| Network Working Group | P. Saint-Andre |
| Internet-Draft | Filament |
| Obsoletes: 7613 (if approved) | A. Melnikov |
| Intended status: Standards Track | Isode Ltd |
| Expires: October 7, 2016 | April 5, 2016 |
Preparation, Enforcement, and Comparison of Internationalized Strings Representing Usernames and Passwords
draft-ietf-precis-7613bis-00
This document describes updated methods for handling Unicode strings representing usernames and passwords. The previous approach was known as SASLprep (RFC 4013) and was based on stringprep (RFC 3454). The methods specified in this document provide a more sustainable approach to the handling of internationalized usernames and passwords. The preparation, enforcement, and comparison of internationalized strings (PRECIS) framework, RFC 7564, obsoletes RFC 3454, and this document obsoletes RFC 7613.
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 7, 2016.
Copyright (c) 2016 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. 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.
Usernames and passwords are widely used for authentication and authorization on the Internet, either directly when provided in plaintext (as in the PLAIN Simple Authentication and Security Layer (SASL) mechanism [RFC4616] and the HTTP Basic scheme [HTTP-BASIC-AUTH]) or indirectly when provided as the input to a cryptographic algorithm such as a hash function (as in the Salted Challenge Response Authentication Mechanism (SCRAM) SASL mechanism [RFC5802] and the HTTP Digest scheme [HTTP-DIGEST-AUTH]).
To increase the likelihood that the input and comparison of usernames and passwords will work in ways that make sense for typical users throughout the world, this document defines rules for preparing, enforcing, and comparing internationalized strings that represent usernames and passwords. Such strings consist of characters from the Unicode character set [Unicode], with special attention to characters outside the ASCII range [RFC20]. The rules for handling such strings are specified through profiles of the string classes defined in the preparation, enforcement, and comparison of internationalized strings (PRECIS) framework specification [RFC7564].
Profiles of the PRECIS framework enable software to handle Unicode characters outside the ASCII range in an automated way, so that such characters are treated carefully and consistently in application protocols. In large measure, these profiles are designed to protect application developers from the potentially negative consequences of supporting the full range of Unicode characters. For instance, in almost all application protocols it would be dangerous to treat the Unicode character SUPERSCRIPT ONE (U+00B9) as equivalent to DIGIT ONE (U+0031), because that would result in false positives during comparison, authentication, and authorization (e.g., an attacker could easy spoof an account "user1@example.com").
Whereas a naive use of Unicode would make such attacks trivially easy, the PRECIS profile defined here for usernames generally protects applications from inadvertently causing such problems. (Similar considerations apply to passwords, although here it is desirable to support a wider range of characters so as to maximize entropy for purposes of authentication.)
The methods defined here might be applicable wherever usernames or passwords are used. However, the methods are not intended for use in preparing strings that are not usernames (e.g., Lightweight Directory Access Protocol (LDAP) distinguished names), nor in cases where identifiers or secrets are not strings (e.g., keys and certificates) or require specialized handling.
This document obsoletes RFC 4013 (the SASLprep profile of stringprep [RFC3454]) but can be used by technologies other than SASL [RFC4422], such as HTTP authentication as specified in [HTTP-BASIC-AUTH] and [HTTP-DIGEST-AUTH].
This document does not modify the handling of internationalized strings in usernames and passwords as prescribed by existing application protocols that use SASLprep. If the community that uses such an application protocol wishes to modernize its handling of internationalized strings to use PRECIS instead of stringprep, it needs to explicitly update the existing application protocol definition (one example is [XMPP-ADDR], which is intended to obsolete [RFC6122]). Non-coordinated updates to protocol implementations are discouraged because they can have a negative impact on interoperability and security.
Many important terms used in this document are defined in [RFC5890], [RFC6365], [RFC7564], and [Unicode]. The term "non-ASCII space" refers to any Unicode code point having a Unicode general category of "Zs", with the exception of U+0020 (here called "ASCII space").
As used here, the term "password" is not literally limited to a word; i.e., a password could be a passphrase consisting of more than one word, perhaps separated by spaces, punctuation, or other non-alphanumeric characters.
