]> USC/Information Sciences Institute

4676 Admiralty Way, Suite 1133 Marina del Rey CA 90292 USA +1 213 587-1057 zihu@usc.edu

USC/Information Sciences Institute

4676 Admiralty Way, Suite 1133 Marina del Rey CA 90292 USA +1 310 448-8323 liangzhu@usc.edu

USC/Information Sciences Institute

4676 Admiralty Way, Suite 1001 Marina del Rey CA 90292 USA +1 310 822-1511 johnh@isi.edu

Verisign Labs

12061 Bluemont Way Reston VA 20190 +1 703 948-3200 amankin@verisign.com

Verisign Labs

12061 Bluemont Way Reston VA 20190 +1 703 948-3200 dwessels@verisign.com

ICANN

paul.hoffman@icann.org

Internet template This document describes the use of TLS to provide privacy for DNS. Encryption provided by TLS eliminates opportunities for eavesdropping and on-path tampering with DNS queries in the network, such as discussed in . In addition, this document specifies two usage profiles for DNS-over-TLS and provides advice on performance considerations to minimize overhead from using TCP and TLS with DNS. This document focuses on securing stub-to-recursive traffic, as per the charter of the DPRIVE working group. It does not prevent future applications of the protocol to recursive-to-authoritative traffic. Note: this document was formerly named draft-ietf-dprive-start-tls-for-dns. Its name has been changed to better describe the mechanism now used. Please refer to working group archives under the former name for history and previous discussion. [RFC Editor: please remove this paragraph prior to publication]

Today, nearly all DNS queries , are sent unencrypted, which makes them vulnerable to eavesdropping by an attacker that has access to the network channel, reducing the privacy of the querier. Recent news reports have elevated these concerns, and recent IETF work has specified privacy considerations for DNS . Prior work has addressed some aspects of DNS security, but until recently there has been little work on privacy between a DNS client and server. DNS Security Extensions (DNSSEC), provide response integrity by defining mechanisms to cryptographically sign zones, allowing end-users (or their first-hop resolver) to verify replies are correct. By intention, DNSSEC does not protect request and response privacy. Traditionally, either privacy was not considered a requirement for DNS traffic, or it was assumed that network traffic was sufficiently private, however these perceptions are evolving due to recent events . Other work that has offered the potential to encrypt between DNS clients and servers includes DNSCurve , DNSCrypt , ConfidentialDNS and IPSECA . In addition to the present draft, the DPRIVE working group has recently adopted a DNS-over-DTLS proposal. This document describes using DNS-over-TLS on a well-known port and also offers advice on performance considerations to minimize overheads from using TCP and TLS with DNS. Initiation of DNS-over-TLS is very straightforward. By establishing a connection over a well-known port, clients and servers expect and agree to negotiate a TLS session to secure the channel. Deployment will be gradual. Not all servers will support DNS-over-TLS and the well-known port might be blocked by some firewalls. Clients will be expected to keep track of servers that support TLS and those that don't. Clients and servers will adhere to the TLS implementation recommendations and security considerations of or its successor. The protocol described here works for queries and responses between stub clients and recursive servers. It might work equally between recursive clients and authoritative servers, but this application of the protocol is out of scope for the DNS PRIVate Exchange (DPRIVE) Working Group per its current charter. This document describes two profiles in providing different levels of assurance of privacy: an opportunistic privacy profile and an out-of-band key-pinned privacy profile. It is expected that a future document based on will further describe additional privacy profiles for DNS over both TLS and DTLS. An earlier version of this document described a technique for upgrading a DNS-over-TCP connection to a DNS-over-TLS session with, essentially, "STARTTLS for DNS". To simplify the protocol, this document now only uses a well-known port to specify TLS use, omitting the upgrade approach. The upgrade approach no longer appears in this document, which now focuses exclusively on the use of a well-known port for DNS-over-TLS.

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.

