Network Working Group S.E. Kille INTERNET-DRAFT Isode Ltd. Expires in six months January 1999 Intended Category: Standard X.509 Authentication SASL Mechanism 1. Status of this Memo This document is an Internet-Draft. Internet-Drafts are working docu- ments of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. 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." To learn the current status of any Internet-Draft, please check the "1id-abstracts.txt" listing contained in the Internet-Drafts Shadow Directories on ds.internic.net (US East Coast), nic.nordu.net (Europe), ftp.isi.edu (US West Coast), or munnari.oz.au (Pacific Rim). 2. Abstract This document defines a SASL [1] authentication mechanism based on X.509 strong authentication [3], providing two way authentication. This mechanism is only for authentication, and has no effect on the protocol encodings and is not designed to provide integrity or confidentiality services. 3. Model The mechanism provides two way strong authentication as defined in X.509. The encoding is based on that used by X.500 in the DAP, DSP, and DISP protocols. The mechanism is based on use of an asymmetric (public key) signing mechanism. This SASL mechanism contains two authentication mechanisms: - Client authentication is where the client provides information to the server, so that the server can authenticate the client. - Server authentication is where the server provides information to the client, so that the client can authenticate the server. Kille [Page 1] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT This mechanism is given three SASL keys for different variants: - "X509-C-" for client authentication only. - "X509-S-" for server authentication only. - "X509-B-" for client and server authentication. In this case client authentication is done prior to server authentication. Each SASL key may be used with a list of algorithms. A list of sup- ported algorithms is given in Section 7. For Client Authentication ("X509-C-"): 1. The client generates the credentials (SASLStrongCredentials) from information on both parties and a random number, and signs the enclosed token with its own private key. 2. The client sends credentials to the server. 3. The server verifies these credentials using the client's public key, and the authentication is complete. For Server Authentication ("X509-S-"): 1. The server generates the credentials (SASLStrongCredentials) from information on both parties and a random number, and signs the enclosed token with its own private key. 2. The server sends credentials to the client. 3. The client verifies these credentials using the server's public key, and the authentication is complete. For most SASL based protocols, server only authentication will not be useful. However, this is included here, as the definition is required for "client and server", and it may be useful for future protocols. For Client and Server Authentication ("X509-B-"), the procedure for "X509-C-" is performed and then followed by the procedure for "X509-S-". The Client needs to go first, as for some protocols the server will need to get information about the client from the client authentication in order to be able to perform the server authentication. 3.1. Encoding The SASLStrongCredentials, which is the definition of the data format exchanged, is encoded using ASN.1 Distinguished Encoding Rules (DER). Kille [Page 2] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT 4. Why this SASL Mechanism is Needed This section discusses the requirements for this SASL mechanism. 4.1. Benefits of a Public Key Mechanism The key benefit of asymmetric (public key) security, is that the secret (private key) only needs to be placed with the entity that is being authenticated. Thus a private key can be issued to a client, which can then be authenticated by ANY server based on a token generated by the client and the generally available public key. Symmetric authentication mechanisms (password mechanisms such as CRAM-MD5) require a shared secret, and the need to maintain it at both endpoints. This means that a secret key for the client needs to be maintained at every server which may need to authenticate the client. This is particularly an issue for protocols such as LDAP, where a client may connect to and be authenticated by a large number of servers. In this situation, the requirement to maintain secret keys on all possible servers is not practical, which makes authentication mechanisms such as CRAM-MD5 unsuitable for LDAP in many situations. 