Client Certificates in DANE TLSA Records

The Transport Layer Security (TLS) protocol optionally supports the authentication of clients using X.509 certificates. TLS Applications currently employing DANE authentication of servers using TLSA records may also desire to authenticate clients using the same mechanism, especially if the client identity is in the form of or can be represented by a DNS domain name. Some design patterns from the Internet of Things (IoT) make use of this form of authentication, where large networks of physical objects identified by DNS names may authenticate themselves using TLS to centralized device management and control platforms. In this document, the term TLS is used generically to describe both the TLS and DTLS (Datagram Transport Layer Security) protocols.

When specifying client identities (i.e. client domain names) in TLSA records, the owner name of the TLSA record has the following format:

The first label identifies the application service name. The remaining labels are composed of the client domain name. Encoding the application service name into the owner name allows the same client domain name to have different authentication credentials for different application services. There is no need to encode the transport label - the same name form is usable with both TLS and DTLS. The _service label could be a custom string for an application, but more commonly is expected to be a service name registered in the IANA Service Name Registry. The RDATA or data field portion of the TLSA record is formed exactly as specified in RFC 6698, and carries the same meaning.

The authentication model assumed in this document is the following: The client is assigned an identity corresponding to a DNS domain name. This domain name doesn't necessarily have any relation to its network layer addresses. Clients often have dynamic or unpredictable addresses, and may move around the network, so tying their identity to network addresses is not feasible or wise in the general case. The client generates (or has generated for it) a private and public key pair, and a certificate binding the name to its public key. This certificate has a corresponding TLSA record published in the DNS, which allows it to be authenticated directly via the DNS (using the DANE-TA or DANE-EE usage modes) or via a PKIX public CA system constraint (using the PKIX-TA or PKIX-EE usage modes).

The client certificate MUST have have the client's DNS name specified in the Subject Alternative Name extension's dNSName type. Or, if an application specific identity is preferred or needed, the SRV-ID (PKIX OtherName SRVName) MUST be used to specify the application service and the client's name, e.g. "_smtp-client.device1.example.com". See and for a discussion of application specific identifiers in X.509 certificates. The initial revision of this document talks mainly about dNSName identifiers, because SRV-ID has not seen much adoption in the Internet to date. However, with TLSA usage modes except for DANE-EE, if there is a need to isolate multiple application specific credentials from each other on the same client (i.e. with the same underlying base domain name), then SRV-ID would need to be employed.

The protocol described in the initial version of this document assumes either that client authentication is mandatory, or that where it is optional, clients can handle a Client Certificate Request message from the server without issues if they are not equipped with client certificates. Technically, the TLS protocol specification states that the client may respond with a Client Certificate message with no certificate, and that the server may at its discretion continue the handshake without client authentication. However in practice, problems may arise. There are deployed client software implementations that do not react gracefully when encountering a certificate request that they did not expect. More importantly, a server may want an explicit indication from the client that it has a DANE record, so as to avoid unnecessary DNS queries in-band with the TLS handshake for clients that don't support this. Hence, to address this issue generally, a client identity signaling solution will need to be devised, whereby the client indicates its DANE identity (i.e. its domain name identity and the fact that this identity has an associated TLSA record) to the server. Application specific protocol enhancements are one way to achieve this, e.g. a new SMTP command. A more general way would be to develop a new TLS extension to convey this information. [Another internet draft is currently being written to define such a TLS extension to convey DANE client identity.]

The following examples are provided in the textual presentation format of the TLSA record. An example TLSA record for the client "device1.example.com." and the application "smtp-client". This record specifies the SHA-256 hash of a PKIX CA certificate to authenticate the client's certificate.

An example TLSA record for the client "client2.example.com." and the application "localsvc". This record specifies the SHA-512 hash of the subject public key component of the client's certificate. The usage mode for this record is 3 (DANE-EE) and hence no PKIX validation for this certificate should be performed.

[Note: As the client identity signaling solution is developed, this section will undergo enhancements to use it. A future revision of this document will explicitly address the additional use case of raw public keys instead of X.509 certificates.] A TLS Client conforming to this specification MUST have a signed DNS TLSA record published corresponding to its DNS name and X.509 certificate. The client presents this certificate in the TLS handshake with the server. The presented client certificate MUST have have the client's DNS name specified either in the Subject Alternative Name extension's dNSName type, or the SRVName type. A TLS Server implementing this specification performs the following steps: Request a client certificate in the TLS handshake (the "Client Certificate Request" message). Extract the client identity from the Subject Alternative Name extension's dNSName or SRVName type in the client certificate. (If no client certificate is provided, then the server may terminate the connection, or at its discretion may continue the handshake without client authentication.) Construct the DNS query name for the corresponding TLSA record. For dNSName, the underscored application service label is prepended to the domain name, corresponding to the application in use. For SRVName, the DNS query name is identical to the content of the SRVName identifier. See for the proposed owner name format. Look up the TLSA record in the DNS. The response MUST be cryptographically validated using DNSSEC. The server could perform the DNSSEC validation itself. It could also be configured to trust responses obtained via a validating resolver to which it has a secure connection. Extract the RDATA of the TLSA record and match it to the presented client certificate according to the rules specified in the DANE TLS protocol. If successfully matched, the client is authenticated and the TLS session proceeds. If not, the session is terminated with a "bad_certificate" alert message. If there are multiple records in the TLSA record set, then the client is authenticated as long as at least one of the TLSA records matches. If the presented client certificate has multiple distinct reference identifier types (e.g. a dNSName, and an rfc822Name) then TLS servers configured to perform DANE authentication according to this specification should only examine and authenticate the dNSName or SRVName identity. If the certificate contains both dNSName and SRVName identities, SRVName should be preferred. See for a description of reference identifiers and matching rules. If the presented client certificate has multiple dNSName or SRVName identities, then the client MUST use an identity signalling mechanism to indicate the intended name to the server. Specific applications may be designed to require more detailed validation steps. For example, a server might want to verify the client's IP address is associated with the certificate in some manner, e.g. by confirming that a secure reverse DNS lookup of that address ties it back to the same domain name, or by requiring an iPAddress component to be included in the certificate. Such details are outside the scope of this document, and should be outlined in other documents specific to the applications that require this behavior. Servers may have their own whitelisting and authorization rules for which certificates they accept. For example a TLS server may be configured to only allow TLS sessions from clients with certificate identities within a specific domain or set of domains.

This specification can also support the use of raw public keys in TLS. This use case employs only usage mode 3 (DANE-EE) and a selector value of 1 (SPKI) in the DANE TLSA record, as described in . It requires the use of the new client identity signaling solution discussed previously.

Should this document also consider client identities in the form of e-mail addresses? The use case might be an SMTP client talking to an SMTP submission server. In that case, the email address of a user would most likely be conveyed in the certificate in a subject alt name rfc822Name type. The corresponding TLSA record would have to then have an owner name format similar to the OPENPGPKEY or SMIMEA records. This use case might be best left to the SMIMEA specification to consider.

This document benefited from discussions with the following people: Duane Wessels, Allison Mankin, Casey Deccio, and Warren Kumari.

This document includes no request to IANA.

This document makes a narrow update to RFC 6698 by defining the usage of the TLSA record for client TLS certificates. There are no security considerations for this document beyond those described in RFC 6698 and in the specifications for TLS and DTLS [RFC5246], [RFC6347].