A DNS server that supports DNS-over-TLS MUST listen for and accept TCP connections on port 853. By mutual agreement with its clients, the server MAY, instead, use a port other than 853 for DNS-over-TLS. DNS clients desiring privacy from DNS-over-TLS from a particular server MUST establish a TCP connection to port 853 on the server. By mutual agreement with its server, the client MAY, instead, use a port other than port 853 for DNS-over-TLS. Such an other port MUST NOT be port 53, but MAY be from the "first-come, first-served" port range. The first data exchange on this TCP connection MUST be the client and server initiating a TLS handshake using the procedure described in . DNS clients and servers MUST NOT use port 853 to transport clear text DNS messages. DNS clients MUST NOT send and DNS servers MUST NOT respond to clear text DNS messages on any port used for DNS-over-TLS (including, for example, after a failed TLS handshake). There are significant security issues in mixing protected and unprotected data and for this reason TCP connections on a port designated by a given server for DNS-over-TLS are reserved purely for encrypted communications. DNS clients SHOULD remember server IP addresses that don't support DNS-over-TLS, including timeouts, connection refusals, and TLS handshake failures, and not request DNS-over-TLS from them for a reasonable period (such as one hour per server). DNS clients following an out-of-band key-pinned privacy profile () MAY be more aggressive about retrying DNS-over-TLS connection failures.

Once the DNS client succeeds in connecting via TCP on the well-known port for DNS-over-TLS, it proceeds with the TLS handshake , following the best practices specified in or its successor. The client will then authenticate the server, if required. This document does not propose new ideas for authentication. Depending on the privacy profile in use , the DNS client may choose not to require authentication of the server, or it may make use of trusted a SPKI Fingerprint pinset. After TLS negotiation completes, the connection will be encrypted and is now protected from eavesdropping. At this point, normal DNS queries SHOULD take place.

All messages (requests and responses) in the established TLS session MUST use the two-octet length field described in Section 4.2.2 of . For reasons of efficiency, DNS clients and servers SHOULD pass the two-octet length field, and the message described by that length field, to the TCP layer at the same time (e.g., in a single "write" system call) to make it more likely that all the data will be transmitted in a single TCP segment (, Section 8). In order to minimize latency, clients SHOULD pipeline multiple queries over a TLS session. When a DNS client sends multiple queries to a server, it should not wait for an outstanding reply before sending the next query (, Section 6.2.1.1). Since pipelined responses can arrive out-of-order, clients MUST match responses to outstanding queries using the ID field, query name, type, and class. Failure by clients to properly match responses to outstanding queries can have serious consequences for interoperability (, Section 7).

For DNS clients that use library functions such as "getaddrinfo()" and "gethostbyname()", current implementations are known to open and close TCP connections each DNS call. To avoid excess TCP connections, each with a single query, clients SHOULD reuse a single TCP connection to the recursive resolver. Alternatively they may prefer to use UDP to a DNS-over-TLS enabled caching resolver on the same machine that then uses a system-wide TCP connection to the recursive resolver. In order to amortize TCP and TLS connection setup costs, clients and servers SHOULD NOT immediately close a connection after each response. Instead, clients and servers SHOULD reuse existing connections for subsequent queries as long as they have sufficient resources. In some cases, this means that clients and servers may need to keep idle connections open for some amount of time. Proper management of established and idle connections is important to the healthy operation of a DNS server. An implementor of DNS-over-TLS SHOULD follow best practices for DNS-over-TCP, as described in . Failure to do so may lead to resource exhaustion and denial-of-service. Whereas client and server implementations from the era are known to have poor TCP connection management, this document stipulates that successful negotiation of TLS indicates the willingness of both parties to keep idle DNS connections open, independent of timeouts or other recommendations for DNS-over-TCP without TLS. In other words, software implementing this protocol is assumed to support idle, persistent connections and be prepared to manage multiple, potentially long-lived TCP connections. This document does not make specific recommendations for timeout values on idle connections. Clients and servers should reuse and/or close connections depending on the level of available resources. Timeouts may be longer during periods of low activity and shorter during periods of high activity. Current work in this area may also assist DNS-over-TLS clients and servers in selecting useful timeout values . Clients and servers that keep idle connections open MUST be robust to termination of idle connection by either party. As with current DNS-over-TCP, DNS servers MAY close the connection at any time (perhaps due to resource constraints). As with current DNS-over-TCP, clients MUST handle abrupt closes and be prepared to reestablish connections and/or retry queries. When reestablishing a DNS-over-TCP connection that was terminated, as discussed in , TCP Fast Open is of benefit. Underlining the requirement for sending only encrypted DNS data on a DNS-over-TLS port (Section 3.2), when using TCP Fast Open the client and server MUST immediately initiate or resume a TLS handshake (clear text DNS MUST NOT be exchanged). DNS servers SHOULD enable fast TLS session resumption and this SHOULD be used when reestablishing connections. When closing a connection, DNS servers SHOULD use the TLS close-notify request to shift TCP TIME-WAIT state to the clients. Additional requirements and guidance for optimizing DNS-over-TCP are provided by , .

