| Internet-Draft | Authenticated ADoT | September 2021 |
| Dickson | Expires 23 March 2022 | [Page] |
This Internet Draft proposes a mechanism for DNS resolvers to discover support for TLS transport to authoritative DNS servers, to validate this indication of support, and to authenticate the TLS certificates involved.¶
This requires that the name server names are in a DNSSEC signed zone.¶
This also requires that the delegation of the zone served is protected by [I-D.dickson-dnsop-ds-hack], since the NS names are the keys used for discovery of TLS transport support.¶
Additional recommendations relate to use of various techniques for efficiency and scalability, and new EDNS options to minimize round trips and for signaling between clients and resolvers.¶
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 https://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 23 March 2022.¶
Copyright (c) 2021 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 (https://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.¶
The Domain Name System (DNS) predates any concerns over privacy, including the possibility of pervasive surveillance. The original transports for DNS were UDP and TCP, unencrypted. Additionally, DNS did not originally have any form of data integrity protection, including against active on-path attackers.¶
DNSSEC (DNS Security extensions) added data integrity protection, but did not address privacy concerns. The original DNS over TLS [RFC7858] and DNS over HTTPS [RFC8484] specifications were limited to client-to-resolver traffic.¶
The remaining privacy component is recursive-to-authoritative servers. This Internet Draft is designed to provide a solution to this problem.¶
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 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The result is that the parental side of the zone cut has records needed for DNS resolution which are not signed and not validatable.¶
Authoritative DNS over TLS is intended to provide the following for communications from recursive resolvers to authoritative servers:¶
The following are new protocol components, which are either included in this document, or are in other documents. Some are strictly required, while others are strongly suggested components to allow better scalability and performance. Some of the new elements are aliases to already documented standards, for purposes of these improvements.¶
| Element | New/Alias/OPT | Format/Base | Required | Description |
|---|---|---|---|---|
| DNST | Alias | SVCB | Spec: Y DNS: N | DNS Transport - support for DoT |
| TLSADOT | Alias | TLSA | Spec: Opt DNS: Yes or TLSA | TLSA without prefixing |
| ADOTD | New | OPT RR | N | Signal desire for ADOT (client-resolver) |
| ADOTA | New | OPT RR | N | Signal availablity of ADOT (resolver-client) |
| NSECD | New | OPT RR | N | Signal desire for NSEC(3) for [RFC8198] |
| NSV | New | DNSKEY Alg | Y | Protect NS - see [I-D.dickson-dnsop-ds-hack] |
This protocol depends on correct configuration and operation of the respective components, and that those are maintained according to Best Current Practices:¶
There are external dependencies that impact the system security of any DNSSEC zone, which are inherently unavoidable in establishing this scheme. Specifially, the original DS record enrollment and any updates to the DS records involved in DNSSEC delegations are presumed secure and outside of the scope of the DNS protocol per se.¶
Other risks relate to normal information security practices, including access controls, role based access, audits, multi-factor authentication, multi-party controls, etc. These are out of scope for this protocol itself.¶
ADoT is a property of DNS servers. The signaling is done at the server level, using a DNS record with the same owner name as the server itself (i.e. where the A and AAAA records for the server are published).¶
In order to support ADoT for a DNS server, it is necessary to publish a record specifyig explicit DoT support. This record also indicates other supported transports for the DNS server, e.g. the standard ports (TCP and UDP port 53).¶
The record type is "DNST" (DNS Transport), which is a specific instance of (aka binding for) SVCB with unique RRTYPE.¶
The zone serving the record MUST be DNSSEC signed. The absence of the DNST RRTYPE is proved by the NSEC(3) record, or else the DNST RRTYPE plus RRSIG is returned in response to a query for this record if it exists.¶
(NB: To be separated out into its own draft and expanded fully.) This SVCB binding will be given the RRTYPE value {TBD} with mnemonic name DNST ("DNS Transport"). Like any SVCB binding, there is a mandatory TargetName (which will normally be ".", indicating the target is the same as the record owner name). The default binding is the standard DNS ports, UDP/53 and TCP/53.¶
The SVCB binding includes support for an optional ADoT port, which is the standard DoT port TCP/853. This is signaled by "alpn=dot". The actual port number may be overridden with the "DOTport=N" SvcParam, and the UDP and TCP ports may also be overridden with optional "UDPport=N" and "TCPport=N" SvcParams.¶
Since DNST is a type-specific binding, it does NOT require the underscore prefix naming that the generic SVCB RRTYPE requires. It may occur anywhere, including at the apex of a DNS zone, and may co-exist with any other type that also permits other types.¶
This scheme uses an SVCB binding for DNS Transport, DNST. The binding has default ports UDP/53 and TCP/53 as the default-alpn. These are assumed unless "no-default-alpn" is added as an optional SvcParam.¶
The DNST binding uses the defined ALPN for DNS-over-TLS with the assigned label "dot". Use of ADoT is signaled if and only if the the SvcParam of "alpn=dot" is present.¶
Suppose the name server ns1.example.net supports only the normal DNS ports, and the name server ns2.example.net supports both the normal ports and ADoT. The zone example.net would include the records:¶
ns1.example.net. IN DNST 1 "."