Some SASL mechanisms (e.g., CRAM-MD5, DIGEST-MD5, and SCRAM) specify that the authentication identity used in the context of such mechanisms is a "simple user name" (see Section 2 of [RFC4422] as well as [RFC4013]). Various application technologies also assume that the identity of a user or account takes the form of a username (e.g., authentication for the Hypertext Transfer Protocol as specified in [HTTP-BASIC-AUTH] and [HTTP-DIGEST-AUTH]), whether or not they use SASL. Note well that the exact form of a username in any particular SASL mechanism or application technology is a matter for implementation and deployment, and that a username does not necessarily map to any particular application identifier (such as the localpart of an email address).
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 [RFC2119].
This document specifies that a username is a string of Unicode code points [Unicode], encoded using UTF-8 [RFC3629], and structured as an ordered sequence of "userparts". A userpart is allowed to contain only code points that are in turn allowed by the PRECIS IdentifierClass defined in Section 4.2 of [RFC7564], and thus consists almost exclusively of letters and digits. A username can consist of a single userpart or a space-separated sequence of userparts.
The syntax for a username is defined as follows, using the Augmented Backus-Naur Form (ABNF) [RFC5234].
username = userpart *(1*SP userpart)
userpart = 1*(idbyte)
;
; an "idbyte" is a byte used to represent a
; UTF-8 encoded Unicode code point that can be
; contained in a string that conforms to the
; PRECIS IdentifierClass
;
All code points and blocks not explicitly allowed in the PRECIS IdentifierClass are disallowed; this includes private use characters, surrogate code points, and the other code points and blocks that were defined as "Prohibited Output" in [RFC4013]. In addition, common constructions such as "user@example.com" (e.g., the Network Access Identifier from [RFC7542]) are allowed as usernames under this specification, as they were under [RFC4013].
A username MUST NOT be zero bytes in length. This rule is to be enforced after any normalization and mapping of code points.
In protocols that provide usernames as input to a cryptographic algorithm such as a hash function, the client will need to perform enforcement of the rules for the UsernameCaseMapped or UsernameCasePreserved profile before applying the algorithm.
This specification defines two profiles for usernames: one that performs case mapping and one that performs case preservation (see further discussion under Section 3.4).
The definition of the UsernameCaseMapped profile of the IdentifierClass is provided in the following sections, including detailed information about preparation, enforcement, and comparison (for details on the distinction between these actions, refer to [RFC7564]).
An entity that prepares a string according to this profile MUST first map fullwidth and halfwidth characters to their decomposition mappings (see Unicode Standard Annex #11 [UAX11]). This is necessary because the PRECIS "HasCompat" category specified in Section 9.17 of [RFC7564] would otherwise forbid fullwidth and halfwidth characters. After applying this width-mapping rule, the entity then MUST ensure that the string consists only of Unicode code points that conform to the PRECIS IdentifierClass defined in Section 4.2 of [RFC7564]. In addition, the entity then MUST encode the string as UTF-8 [RFC3629].
An entity that performs enforcement according to this profile MUST prepare a string as described in Section 3.2.1 and MUST also apply the rules specified below for the UsernameCaseMapped profile (these rules MUST be applied in the order shown):
An entity that performs comparison of two strings according to this profile MUST prepare each string as specified in Section 3.2.1 and then enforce the rules specified in Section 3.2.2. The two strings are to be considered equivalent if they are an exact octet-for-octet match (sometimes called "bit-string identity").
The definition of the UsernameCasePreserved profile of the IdentifierClass is provided in the following sections, including detailed information about preparation, enforcement, and comparison (for details on the distinction between these actions, refer to [RFC7564]).
An entity that prepares a string according to this profile MUST first map fullwidth and halfwidth characters to their decomposition mappings (see Unicode Standard Annex #11 [UAX11]). This is necessary because the PRECIS "HasCompat" category specified in Section 9.17 of [RFC7564] would otherwise forbid fullwidth and halfwidth characters. After applying this width-mapping rule, the entity then MUST ensure that the string consists only of Unicode code points that conform to the PRECIS IdentifierClass defined in Section 4.2 of [RFC7564]. In addition, the entity then MUST encode the string as UTF-8 [RFC3629].
An entity that performs enforcement according to this profile MUST prepare a string as described in Section 3.3.1 and MUST also apply the rules specified below for the UsernameCasePreserved profile (these rules MUST be applied in the order shown):
An entity that performs comparison of two strings according to this profile MUST prepare each string as specified in Section 3.3.1 and then enforce the rules specified in Section 3.3.2. The two strings are to be considered equivalent if they are an exact octet-for-octet match (sometimes called "bit-string identity").