This protocol provides flexibility to accommodate several different use cases. This document defines two usage profiles: (1) opportunistic privacy, and (2) out-of-band key-pinned authentication that can be used to obtain stronger privacy guarantees if the client has a trusted relationship with a DNS server supporting TLS. Additional methods of authentication will be defined in a forthcoming draft .

For opportunistic privacy, analogous to SMTP opportunistic encryption one does not require privacy, but one desires privacy when possible. With opportunistic privacy, a client might learn of a TLS-enabled recursive DNS resolver from an untrusted source (such as DHCP while roaming), it might or might not validate the resolver. These choices maximize availability and performance, but they leave the client vulnerable to on-path attacks that remove privacy. Opportunistic privacy can be used by any current client, but it only provides guaranteed privacy when there are no on-path active attackers.

The out-of-band key-pinned privacy profile can be used in environments where an established trust relationship already exists between DNS clients and servers (e.g., stub-to-recursive in enterprise networks, actively-maintained contractual service relationships, or a client using a public DNS resolver). The result of this profile is that the client has strong guarantees about the privacy of its DNS data by connecting only to servers it can authenticate. In this profile, clients authenticate servers by matching a set of Subject Public Key Info (SPKI) Fingerprints in an analogous manner to that described in . With this out-of-band key-pinned privacy profile, client administrators SHOULD deploy a backup pin along with the primary pin, for the reasons explained in . A backup pin is especially helpful in the event of a key rollover, so that a server operator does not have to coordinate key transitions with all its clients simultaneously. After a change of keys on the server, an updated pinset SHOULD be distributed to all clients in some secure way in preparation for future key rollover. The mechanism for out-of-band pinset update is out of scope for this document. Such a client will only use DNS servers for which an SPKI Fingerprint pinset has been provided. The possession of trusted pre-deployed pinset allows the client to detect and prevent person-in-the-middle and downgrade attacks. However, a configured DNS server may be temporarily unavailable when configuring a network. For example, for clients on networks that require authentication through web-based login, such authentication may rely on DNS interception and spoofing. Techniques such as those used by DNSSEC-trigger MAY be used during network configuration, with the intent to transition to the designated DNS provider after authentication. The user MUST be alerted that the DNS is not private during such bootstrap. Upon successful TLS connection and handshake, the client computes the SPKI Fingerprints for the public keys found in the validated server's certificate chain (or in the raw public key, if the server provides that instead). If a computed fingerprint exactly matches one of the configured pins the client continues with the connection as normal. Otherwise, the client MUST treat the SPKI validation failure as a non-recoverable error. provides a detailed example of how this authentication could be performed in practice.

DNS-over-TLS incurs additional latency at session startup. It also requires additional state (memory) and increased processing (CPU). Latency: Compared to UDP, DNS-over-TCP requires an additional round-trip-time (RTT) of latency to establish a TCP connection. TCP Fast Open can eliminate that RTT when information exists from prior connections. The TLS handshake adds another two RTTs of latency. Clients and servers should support connection keepalive (reuse) and out-of-order processing to amortize connection setup costs. Fast TLS connection resumption further reduces the setup delay and avoids the DNS server keeping per-client session state. TLS False Start can also lead to a latency reduction in certain situations. State: The use of connection-oriented TCP requires keeping additional state at the server in both the kernel and application. The state requirements are of particular concern on servers with many clients, although memory-optimized TLS can add only modest state over TCP. Smaller timeout values will reduce the number of concurrent connections, and servers can preemptively close connections when resource limits are exceeded. Processing: Use of TLS encryption algorithms results in slightly higher CPU usage. Servers can choose to refuse new DNS-over-TLS clients if processing limits are exceeded. Number of connections: To minimize state on DNS servers and connection startup time, clients SHOULD minimize creation of new TCP connections. Use of a local DNS request aggregator (a particular type of forwarder) allows a single active DNS-over-TLS connection from any given client computer to its server. Additional guidance can be found in . A full performance evaluation is outside the scope of this specification. A more detailed analysis of the performance implications of DNS-over-TLS (and DNS-over-TCP) is discussed in and .