ns2.example.net. IN DNST 1 "." alpn=dot
¶
And similarly, if another zone with many name server names wanted to have a policy of all-ADoT support (i.e. every name server supports ADoT), this could be encoded as:¶
*.example2.net DNST 1 "." alpn=dot¶
In each case, the first parameter is the SvcPriority, which must be non-zero (zero indicates AliasForm SVCB record type).¶
Note that it is possible for the resolver to use alter or ignore SvcPriority based on its own local policy. For instance, a resolver to prefer non-DoT over DoT or vice versa. Local policy might be to override SvcPriority ordering, and/or ignore some of the records. For example, a resolver might prefer to support mandatory use of DoT if present on any server in the NS RRset.¶
The presence of ADoT requires additionally that a TLSA [RFC6698] record be provided. A new RRTYPE is to be created for this as an alias of TLSA, with mnemonic of "TLSADOT" (TLS ADOT Certificate). This record will be published at the location NSNAME, where NSNAME is the name of the name server. Any valid TLSA RDATA is permitted. The use of Certificate Usage types PKIX-TA and PKIX-EE is NOT RECOMMENDED. The use of Certificate Usage types DANE-TA TLSA records may provide more flexibility in provisioning, including use of wild cards. Per [RFC7218][RFC7671] the RECOMMENDED Selector and Matching types for this are SPKI and SHA2-256, giving the recommended TLSA record type of DANE-TA SPKI SHA2-256, with the full encoded certificate.¶
Note that this TLSADOT record is "wildcard friendly". Use of aggressive synthesis by resolvers (per [RFC8198]) allows RRTYPE-specific wildcards to be used, avoiding repetitive entries where the RDATA is identical.¶
In the above example, ns2.example.net supports DNS over TLS, and thus would need to have a TLSA record. The zone would include:¶
ns2.example.net. IN TLSADOT DANE-TA SPKI SHA2-256 (signing cert)¶
If there were another zone containing many DNS server names, example2.net, it would be relatively simple to apply a wildcard record and use a signing cert (rather than end-entity cert) in the TLSADOT record). This would allow DNS caching to avoid repeated queries to the authoritative server for the zone containing the DNS server names, to obtain the TLSA-type information. This would look like the following:¶
*.example2.net IN TLSADOT DANE-TA SPKI SHA2-256 (signing cert)
ns1.example2.net IN A IP4_ADDRESS1
ns2.example2.net IN A IP4_ADDRESS2
ns3.example2.net IN A IP4_ADDRESS3
ns4.example2.net IN A IP4_ADDRESS4
ns1.example2.net IN AAAA IP6_ADDRESS1
ns2.example2.net IN AAAA IP6_ADDRESS2
ns3.example2.net IN AAAA IP6_ADDRESS3
ns4.example2.net IN AAAA IP6_ADDRESS4
¶
This transport signaling MUST only be trusted if the name server names for the domain containing the relevant name servers' names are protected with [I-D.dickson-dnsop-ds-hack] and validated. The name servers must also be in a DNSSEC signed zone (i.e. securely delegated where the delegation has been successfully DNSSEC validated).¶
The specific DNS transport that a name server supports is indicated via use of an RRSet of RRTYPE "DNST". This is a SVCB binding, and normally will use the TargetName of "." (meaning the same name). The default ALPN (transport mechanisms) are TCP/53 and UDP/53. The ADoT transport support is signaled by "alpn=dot". There is an existing entry for "dot" in the ALPN table, with port TCP/853.¶
Note that this RRTYPE is also "wildcard friendly". If a DNS zone containing the names of many servers with identical policy (related to ADoT support), those could be managed via one or more wildcard entries.¶
We re-use the same examples from above, indicating whether or not individual authoritative name servers support DoT:¶
ns1.example.net. IN DNST 1 "."
