DoT/DoDTLS Profiles: Usage Profiles for DNS over TLS and DNS over DTLS
Two usage profiles, Strict and Opportunistic, and the ways a DNS client can check it is really talking to its encrypted DNS server.
Dickinson, S., Gillmor, D., and T. Reddy, "Usage Profiles for DNS over TLS and DNS over DTLS", RFC 8310, DOI 10.17487/RFC8310, March 2018, <https://www.rfc-editor.org/info/rfc8310>.
@misc{rfc8310,
series = {Request for Comments},
number = 8310,
howpublished = {RFC 8310},
publisher = {RFC Editor},
doi = {10.17487/RFC8310},
url = {https://www.rfc-editor.org/info/rfc8310},
author = {S. Dickinson and D. Gillmor and T. Reddy},
title = {{Usage Profiles for DNS over TLS and DNS over DTLS}},
pagetotal = 27,
year = 2018,
month = mar,
} RFC text
Contents
Internet Engineering Task Force (IETF) S. Dickinson
Request for Comments: 8310 Sinodun
Updates: 7858 D. Gillmor
Category: Standards Track ACLU
ISSN: 2070-1721 T. Reddy
McAfee
March 2018
Usage Profiles for DNS over TLS and DNS over DTLS
Abstract
This document discusses usage profiles, based on one or more
authentication mechanisms, which can be used for DNS over Transport
Layer Security (TLS) or Datagram TLS (DTLS). These profiles can
increase the privacy of DNS transactions compared to using only
cleartext DNS. This document also specifies new authentication
mechanisms -- it describes several ways that a DNS client can use an
authentication domain name to authenticate a (D)TLS connection to a
DNS server. Additionally, it defines (D)TLS protocol profiles for
DNS clients and servers implementing DNS over (D)TLS. This document
updates RFC 7858.
Status of This Memo
This is an Internet Standards Track document.
This document is a product of the Internet Engineering Task Force
(IETF). It represents the consensus of the IETF community. It has
received public review and has been approved for publication by the
Internet Engineering Steering Group (IESG). Further information on
Internet Standards is available in Section 2 of RFC 7841.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
https://www.rfc-editor.org/info/rfc8310.
Copyright Notice
Copyright (c) 2018 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.
Table of Contents
1. Introduction ....................................................4
2. Terminology .....................................................6
3. Scope ...........................................................7
4. Discussion ......................................................8
5. Usage Profiles ..................................................8
5.1. DNS Resolution ............................................11
6. Authentication in DNS over (D)TLS ..............................11
6.1. DNS-over-(D)TLS Startup Configuration Problems ............11
6.2. Credential Verification ...................................12
6.3. Summary of Authentication Mechanisms ......................12
6.4. Combining Authentication Mechanisms .......................15
6.5. Authentication in Opportunistic Privacy ...................15
6.6. Authentication in Strict Privacy ..........................16
6.7. Implementation Guidance ...................................16
7. Sources of Authentication Domain Names .........................17
7.1. Full Direct Configuration .................................17
7.2. Direct Configuration of ADN Only ..........................17
7.3. Dynamic Discovery of ADN ..................................17
7.3.1. DHCP ...............................................18
8. Credential Verification Based on Authentication Domain Name ....18
8.1. Authentication Based on PKIX Certificate ..................18
8.2. DANE ......................................................19
8.2.1. Direct DNS Meta-Queries ............................20
8.2.2. TLS DNSSEC Chain Extension .........................20
9. (D)TLS Protocol Profile ........................................20
10. IANA Considerations ...........................................21
11. Security Considerations .......................................21
11.1. Countermeasures to DNS Traffic Analysis ..................22
12. References ....................................................22
12.1. Normative References .....................................22
12.2. Informative References ...................................24
Appendix A. Server Capability Probing and Caching by DNS Clients ..26
Acknowledgments ...................................................27
Authors' Addresses ................................................27
1. Introduction
DNS privacy issues are discussed in [RFC7626]. The specific issues
described in [RFC7626] that are most relevant to this document are
o Passive attacks that eavesdrop on cleartext DNS transactions on
the wire (Section 2.4 of [RFC7626]) and
o Active attacks that redirect clients to rogue servers to monitor
DNS traffic (Section 2.5.3 of [RFC7626]).
Mitigating these attacks increases the privacy of DNS transactions;
however, many of the other issues raised in [RFC7626] still apply.
Two documents that provide ways to increase DNS privacy between DNS
clients and DNS servers are
o "Specification for DNS over Transport Layer Security (TLS)"
[RFC7858], referred to here as simply "DNS over TLS".
o "DNS over Datagram Transport Layer Security (DTLS)" [RFC8094],
referred to here as simply "DNS over DTLS". Note that [RFC8094]
is an Experimental specification.