4.2. Why Authentication Only? This service provides authentication only. The primary reason for this is that it makes the mechanism very simple. It would be possible to define a more complex mechanism which exchanged session keys and also provided confidentiality and/or integrity. There are a number of places where an authentication only service is useful: - Where confidentiality and integrity are provided by lower layers (e.g., TLS or IPSec). - Where confidentiality or integrity services are provided by the application (e.g., X.500 signed operations). - Where physical and other security aspects of the environment do no require confidentiality and integrity services. - For legacy applications where changes to the data exchange would be difficult to integrate. 4.3. Relationship to TLS The functionality defined here can be provided by TLS, and it is impor- tant to consider why it is useful to have it in both places. There are Kille [Page 3] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT a number of reasons for this: - SASL. SASL also duplicates TLS functionality, and the rationale for this is clearly given in RFC 2222 [1]. These arguments apply here. - Simplicity. This mechanism is a great deal simpler than TLS. If there is only a requirement for this functionality (as distinct from all of TLS), this simplicity will facilitate deployment. - Layering. The SASL mechanism to establish authentication works cleanly with most protocols. TLS often layers awkwardly, and does not provide the authentication dialogue at the right stage in the protocol negotiation. - Proxy support. Proxys can be cleanly supported with this mechan- ism, but not with TLS. This works because the proxy can authenti- cate the client, and then simply pass the credentials on the server, using the previous token member. - No Data Confidentiality and integrity required. In many situa- tions, the data confidentiality and integrity provided by TLS is not needed, and in this situation use of TLS provides an unneces- sary overhead and complexity. - Export. Because of its data confidentiality functionality, TLS will often have export problems. When used with a signing algo- rithm such as DSS that cannot be used for data confidentiality, export problems for this mechanism will be much less. For this reason, it will be much easier to globally deploy this mechanism than TLS. It is also useful that if an "export product" uses "weak" data confidentiality, that a separate authentication mechan- ism will mean that authentication does not need to be weakened. 5. Token Definition The SASLStrongCredentials defined here are based on the StrongCreden- tials defined in X.511, making use of the SIGNED Macro and Certification Path definitions of X.509. Two optional fields have been added, the second of which makes use of GeneralName defined in X.509 [6]. The credentials definition is given here for clarity. The formal defini- tions of CertificationPath, AlgorithmIdentifier, and DistinguishedName are by reference to X.511. The formal definition of GeneralName is given in X.509. There are a number of names referenced in these definitions. There are two entities involved in the interaction: Kille [Page 4] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT 1. Signer. This is the entity that is creating the SASLStrongCreden- tials, and singing it with its private key. The signer is also referred to as the "subject", in line with PKIX terminology, as the signer is essentially proving its own identity. 2. Target. This is the entity which is expected to verify the signa- ture. The target's name is included in order to prevent replay attacks, and so it is only the target that can securely verify the SASLStrongCredentials (other entities can verify the signature, but they would not be able to detect replay attack). This enables the token to be used to build trust chains. SASLStrongCredentials ::= SET { certification-path [0] CertificationPath OPTIONAL, bind-token [1] SASLToken, name [2] DistinguishedName OPTIONAL} -- signer SASLToken ::= SIGNED { SEQUENCE { algorithm [0] AlgorithmIdentifier, name [1] DistinguishedName, -- target time [2] UTCTime, random [3] BITSTRING, response [4] BITSTRING OPTIONAL, target-name [20] GeneralName OPTIONAL, signer-name [21] GeneralName OPTIONAL, previous-token [22] SASLStrongCredentials OPTIONAL}} PreviousToken ::= SEQUENCE { token [0] SASLToken, protocol [1] Protocol OPTIONAL } Protocol ::= INTEGER GeneralName ::= CHOICE { otherName [0] OtherName, rfc822Name [1] IA5String, dNSName [2] IA5String, x400Address [3] ORAddress, directoryName [4] Name, edipartyname [5] EDIPartyName, uniformresouceidentifier [6] IA5String, iPAddress [7] OCTET STRING, registeredID [8] OBJECT IDENTIFIER } The elements of SASLStrongCredentials are as follows: certification-path: This provides a mechanism for exchange of certificates, which may Kille [Page 5] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT help the recipient to verify the credentials. If this is included, it must be consistent with the name in SASLStrongCredentials.name. This information is provided by the entity generating the token to facilitate verification. The entity verifying the token is not required to use this information. bind-token: This is the signed token, which is the core of the credentials. name:This is the distinguished name of the signer of the token. For client authentication, this will need to be included unless the information is carried in another protocol element of the exchange (which will typically not be the case). For server authentication, this will not normally be needed, as the client will have a priori knowledge of the server it is connecting to. The entity verifying the token shall ensure that this name is con- sistent with the certificate, as a part of the verification pro- cess. The signed token contains the following elements. algorithm: This is the signature algorithm used to sign the token. This is included for compatibility with X.509, and generally implies both an asymmetric algorithm and a hash algorithm. The value SHALL be consistent with the algorithm defined by the SASL mechanism (e.g., DSA-SHA1). name:This is the distinguished name of the target (which will verify the token). For client authentication, this will be the name of the server. This element is mandatory for compatibility with X.511. If a name form other than Distinguished Name is used, this element should contain a null distinguished name, and a name included in the signer-name parameter. time:This is the time that the token expires. random: This is a random number, which must be unique for the target over the valid life of the token. This is included to prevent replay attack. It is recommended that this number is at least as long as the block size of the hash algorithm used. response: This is used to carry a number derived from random if challenge Kille [Page 6] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT response of authentication is required. This shall be used in the client phase of X509-B- and in no other circumstances. In this case, the value used in this field in the client authentication shall be used in the SASLToken.random field of the server authenti- cation. items 5-19: There is a gap in the sequence numbering. Items 6-9 are used in X.500 DAP, but are not appropriate here. target-name: This is a very general definition of a name, taken from X.509(v3). This definition is being used by ongoing work on PKI. This enables authentication identifiers other than distinguished names to be used. This will be important when the target does not have a dis- tinguished name. This field is needed when access control is to be applied on the basis of a name different to the one doing the signing. The name of the entity doing the signing is in SASLStrongCredentials.name, and this is verified by the signing process. This field, verified by the signature, is an alternate name to be used for access con- trol purposes in the authentication and for ongoing purposes with the association established. In SASL terminology, this is the "authorization entity". Note that this description is for tutorial purposes only, and the formal definition is taken from X.511. signer-name This has the same function as name, but allows for encodings other than Distinguished Name. previous-token This gives a mechanism to include a previous token, which includes a SASLToken, and optionally a protocol if this is different from the current protocol. The mechanism might be used in a firewall, which does protocol relay. The initial token is generated by the client, which is then encapsulated in another token generated by the firewall. This enables a signed trust chain to be built. The "change of protocol" enables a server to use a different proto- col on behalf of its client (e.g., an ACAP server performing a directory lookup on behalf of the ACAP client). Protocols