ns2.example.net. IN DNST 1 "." alpn=dot
¶
And similarly, if another zone with many name server names wanted to have a policy of all-ADoT support (i.e. every name server supports ADoT), this could be encoded as:¶
*.example2.net DNST 1 "." alpn=dot¶
A domain inherits the signaled transport for the name servers serving the domain.¶
This transport signaling MUST only be trusted for use of ADoT if the delegated name server names for the domain are protected with [I-D.dickson-dnsop-ds-hack].¶
The delegation to NS names "A" and "B", along with the DS record protecting/encoding "A" and "B", results in the DNS transport that is signaled for "A" and "B" being applied to the domain being delegated. This transport will include ADoT IFF the transport for "A" and "B" has included ADoT via DNS records.¶
No additional configuration is needed, beyond use of authority servers which signal DoT support. The following examples assumes the previous DNS records are provisioned:¶
example.com NS ns1.example.net. // does not support ADoT
example.com NS ns2.example.net. // supports ADoT
example2.com NS ns1.example2.net. // all support ADoT
example2.com NS ns2.example2.net. // all support ADoT
¶
In this example, ns1 does not have ADoT support (since the DNS record excludes the "alpn=dot" parameter), while ns2 does support ADoT (since it includes "alpn=dot").¶
These records are used to validate corresponding delegation records, DNS records, and TLSA records, as follows:¶
Suppose a client requests resolution for the IP address of "sensitive-name.example.com". Suppose the client's resolver has a "cold" cache without any entries beyond the standard Root Zone and relevant TLD name server records.¶
Suppose the following entries are present at their respective TLD authority servers, delegating to the respective authority servers:¶
// (Single NS for brevity only, please use 2 NS minimum )
// Unsigned delegations to various single-operator servers
example2.com NS ns1.example2.net. // all support ADoT
example3.com NS ns2.example2.net. // all support ADoT
example4.com NS ns3.example2.net. // all support ADoT
example5.com NS ns4.example2.net. // all support ADoT
// Zone serving NS data for single-operator's servers
example2.net NS ns1.infra2.example
example2.net NS ns2.infra2.example
example2.net DS (DS record data)
// glueless name servers are used
// Special zone serving NS data for previous zone
infra2.example NS ns1-glue.infra2.example
infra2.example NS ns2-glue.infra2.example
infra2.example DS (DS record data)
// Note use of glue for only this zone's delegation
ns1-glue.infra2.example A (glue A data)
ns1-glue.infra2.example AAAA (glue AAAA data)
ns2-glue.infra2.example A (glue A data)
ns2-glue.infra2.example AAAA (glue AAAA data)
¶
Suppose the following additional entries are in the respective authority servers for the ADOT signaling/certs:¶
example2.net SOA ( SOA record data )
// glueless name servers are used
example2.net NS ns1.infra2.example
example2.net NS ns2.infra2.example
//
// SVCB records (DNS Transport) for discovery of support
// wildcard used for efficiency and caching performance
*.example2.net DNST 1 "." alpn=dot
//
// ADOT TLSA signing cert
// wildcard used for efficiency and caching performance
*.example2.net IN TLSADOT DANE-TA SPKI SHA2-256 (signing cert)
//
// Addresses of name servers serving customer zones
// E.g. example2.com to example5.com served on these
ns1.example2.net IN A IP4_ADDRESS1
ns2.example2.net IN A IP4_ADDRESS2
ns3.example2.net IN A IP4_ADDRESS3
ns4.example2.net IN A IP4_ADDRESS4
ns1.example2.net IN AAAA IP6_ADDRESS1
ns2.example2.net IN AAAA IP6_ADDRESS2
ns3.example2.net IN AAAA IP6_ADDRESS3
ns4.example2.net IN AAAA IP6_ADDRESS4
//
// plus RRSIGs and NSEC(3) records and their RRSIGs
infra2.example SOA ( SOA record data )
infra2.example NS ns1-glue.infra2.example
infra2.example NS ns2-glue.infra2.example
ns1-glue.infra2.example A (same as glue A data)
ns1-glue.infra2.example AAAA (same as glue AAAA data)
ns2-glue.infra2.example A (same as glue A data)
ns2-glue.infra2.example AAAA (same as glue AAAA data)
//
// name server info for example2.net zone
ns1.infra2.example A (glueless A data)
ns1.infra2.example AAAA (glueless AAAA data)