Both documents are limited in scope to communications between stub
clients and recursive resolvers, and the same scope is applied to
this document (see Sections 2 and 3). The proposals here might be
adapted or extended in future to be used for recursive clients and
authoritative servers, but this application was out of scope for the
DNS PRIVate Exchange (dprive) Working Group charter at the time this
document was published.
This document specifies two usage profiles (Strict Privacy and
Opportunistic Privacy) for DTLS [RFC6347] and TLS [RFC5246] that
provide improved levels of mitigation for the attacks described above
compared to using only cleartext DNS.
Section 5 presents a generalized discussion of usage profiles by
separating the usage profile, which is based purely on the security
properties it offers the user, from the specific mechanism or
mechanisms that are used for DNS server authentication. The profiles
described are
o A Strict Privacy profile, which requires an encrypted connection
and successful authentication of the DNS server; this mitigates
both passive eavesdropping and client redirection (at the expense
of providing no DNS service if an encrypted, authenticated
connection is not available).
o An Opportunistic Privacy profile, which will attempt, but does not
require, encryption and successful authentication; it therefore
provides limited or no mitigation for such attacks but maximizes
the chance of DNS service.
The above usage profiles attempt authentication of the server using
at least one authentication mechanism. Section 6.4 discusses how to
combine authentication mechanisms to determine the overall
authentication result. Depending on that overall authentication
result (and whether encryption is available), the usage profile will
determine if the connection should proceed, fall back, or fail.
One authentication mechanism is already described in [RFC7858].
[RFC7858] specifies an authentication mechanism for DNS over TLS that
is based on Subject Public Key Info (SPKI) in the context of a
specific case of a Strict Privacy profile using that single
authentication mechanism. Therefore, the "out-of-band key-pinned
privacy profile" described in [RFC7858] would qualify as a "Strict
Privacy profile" that used SPKI pinning for authentication.
This document extends the use of authentication based on SPKI
pin sets, so that it is considered a general authentication mechanism
that can be used with either DNS-over-(D)TLS usage profile. That is,
the mechanism for SPKI pin sets as described in [RFC7858] MAY be used
with DNS over (D)TLS.
This document also describes a number of additional authentication
mechanisms, all of which specify how a DNS client should authenticate
a DNS server based on an "authentication domain name". In
particular, the following topics are described:
o How a DNS client can obtain the combination of an authentication
domain name and IP address for a DNS server. See Section 7.
o What acceptable credentials a DNS server can present to prove its
identity for (D)TLS authentication based on a given authentication
domain name. See Section 8.
o How a DNS client can verify that any given credential matches the
authentication domain name obtained for a DNS server. See
Section 8.
This document defines a (D)TLS protocol profile for use with DNS; see
Section 9. This profile defines the configuration options and
protocol extensions required of both parties to (1) optimize
connection establishment and session resumption for transporting DNS
and (2) support all currently specified authentication mechanisms.
2. Terminology
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.
Several terms are used specifically in the context of this document:
o DNS client: A DNS stub resolver or forwarder. In the case of a
forwarder, the term "DNS client" is used to discuss the side that
sends queries.
o DNS server: A DNS recursive resolver or forwarder. In the case of
a forwarder, the term "DNS server" is used to discuss the side
that responds to queries. Note that, as used in this document,
this term does not apply to authoritative servers.
o Privacy-enabling DNS server: A DNS server that implements
DNS over TLS [RFC7858] and may optionally implement DNS over DTLS
[RFC8094]. The server should also offer at least one of the
credentials described in Section 8 and implement the (D)TLS
profile described in Section 9.
o (D)TLS: Used for brevity; refers to both Transport Layer Security
[RFC5246] and Datagram Transport Layer Security [RFC6347].
Specific terms will be used for any text that applies to either
protocol alone.
o DNS over (D)TLS: Used for brevity; refers to both DNS over TLS
[RFC7858] and DNS over DTLS [RFC8094]. Specific terms will be
used for any text that applies to either protocol alone.
o Authentication domain name: A domain name that can be used to
authenticate a privacy-enabling DNS server. Sources of
authentication domain names are discussed in Section 7.
o SPKI pin sets: [RFC7858] describes the use of cryptographic
digests to "pin" public key information in a manner similar to
HTTP Public Key Pinning (HPKP) [RFC7469]. An SPKI pin set is a
collection of these pins that constrains a DNS server.
o Authentication information: Information a DNS client may use as
the basis of an authentication mechanism. In this context, this
information can be either
* an SPKI pin set or
* an authentication domain name
o Reference identifier: A reference identifier as described in
[RFC6125], constructed by the DNS client when performing TLS
authentication of a DNS server.
o Credential: Information available for a DNS server that proves its
identity for authentication purposes. Credentials discussed here
include
* a PKIX certificate
* a DNSSEC-validated chain to a TLSA record
but may also include SPKI pin sets.