In order to accommodate the widest range of username constructs in applications, this document defines two username profiles: UsernameCaseMapped and UsernameCasePreserved. These two profiles differ only in the Case-Mapping Rule and are otherwise identical.
Case mapping is a matter for the application protocol, protocol implementation, or end deployment. In general, this document suggests that it is preferable to apply the UsernameCaseMapped profile and therefore perform case mapping, because not doing so can lead to false positives during authentication and authorization (as described in [RFC6943]) and can result in confusion among end users, given the prevalence of case mapping in many existing protocols and applications. However, there can be good reasons to apply the UsernameCasePreserved profile and thus not perform case mapping, such as backward compatibility with deployed infrastructure.
In particular:
If the specification for a SASL mechanism, SASL application protocol, or non-SASL application protocol uses the UsernameCaseMapped profile, it MUST clearly describe whether case mapping is to be applied at the level of the protocol itself, implementations thereof, or service deployments (each of these approaches can be legitimate, depending on the application in question).
Both the UsernameCaseMapped and UsernameCasePreserved profiles enable an application protocol, implementation, or deployment to create application-layer constructs such as a username that is a space-separated set of userparts like "Firstname Middlename Lastname". Although such a construct is not a profile of the PRECIS IdentifierClass (because U+0020 SPACE is not allowed in the IdentifierClass), it can be created at the application layer because U+0020 SPACE can be used as a separator between instances of the PRECIS IdentifierClass (e.g., userparts as defined in this specification).
The following examples illustrate a small number of userparts (not usernames) that are consistent with the format defined above (note that the characters "<" and ">" are used here to delineate the actual userparts and are not part of the userpart strings).
+--------------------------+---------------------------------+
| # | Userpart | Notes |
+--------------------------+---------------------------------+
| 1 | <juliet@example.com> | The at-sign is allowed in the |
| | | PRECIS IdentifierClass |
+--------------------------+---------------------------------+
| 2 | <fussball> | |
+--------------------------+---------------------------------+
| 3 | <fußball> | The third character is LATIN |
| | | SMALL LETTER SHARP S (U+00DF) |
+--------------------------+---------------------------------+
| 4 | <π> | A userpart of GREEK SMALL |
| | | LETTER PI (U+03C0) |
+--------------------------+---------------------------------+
| 5 | <Σ> | A userpart of GREEK CAPITAL |
| | | LETTER SIGMA (U+03A3) |
+--------------------------+---------------------------------+
| 6 | <σ> | A userpart of GREEK SMALL |
| | | LETTER SIGMA (U+03C3) |
+--------------------------+---------------------------------+
| 7 | <ς> | A userpart of GREEK SMALL |
| | | LETTER FINAL SIGMA (U+03C2) |
+--------------------------+---------------------------------+
Table 1: A Sample of Legal Userparts
Several points are worth noting. Regarding examples 2 and 3: although in German the character eszett (LATIN SMALL LETTER SHARP S (U+00DF)) can mostly be used interchangeably with the two characters "ss", the userparts in these examples are different and (if desired) a server would need to enforce a registration policy that disallows one of them if the other is registered. Regarding examples 5, 6, and 7: optional case-mapping of GREEK CAPITAL LETTER SIGMA (U+03A3) to lowercase (i.e., to GREEK SMALL LETTER SIGMA (U+03C3)) during comparison would result in matching the userparts in examples 5 and 6; however, because the PRECIS mapping rules do not account for the special status of GREEK SMALL LETTER FINAL SIGMA (U+03C2), the userparts in examples 5 and 7 or examples 6 and 7 would not be matched during comparison.
The following examples illustrate strings that are not valid userparts (not usernames) because they violate the format defined above.
