 38 7a 93 2f 0000200 39 be 12 34 09 6e 0b 57 10 b7 a3 7b f2 c6 ee d6 0000220 c1 e5 ec ae c5 9c 83 14 f4 6b 58 e2 de f2 ff c9 0000240 77 07 e3 f3 4c 97 cf 1a 28 9e 38 a1 b3 93 41 75 0000260 a1 a4 76 3f 4d 78 d7 44 d6 1a e3 ce e2 5d c5 78 0000300 4c b5 31 22 2e c7 4b 8c 6f 56 78 5c a1 c4 c0 1d 0000320 ca e5 b9 44 d7 e9 90 9c bc ee b0 a2 b1 dc da 6d 0000340 a0 0f f6 ad 1e 2c 12 a2 a7 66 60 3e 36 d4 91 41 0000360 c2 f2 e7 69 39 2c 9d d2 df b5 a3 44 95 48 7c 87 0000400 64 89 dd bf 05 01 ee dd 02 03 01 00 01 0000415 0000000 c4 f6 dd 00 8f 7e b5 a0 95 cf 09 c5 8d d7 95 ba 0000020 1d 06 6c 11 50 cf 2a e3 d3 07 9b fa af e6 2b 1a
+-------------------------------------------------------------------+ | URI: | | ni:///sha-256;xPbdAI9-taCVzwnFjdeVuh0GbBFQzyrj0web-q_mKxo | +-------------------------------------------------------------------+ | .well-known URL (split over 2 lines): | | http://example.com/.well-known/ni/sha256/ | | xPbdAI9-taCVzwnFjdeVuh0GbBFQzyrj0web-q_mKxo | +-------------------------------------------------------------------+ | URL Segment: | | sha-256;xPbdAI9-taCVzwnFjdeVuh0GbBFQzyrj0web-q_mKxo | +-------------------------------------------------------------------+ | Binary name (ASCII hex encoded) with 120-bit truncated hash value | | which is Suite ID 0x03: | | 03c4 f6dd 008f 7eb5 a095 cf09 c58d d795 | +-------------------------------------------------------------------+ | Human-readable form of a name for this key (truncated to 120 bits | | in length) with checksum: | | nih:sha-256-120;c4f6dd008f7eb5a095cf09c58dd795;8b04 | +-------------------------------------------------------------------+ | Human-readable form of a name for this key (truncated to 32 bits | | in length) with checksum: | | nih:sha-256-32;c4f6dd00;1fa8 | +-------------------------------------------------------------------+ | Human-readable form using decimal presentation of the | | algorithm ID (sha-256-120) with checksum: | | nih:3;c4f6dd008f7eb5a095cf09c58dd795;1819 | +-------------------------------------------------------------------+
The procedures for registration of a URI scheme are specified in RFC 4395 [RFC4395]. The following is the proposed assignment template.
URI scheme name: ni
Status: Permanent
URI scheme syntax. See Section 3
URI scheme semantics. See Section 3
Encoding considerations. See Section 3
Applications/protocols that use this URI scheme name: General applicability with initial use cases provided by CoAP and DECADE
Interoperability considerations: Defined here.
Security considerations: See Section 10
Contact: stephen.farrell@cs.tcd.ie
Author/Change controller: IETF
References: As specified in this document
The procedures for registration of a URI scheme are specified in RFC 4395 [RFC4395]. The following is the proposed assignment template.
URI scheme name: nih
Status: Permanent
URI scheme syntax. See Section 7
URI scheme semantics. See Section 7
Encoding considerations. See Section 7
Applications/protocols that use this URI scheme name: General applicability with initial use cases provided by CoAP and DECADE
Interoperability considerations: Defined here.
Security considerations: See Section 10
Contact: stephen.farrell@cs.tcd.ie
Author/Change controller: IETF
References: As specified in this document
The procedures for registration of a Well Known URI entry are specified in RFC 5785 [RFC5785]. The following is the proposed assignment template.
URI suffix: ni
Change controller: IETF
Specification document(s): This document
Related information: None
IANA is requested to create a new registry for hash algorithms as used in the name formats specified here. This registry has four fields, the binary suite ID, the hash algorithm name string, the truncation length and the underlying algorithm reference. Future assignments are to be made through expert review [RFC5226]. Initial values are specified below.
Since there are only 63 possible binary suite ID field values allowed by the binary format specified here, the suite ID field value is OPTIONAL. Where the binary format is not expected to be used for a given hash algorithm, this field SHOULD be omitted.
ID Hash name string Value length Reference 0 Reserved 1 sha-256 256 bits [RFC4055] 2 sha-256-128 128 bits [RFC4055] 3 sha-256-120 120 bits [RFC4055] 4 sha-256-96 96 bits [RFC4055] 5 sha-256-64 64 bits [RFC4055] 6 sha-256-32 32 bits [RFC4055] 32 Reserved
The Suite ID value 32 is reserved for compatibility with ORCHIDs [RFC4843]. The referenced hash algorithm matching to the Suite ID, truncated to the length indicated, according to the description given in Section 2, MUST be used for generating the hash.
No secret information is required to generate or verify a name of the form described here. Therefore a name like this can only provide evidence for the integrity for the referenced object and the proof of integrity provided is only as good as the proof of integrity for the name from which we started. In other words, the hash value can provide a name-data integrity binding between the name and the bytes returned when the name is de-referenced using some protocol.
Disclosure of a name value does not necessarily entail disclosure of the referenced object but may enable an attacker to determine the contents of the referenced object by reference to a search engine or other data repository or, for a highly formatted object with little variation, by simply guessing the value and checking if the digest value matches. So the fact that these names contain hashes does not protect the confidentiality of the object that was input to the hash.
The integrity of the referenced content would be compromised if a weak hash function were used. So don't use those. SHA-256 is currently our preferred hash algorithm which is why we've only added SHA-256 based suites to the initial IANA registry.
If a truncated hash value is used, certain security properties will be affected. In general a hash algorithm is designed to produce sufficient bits to prevent a 'birthday attack' collision occurring. To ensure that the difficulty of discovering two pieces of content that result in the same digest with a work factor O(2^x) by brute force requires a digest length of 2x. Many security applications only require protection against a 2nd pre-image attack which only requires a digest length of x to achieve the same work factor. Basically, the shorter the hash value used, the less security benefit you can possibly get.
This work has been supported by the EU FP7 project SAIL. The authors would like to thank SAIL participants to our naming discussions, especially Jean-Francois Peltier, for their input.
The authors would also like to thank Bob Moskowitz, Tero Kivinen, Zach Shelby, Carsten Bormann, David McGrew, Eric Rescorla, Tobias Heer, Martin Thomas and James Manger for their comments and input to the document.