IANA is requested to add the following value to the "Service Name and Transport Protocol Port Number Registry" registry in the System Range. The registry for that range requires IETF Review or IESG Approval and such a review was requested using the Early Allocation process for the well-known TCP port in this document. We further recommend that IANA reserve the same port number over UDP for the proposed DNS-over-DTLS protocol . IANA responded to the early allocation request with the following TEMPORARY assignment:

The TEMPORARY assignment expires 2016-10-08. IANA is requested to make the assigmnent permanent upon publication of this document as an RFC.

[Note to RFC Editor: please do not remove this section prior to publication as it may be useful to future Foo-over-TLS efforts] Earlier versions of this document proposed an upgrade-based approach to establishing a TLS session. The client would signal its interest in TLS by setting a "TLS OK" bit in the EDNS0 flags field. A server would signal its acceptance by responding with the TLS OK bit set. Since we assume the client doesn't want to reveal (leak) any information prior to securing the channel, we proposed the use of a "dummy query" that clients could send for this purpose. The proposed query name was STARTTLS, query type TXT, and query class CH. The TLS OK signaling approach has both advantages and disadvantages. One important advantage is that clients and servers could negotiate TLS. If the server is too busy, or doesn't want to provide TLS service to a particular client, it can respond negatively to the TLS probe. An ancillary benefit is that servers could collect information on adoption of DNS-over-TLS (via the TLS OK bit in queries) before implementation and deployment. Another anticipated advantage is the expectation that DNS-over-TLS would work over port 53. That is, no need to "waste" another port and deploy new firewall rules on middleboxes. However, at the same time, there was uncertainty whether or not middleboxes would pass the TLS OK bit, given that the EDNS0 flags field has been unchanged for many years. Another disadvantage is that the TLS OK bit may make downgrade attacks easy and indistinguishable from broken middleboxes. From a performance standpoint, the upgrade-based approach had the disadvantage of requiring 1xRTT additional latency for the dummy query. Following this proposal, DNS-over-DTLS was proposed separately. DNS-over-DTLS claimed it could work over port 53, but only because a non-DTLS server interprets a DNS-over-DTLS query as a response. That is, the non-DTLS server observes the QR flag set to 1. While this technically works, it seems unfortunate and perhaps even undesirable. DNS over both TLS and DTLS can benefit from a single well-known port and avoid extra latency and mis-interpreted queries as responses.

[Note to RFC Editor: please remove this section and reference to RFC 6982 prior to publication.] This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in RFC 6982. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist. According to RFC 6982, "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".

The Unbound recursive name server software added support for DNS-over-TLS in version 1.4.14. The unbound.conf configuration file has the following configuration directives: ssl-port, ssl-service-key, ssl-service-pem, ssl-upstream. See https://unbound.net/documentation/unbound.conf.html.

Sinodun Internet Technologies has implemented DNS-over-TLS in the ldns library from NLnetLabs. This also gives DNS-over-TLS support to the drill DNS client program. Patches available at https://portal.sinodun.com/stash/projects/TDNS/repos/dns-over-tls_patches/browse.

The digit DNS client from USC/ISI supports DNS-over-TLS. Source code available at http://www.isi.edu/ant/software/tdns/index.html.

The getdns API implementation supports DNS-over-TLS. Source code available at https://getdnsapi.net.