3. Scope
This document is limited to describing
o Usage profiles based on general authentication mechanisms.
o The details of domain-name-based authentication of DNS servers by
DNS clients (as defined in Section 2).
o The (D)TLS profiles needed to support authentication in
DNS over (D)TLS.
As such, the following topics are out of scope for this document:
o Authentication of authoritative servers by recursive resolvers.
o Authentication of DNS clients by DNS servers.
o The details of how to perform authentication based on SPKI
pin sets. This is defined in [RFC7858].
o Any server identifier other than domain names, including IP
addresses, organizational names, country of origin, etc.
4. Discussion
One way to mitigate eavesdropping on cleartext DNS transactions by
passive attackers is to encrypt the query (and response). Such
encryption typically provides integrity protection as a side effect;
this means that on-path attackers cannot simply inject bogus DNS
responses. To also mitigate active attackers pretending to be the
server, the client must authenticate the (D)TLS connection to the
server.
This document discusses usage profiles, which provide differing
levels of attack mitigation to DNS clients, based on the requirements
for authentication and encryption, regardless of the context (for
example, which network the client is connected to). A usage profile
is a concept distinct from a usage policy or usage model; a usage
policy or usage model might dictate which profile should be used in a
particular context (enterprise vs. coffee shop), with a particular
set of DNS servers or with reference to other external factors. A
description of the variety of usage policies is out of scope for this
document but may be the subject of future work.
The term "privacy-enabling DNS server" is used throughout this
document. This is a DNS server that
o MUST implement DNS over TLS [RFC7858].
o MAY implement DNS over DTLS [RFC8094].
o SHOULD offer at least one of the credentials described in
Section 8.
o Implements the (D)TLS profile described in Section 9.
5. Usage Profiles
A DNS client has a choice of usage profiles available to increase the
privacy of DNS transactions. This choice is briefly discussed in
both [RFC7858] and [RFC8094]. These usage profiles are
o Strict Privacy profile: The DNS client requires both an encrypted
and authenticated connection to a privacy-enabling DNS server. A
hard failure occurs if this is not available. This requires the
client to securely obtain authentication information it can use to
authenticate the server. This profile mitigates both passive and
active attacks, thereby providing the client with the best
available privacy for DNS. This profile is discussed in detail in
Section 6.6.
o Opportunistic Privacy profile: The DNS client uses Opportunistic
Security as described in [RFC7435].
* "... the use of cleartext as the baseline communication
security policy, with encryption and authentication negotiated
and applied to the communication when available."
As described in [RFC7435], it might result in
* an encrypted and authenticated connection
* an encrypted connection
* a cleartext connection
depending on the fallback logic of the client, the available
authentication information, and the capabilities of the DNS
server. In all these cases, the DNS client is willing to continue
with a connection to the DNS server and perform resolution of
queries. The use of Opportunistic Privacy is intended to support
incremental deployment of increased privacy with a view to
widespread adoption of the Strict Privacy profile. It should be
employed when the DNS client might otherwise settle for cleartext;
it provides the maximum protection available, depending on the
combination of factors described above. If all the configured DNS
servers are DNS privacy servers, then it can provide protection
against passive attacks and might protect against active ones.
Both profiles can include an initial meta-query (performed using
Opportunistic Privacy) to obtain the IP address for the privacy-
enabling DNS server to which the DNS client will subsequently
connect. The rationale for permitting this for the Strict Privacy
profile is that requiring such meta-queries to also be performed
using the Strict Privacy profile would introduce significant
deployment obstacles. However, it should be noted that in this
scenario an active attack on the meta-query is possible. Such an
attack could result in a Strict Privacy profile client connecting to
a server it cannot authenticate (and therefore not obtaining DNS
service) or an Opportunistic Privacy client connecting to a server
controlled by the attacker. DNSSEC validation can detect the attack
on the meta-query, which may result in the client not obtaining DNS
service (for both usage profiles), depending on its DNSSEC validation
policy. See Section 7.2 for more discussion.
To compare the two usage profiles, Table 1 below shows a successful
Strict Privacy profile alongside the three possible outcomes of an
Opportunistic Privacy profile. In the best-case scenario for the
Opportunistic Privacy profile (an authenticated and encrypted
connection), it is equivalent to the Strict Privacy profile. In the
worst-case scenario, it is equivalent to cleartext. Clients using
the Opportunistic Privacy profile SHOULD try for the best case but
MAY fall back to the intermediate case and, eventually, the worst-
case scenario, in order to obtain a response. One reason to fall
back without trying every available privacy-enabling DNS server is if
latency is more important than attack mitigation; see Appendix A.
The Opportunistic Privacy profile therefore provides varying
protection, depending on what kind of connection is actually used,
including no attack mitigation at all.