ns2.infra2.example A (glueless A data)
ns2.infra2.example AAAA (glueless AAAA data)
//
// plus RRSIGs and NSEC(3) records and their RRSIGs
¶
The following are the necessary queries to various servers necessary to do a private TLS-protected lookup.¶
Several examples are provided in order, from a presumed cold cache state. Root Priming and TLD queries are presumed to already have been complete.¶
Query for sensitive-name.example2.com:¶
Query for sensitive-name.example3.com:¶
Query for sensitive-name.example4.com:¶
Query for sensitive-name.example5.com:¶
Query for sensitive-name2.example2.com:¶
Once the initial query or queries for a name server zone has been done, if that zone uses wildcards for DNST and TLSADOT, the only queries needed for a new name server are the A and/or AAAA records. Once the initial query for a name server has been done, all of the address and TLS information is available in the cache, and the DOT query can be made upon receipt of the TLD delegation record. Once the initial query for a second-level domain has been done, the TLD delegation and all of the address and TLS information is available in the cache, and the DOT query can be made immediately.¶
Once a cache is populated with wildcards from the name server domain, additional delegation queries require no more trips than those needed for normal UDP queries:¶
Once a cache is populated with name server addresses and wildcards, additional delegation queries require no more trips than those needed for normal UDP queries:¶
The following presume some new OPT sub-types, to be added to the IANA action section or to be split out as separate drafts. The sub-type mnemonics are "ADOTA" (available) and "ADOTD" (desired), each with an enumerated set of values and mnemonic codes. Respectively those are: "Always", "Upon Request", and "Never"; and "Force", "If Available", and "Never".¶
A DNS server (e.g. recursive resolver or forwarder) MAY signal to clients that it offers the use of ADoT. The mechanism used is to set the EDNS option "ADOTA". The values for this option are "Always", "Upon Request", and "Never". The value "Always" indicates the server will always attempt to use ADoT without regards to client requests for ADoT. The value "Upon Request" indicates that the server will ONLY use ADoT for upstream queries if the client requests that ADoT be used. These values have no effect on answers served from the resolver's cache. (The "Never" case is unusual, in that it signals the server understands the option, but does not perform ADoT. Generally this would be used to allow a client to track changes in the status, if the client is interested in uses of ADoT.)¶
A DNS client (e.g. stub or forwarder) MAY signal the desire to have the resolver use ADoT. The mechanism used is to set the EDNS option "ADOTD". The values for this option are "Force", "If Available", and "Never". The value "Force" indicates the server should attempt to use ADoT, and return a failure code of XXXX and an EDE value of YYYY if the authoritative server does not offer ADoT, or any other ADoT failure occurs. The value "If Available" indicates that the server should use ADoT for upstream queries if it is availble, but SHOULD NOT allow any downgrades if the authoritative server signals that ADoT is available. These values have no effect on answers served from the resolver's cache. (The "Never" case is unusual, in that it signals the client understands the option, but does not perform ADoT. Generally this would be used to allow a server to track changes in the client base, so the server operator can make informed decisions about enabling ADoT.)¶
As outlined above, there could be security issues in various use cases.¶
This document may or many not have any IANA actions. (e.g. if the RRTYPEs, EDNS subtypes, DNSKEY algorithms, etc., are defined in other documents, no IANA actions are needed.)¶
Thanks to everyone who helped create the tools that let everyone use Markdown to create Internet Drafts, and the RFC Editor for xml2rfc.¶
Thanks to Dan York for his Tutorial on using Markdown (specificially mmark) for writing IETF drafts.¶
Thanks to YOUR NAME HERE for contributions, reviews, etc.¶