Use of DNS-over-TLS is designed to address the privacy risks that arise out of the ability to eavesdrop on DNS messages. It does not address other security issues in DNS, and there are a number of residual risks that may affect its success at protecting privacy: There are known attacks on TLS, such as person-in-the-middle and protocol downgrade. These are general attacks on TLS and not specific to DNS-over-TLS; please refer to the TLS RFCs for discussion of these security issues. Clients and servers MUST adhere to the TLS implementation recommendations and security considerations of or its successor. DNS clients keeping track of servers known to support TLS enables clients to detect downgrade attacks. For servers with no connection history and no apparent support for TLS, depending on their Privacy Profile and privacy requirements, clients may choose to (a) try another server when available, (b) continue without TLS, or (c) refuse to forward the query. Middleboxes are present in some networks and have been known to interfere with normal DNS resolution. Use of a designated port for DNS-over-TLS should avoid such interference. In general, clients that attempt TLS and fail can either fall back on unencrypted DNS, or wait and retry later, depending on their Privacy Profile and privacy requirements. Any DNS protocol interactions performed in the clear can be modified by a person-in-the-middle attacker. For example, unencrypted queries and responses might take place over port 53 between a client and server. For this reason, clients MAY discard cached information about server capabilities advertised in clear text. This document does not itself specify ideas to resist known traffic analysis or side channel leaks. Even with encrypted messages, a well-positioned party may be able to glean certain details from an analysis of message timings and sizes. Clients and servers may consider the use of a padding method to address privacy leakage due to message sizes

The below individuals contributed significantly to the draft. The RFC Editor prefers a maximum of 5 names on the front page, and so we have listed additional authors in this section.

The authors would like to thank Stephane Bortzmeyer, John Dickinson, Brian Haberman, Christian Huitema, Shumon Huque, Kim-Minh Kaplan, Simon Joseffson, Simon Kelley, Warren Kumari, John Levine, Ilari Liusvaara, Bill Manning, George Michaelson, Eric Osterweil, Jinmei Tatuya, Tim Wicinski, and Glen Wiley for reviewing this Internet-draft. They also thank Nikita Somaiya for early work on this idea. Work by Zi Hu, Liang Zhu, and John Heidemann on this document is partially sponsored by the U.S. Dept. of Homeland Security (DHS) Science and Technology Directorate, HSARPA, Cyber Security Division, BAA 11-01-RIKA and Air Force Research Laboratory, Information Directorate under agreement number FA8750-12-2-0344, and contract number D08PC75599.

&RFC2119; &RFC1034; &RFC1035; &RFC5246; &RFC5077; &RFC6335; &RFC7120; &RFC7469; &RFC7525; &RFC2818; &RFC3234; &RFC4033; &RFC5280; &RFC5966; &RFC6698; &RFC7258; &RFC7413; &RFC7435; &RFC7626; VeriSign, Inc

VeriSign, Inc

Verisign, Inc

VeriSign, Inc

SUNY Buffalo

NLnet Labs

OpenDNS, INC.

Sinodun Internet Technologies

ACLU

Cisco Systems, Inc.

NLnet Labs

Red Hat

Dyn Inc.

Sinodun

ISC

Google Switzerland GmbH

Google Inc.

Cisco

Cisco

Cisco

This section presents an example of how the out-of-band key-pinned privacy profile could work in practice based on a minimal pinset (two pins). Operators of a DNS-over-TLS service in this profile are expected to provide pins that are specific to the service being pinned (i.e., public keys belonging directly to the end-entity or to a service-specific private CA) and not to public key(s) of a generic public CA. A DNS client system is configured with an out-of-band key-pinned privacy profile from a network service, using a pinset containing two pins. Represented in HPKP style, the pins are: pin-sha256="FHkyLhvI0n70E47cJlRTamTrnYVcsYdjUGbr79CfAVI=" pin-sha256="dFSY3wdPU8L0u/8qECuz5wtlSgnorYV2f66L6GNQg6w=" The client also configures the IP addresses of its expected DNS server, 192.0.2.3 and 192.0.2.4. The client connects to 192.0.2.3 on TCP port 853 and begins the TLS handshake, negotiation TLS 1.2 with a diffie-hellman key exchange. The server sends a Certificate message with a list of three certificates (A, B, and C), and signs the ServerKeyExchange message correctly with the public key found certificate A. The client now takes the SHA-256 digest of the SPKI in cert A, and compares it against both pins in the pinset. If either pin matches, the verification is successful; the client continues with the TLS connection and can make its first DNS query. If neither pin matches the SPKI of cert A, the client verifies that cert A is actually issued by cert B. If it is, it takes the SHA-256 digest of the SPKI in cert B and compares it against both pins in the pinset. If either pin matches, the verification is successful. Otherwise, it verifes that B was issued by C, and then compares the pins against the digest of C's SPKI. If none of the SPKIs in the cryptographically-valid chain of certs match any pin in the pinset, the client closes the connection with an error, and marks the IP address as failed.