+--------------------------+---------------------------------+
| # | Non-Userpart String | Notes |
+--------------------------+---------------------------------+
| 8 | <foo bar> | Space (U+0020) is disallowed in |
| | | the userpart |
+--------------------------+---------------------------------+
| 9 | <> | Zero-length userpart |
+--------------------------+---------------------------------+
| 10| <henryⅣ> | The sixth character is ROMAN |
| | | NUMERAL FOUR (U+2163) |
+--------------------------+---------------------------------+
| 11| <♚> | A localpart of BLACK CHESS KING |
| | | (U+265A) |
+--------------------------+---------------------------------+
Table 2: A Sample of Strings That Violate the Userpart Rule
Here again, several points are worth noting. Regarding example 8: although this is not a valid userpart, it is a valid username because it is a space-separated sequence of userparts. Regarding example 10: the Unicode character ROMAN NUMERAL FOUR (U+2163) has a compatibility equivalent of the string formed of LATIN CAPITAL LETTER I (U+0049) and LATIN CAPITAL LETTER V (U+0056), but characters with compatibility equivalents are not allowed in the PRECIS IdentifierClass. Regarding example 11: symbol characters such as BLACK CHESS KING (U+265A) are not allowed in the PRECIS IdentifierClass.
This document specifies that a password is a string of Unicode code points [Unicode], encoded using UTF-8 [RFC3629], and conformant to the OpaqueString profile (specified below) of the PRECIS FreeformClass defined in Section 4.3 of [RFC7564].
The syntax for a password is defined as follows, using the Augmented Backus-Naur Form (ABNF) [RFC5234].
password = 1*(freebyte)
;
; a "freebyte" is a byte used to represent a
; UTF-8 encoded Unicode code point that can be
; contained in a string that conforms to the
; PRECIS FreeformClass
;
All code points and blocks not explicitly allowed in the PRECIS FreeformClass are disallowed; this includes private use characters, surrogate code points, and the other code points and blocks defined as "Prohibited Output" in Section 2.3 of RFC 4013 (when corrected per [Err1812]).
A password MUST NOT be zero bytes in length. This rule is to be enforced after any normalization and mapping of code points.
In protocols that provide passwords as input to a cryptographic algorithm such as a hash function, the client will need to perform enforcement of the rules for the OpaqueString profile before applying the algorithm, because the password is not available to the server in plaintext form.
The definition of the OpaqueString profile is provided in the following sections, including detailed information about preparation, enforcement, and comparison (for details on the distinction between these actions, refer to [RFC7564]).
An entity that prepares a string according to this profile MUST ensure that the string consists only of Unicode code points that conform to the FreeformClass base string class defined in [RFC7564]. In addition, the entity MUST encode the string as UTF-8 [RFC3629].
An entity that performs enforcement according to this profile MUST prepare a string as described in Section 4.2.1 and MUST also apply the rules specified below for the OpaqueString profile (these rules MUST be applied in the order shown):
An entity that performs comparison of two strings according to this profile MUST prepare each string as specified in Section 4.2.1 and then enforce the rules specified in Section 4.2.2. The two strings are to be considered equivalent if they are an exact octet-for-octet match (sometimes called "bit-string identity").
The following examples illustrate a small number of passwords that are consistent with the format defined above (note that the characters "<" and ">" are used here to delineate the actual passwords and are not part of the password strings).
+------------------------------------+------------------------------+
| # | Password | Notes |
+------------------------------------+------------------------------+
| 12| <correct horse battery staple> | ASCII space is allowed |
+------------------------------------+------------------------------+
| 13| <Correct Horse Battery Staple> | Differs by case from |
| | | example 12 |
+------------------------------------+------------------------------+
| 14| <πßå> | Non-ASCII letters are OK |
| | | (e.g., GREEK SMALL LETTER |
| | | PI (U+03C0)) |
+------------------------------------+------------------------------+
| 15| <Jack of ♦s> | Symbols are OK (e.g., BLACK |
| | | DIAMOND SUIT (U+2666)) |
+------------------------------------+------------------------------+
| 16| <foo bar> | OGHAM SPACE MARK (U+1680) is |
| | | mapped to U+0020, and thus |
| | | the full string is mapped to |
| | | <foo bar> |
+------------------------------------+------------------------------+
Table 3: A Sample of Legal Passwords
The following example illustrates a string that is not a valid password because it violates the format defined above.
+------------------------------------+------------------------------+
| # | Password | Notes |
+------------------------------------+------------------------------+
| 17| <my cat is a 	by> | Controls are disallowed |
+------------------------------------+------------------------------+
Table 4: A String That Violates the Password Rules
This specification defines only the PRECIS-based rules for the handling of strings conforming to the UsernameCaseMapped and UsernameCasePreserved profiles of the PRECIS IdentifierClass, and strings conforming to the OpaqueString profile of the PRECIS FreeformClass. It is the responsibility of an application protocol to specify the protocol slots in which such strings can appear, the entities that are expected to enforce the rules governing such strings, and at what points during protocol processing or interface handling the rules need to be enforced. See Section 6 of [RFC7564] for guidelines on using PRECIS profiles in applications.