are represented as Integers, identified by the TCP Port Number. Kille [Page 7] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT 6. Distinguished Names The X.509 strong authentication mechanism makes use of distinguished names to identify the target. For some protocols, such as LDAP [2], this is natural. For protocols which make use of internet domain names to identify objects, the representation defined in RFC 2247 [5] MAY be used as an alternative to subject-name in the token. For an Internet Mailbox the local part must be encoded as a domain component. For exam- ple "J.Bloggs@widget.com" is represented as the distinguished name "DC=J.Bloggs,DC=widget,DC=com". 7. Supported Algorithms The following signature algorithms are recognized for use with this mechanism, and identified by a key. Each key would be combined to make three possible SASL mechanisms. For example for the DSA-SHA1 algorithm would give X509-C-DSA-SHA1, X509-S-DSA-SHA1, and X509-B-DSA-SHA1. All algorithm names are constrained to 13 characters, to keep within the total SASL limit of 20 characters. The following table gives a list of algorithm keys, noting the object identifier and the body which assigned the identifier. Key Object Id Body RSA-MD4 1.3.14.2.2.2 OIW RSA-MD5 1.3.14.2.2.3 OIW RSA-MD4-ENCR 1.3.14.2.2.4 OIW DSA-SHA 1.3.14.2.2.13 OIW DSA-SHA-COMM 1.3.14.2.2.20 OIW RSA-MD2 1.3.14.7.3.1 OIW ELGAMAL-MD2 1.3.14.7.3.2 OIW RSA-MD2-ENCR 1.2.840.113549.1.1.2 RSA RSA-MD5-ENCR 1.2.840.113549.1.1.4 RSA RSA-SHA1-ENC 1.2.840.113549.1.1.5 RSA MSP-SDNS 2.16.840.1.101.2.1.1.1 DMS MSP-MOSAIC 2.16.840.1.101.2.1.1.2 DMS DSA-SHA1 1.2.840.10040.4.3 ANSI Two special algorithm keys are defined: - IMPLICIT. This is used when the signer has a priori knowledge of the algorithm to use. The algorithms is then identified solely by the AlgrithmIdentifier Object Identifier in the token. - X-*. Any algorithm starting with "X-" is reserved for private extension. Kille [Page 8] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT Support of the DSS-SHA1 algorithm is recommended for use with this mechanism. 8. Example The following example shows use with IMAP4. The example is designed to illustrate the protocol interaction and does not provide valid encoding examples. S: * OK IMAP4 server ready C: AOO1 CAPABILITY S: * CAPABILITY IMAP4 AUTH=CRAM-MD5 AUTH=X509-C-DSA-SHA1 AUTH=X509-C-RSA-MD5 S: A001 OK done C: AOO2 AUTHENTICATE X509-C-DSS-SHA1 S: + C: AE31FF05.......== S: + 13c3FF44.......== S: AOO3 OK Welcome, authenticated user: CN=Joe Bloggs,O=Widget,C=GB Editor's Note: The ASN.1 values here are fake. A real example should be used (perhaps include ASN.1 value notation), when one can be generated from a prototype implementation. 9. Security Considerations These algorithms are designed to be used for authentication where the underlying transport service cannot guarantee confidentiality. These mechanisms do not prevent an authenticated association from being hijacked. 10. Acknowledgments Design ideas included in this document are based on those from ITU and ISO, and the IETF ASID Working Groups. Useful ideas were taken from a note "X.500 Strong Authentication Mechanisms for LDAPv3" by Mark Wahl. The contributions of individuals in these working groups, including Harald Alvestrand (Maxware), Alexis Bor (Directory Works), David Boyce (Isode), William Curtin (DISA), Bruce Greenblatt (RSA), Steve Hole (Esys), Tim Howes (Netscape), John Myers (Netscape), Chris Newman (Inno- soft), Frank Siebeblist (DASCOM), Erik Skovgaard (Geotrain), and Sean Turner (IECA) are gratefully acknowledged. 11. Bibliography [1] J. Meyers, "Simple Authentication and Security Layer", RFC 2222, October 1997. Kille [Page 9] Expires July 1999 IP X.509 SASL Authentication INTERNET DRAFT [2] M. Wahl, T. Howes, S. Kille, "Lightweight Directory Access Protocol (v3)", RFC 2252, December 1997. [3] ITU-T Recommendation X.509 (1997) | ISO/IEC 9594-8:1997, Information Technology - Open Systems Interconnection - The Directory: Authentica- tion Framework. [4] ITU-T Recommendation X.511 (1997) | ISO/IEC 9594-8:1997, Information Technology - Open Systems Interconnection - The Directory: Abstract Service Definition. [5] S. Kille, M.Wahl, A. Grimstad, R. Huber, S. Sataluri, "Using Domains in LDAP/X.500 Distinguished Names", RFC 2247, January 1998. 12. Author's Address Steve Kille Isode Ltd The Dome, The Square Richmond, Surrey, TW9 1DT, UK Phone: +44-181-332-9091 Email: S.Kille@isode.com Kille [Page 10]