Above and beyond the PRECIS-based rules specified here, application protocols can also define application-specific rules governing such strings (rules regarding minimum or maximum length, further restrictions on allowable characters or character ranges, safeguards to mitigate the effects of visually similar characters, etc.), application-layer constructs (see Section 3.5), and related matters.
Some PRECIS profile definitions encourage entities that enforce the rules to be liberal in what they accept. However, for usernames and passwords such a policy can be problematic, because it can lead to false positives. An in-depth discussion can be found in [RFC6943].
The rules defined in this specification differ slightly from those defined by the SASLprep specification [RFC4013]. The following sections describe these differences, along with their implications for migration, in more detail.
Deployments that currently use SASLprep for handling usernames might need to scrub existing data when they migrate to the rules defined in this specification. In particular:
Under SASLprep, the use of NFKC also handled the mapping of fullwidth and halfwidth code points to their decomposition mappings.
Depending on local service policy, migration from RFC 4013 to this specification might not involve any scrubbing of data (because passwords might not be stored in the clear anyway); however, service providers need to be aware of possible issues that might arise during migration. In particular:
Under SASLprep, the use of NFKC also handled the mapping of fullwidth and halfwidth code points to their decomposition mappings. Although it is expected that code points with compatibility equivalents are rare in existing passwords, some passwords that matched when SASLprep was used might no longer work when the rules in this specification are applied.
IANA has made the updates described below.
IANA has added the following entry to the "PRECIS Profiles" registry.
IANA has added the following entry to the "PRECIS Profiles" registry.
IANA has added the following entry to the "PRECIS Profiles" registry.
The stringprep specification [RFC3454] did not provide for entries in the "Stringprep Profiles" registry to have any state except "Current" or "Not Current". Because this document obsoletes RFC 4013, which registered the SASLprep profile of stringprep, IANA has marked that profile as "Not Current" and cited this document as an additional reference.
The ability to include a wide range of characters in passwords and passphrases can increase the potential for creating a strong password with high entropy. However, in practice, the ability to include such characters ought to be weighed against the possible need to reproduce them on various devices using various input methods.
The process of comparing identifiers (such as SASL simple user names, authentication identifiers, and authorization identifiers) can lead to either false negatives or false positives, both of which have security implications. A more detailed discussion can be found in [RFC6943].
The security considerations described in [RFC7564] apply to the IdentifierClass and FreeformClass base string classes used in this document for usernames and passwords, respectively.
The security considerations described in [UTS39] apply to the use of Unicode characters in usernames and passwords.
This document builds upon the PRECIS framework defined in [RFC7564], which differs fundamentally from the stringprep technology [RFC3454] used in SASLprep [RFC4013]. The primary difference is that stringprep profiles allowed all characters except those characters that were explicitly disallowed, whereas PRECIS profiles disallow all characters except those characters that are explicitly allowed (this "inclusion model" was originally used for internationalized domain names in [RFC5891]; see [RFC5894] for further discussion). It is important to keep this distinction in mind when comparing the technology defined in this document to SASLprep [RFC4013].
The following substantive modifications were made from RFC 4013.
This document borrows some text from [RFC4013] and [RFC6120].
The following individuals provided helpful feedback on this document: Marc Blanchet, Ben Campbell, Alan DeKok, Joe Hildebrand, Jeffrey Hutzelman, Simon Josefsson, Jonathan Lennox, James Manger, Matt Miller, Chris Newman, Yutaka OIWA, Pete Resnick, Andrew Sullivan, Nico Williams, and Yoshiro YONEYA. Nico Williams in particular deserves special recognition for providing text that was used in Section 3.4. Thanks also to Takahiro NEMOTO and Yoshiro YONEYA for implementation feedback.
Robert Sparks and Derek Atkins reviewed the document on behalf of the General Area Review Team and the Security Directorate, respectively.
Benoit Claise and Stephen Farrell provided helpful input during IESG review.
Thanks to Matt Miller as document shepherd, Marc Blanchet and Yoshiro YONEYA as working group chairs, and Pete Resnick and Barry Leiba as area directors.
Peter Saint-Andre wishes to acknowledge Cisco Systems, Inc., for employing him during his work on earlier draft versions of this document.