Note that there is no requirement in Opportunistic Security to notify
the user regarding what type of connection is actually used; the
"detection" described below is only possible if such connection
information is available. However, if it is available and the user
is informed that an unencrypted connection was used to connect to a
server, then the user should assume (detect) that the connection is
subject to both active and passive attacks, since the DNS queries are
sent in cleartext. This might be particularly useful if a new
connection to a certain server is unencrypted when all previous
connections were encrypted. Similarly, if the user is informed that
an encrypted but unauthenticated connection was used, then the user
can detect that the connection may be subject to active attacks. In
other words, for the cases where no protection is provided against an
attacker (N), it is possible to detect that an attack might be
happening (D). This is discussed in Section 6.5.
+---------------+------------+------------------+-----------------+
| Usage Profile | Connection | Passive Attacker | Active Attacker |
+---------------+------------+------------------+-----------------+
| Strict | A, E | P | P |
| Opportunistic | A, E | P | P |
| Opportunistic | E | P | N, D |
| Opportunistic | | N, D | N, D |
+---------------+------------+------------------+-----------------+
P == Protection; N == No protection; D == Detection is possible;
A == Authenticated connection; E == Encrypted connection
Table 1: Attack Protection by Usage Profile and Type of Attacker
The Strict Privacy profile provides the best attack mitigation and
therefore SHOULD always be implemented in DNS clients that implement
the Opportunistic Privacy profile.
A DNS client that implements DNS over (D)TLS SHOULD NOT be configured
by default to use only cleartext.
The choice between the two profiles depends on a number of factors,
including which is more important to the particular client:
o DNS service, at the cost of no attack mitigation (Opportunistic
Privacy) or
o Best available attack mitigation, at the potential cost of no DNS
service (Strict Privacy).
Additionally, the two profiles require varying levels of
configuration (or a trusted relationship with a provider) and DNS
server capabilities; therefore, DNS clients will need to carefully
select which profile to use based on their communication needs.
A DNS server that implements DNS over (D)TLS SHOULD provide at least
one credential (Section 2) so that those DNS clients that wish to use
the Strict Privacy profile are able to do so.
5.1. DNS Resolution
A DNS client SHOULD select a particular usage profile when resolving
a query. A DNS client MUST NOT fall back from Strict Privacy to
Opportunistic Privacy during the resolution of a given query, as this
could invalidate the protection offered against attackers. It is
anticipated that DNS clients will use a particular usage profile for
all queries to all configured servers until an operational issue or
policy update dictates a change in the profile used.
6. Authentication in DNS over (D)TLS
This section describes authentication mechanisms and how they can be
used in either Strict or Opportunistic Privacy for DNS over (D)TLS.
6.1. DNS-over-(D)TLS Startup Configuration Problems
Many (D)TLS clients use PKIX authentication [RFC6125] based on an
authentication domain name for the server they are contacting. These
clients typically first look up the server's network address in the
DNS before making this connection. Such a DNS client therefore has a
bootstrap problem, as it will typically only know the IP address of
its DNS server.
In this case, before connecting to a DNS server, a DNS client needs
to learn the authentication domain name it should associate with the
IP address of a DNS server for authentication purposes. Sources of
authentication domain names are discussed in Section 7.
One advantage of this domain-name-based approach is that it
encourages the association of stable, human-recognizable identifiers
with secure DNS service providers.
6.2. Credential Verification
Verification of SPKI pin sets is discussed in [RFC7858].
In terms of domain-name-based verification, once an authentication
domain name is known for a DNS server, a choice of authentication
mechanisms can be used for credential verification. Section 8
discusses these mechanisms -- namely, PKIX certificate-based
authentication and DNS-Based Authentication of Named Entities (DANE)
-- in detail.
Note that the use of DANE adds requirements on the ability of the
client to get validated DNSSEC results. This is discussed in more
detail in Section 8.2.
6.3. Summary of Authentication Mechanisms
This section provides an overview of the various authentication
mechanisms. Table 2 below indicates how the DNS client obtains
information to use for authentication for each option: either
statically via direct configuration or dynamically. Of course, the
Opportunistic Privacy profile does not require authentication, and so
a client using that profile may choose to connect to a
privacy-enabling DNS server on the basis of just an IP address.
+---+------------+-------------+------------------------------------+
| # | Static | Dynamically | Short name: Description |
| | Config | Obtained | |
+---+------------+-------------+------------------------------------+
| 1 | SPKI + IP | | SPKI: SPKI pin set(s) and IP |
| | | | address obtained out of band |
| | | | [RFC7858] |
| | | | |
| 2 | ADN + IP | | ADN: ADN and IP address obtained |
| | | | out of band (see Section 7.1) |
| | | | |
| 3 | ADN | IP | ADN only: Opportunistic Privacy |
| | | | meta-queries to a NP DNS server |
| | | | for A/AAAA (see Section 7.2) |
| | | | |
| 4 | | ADN + IP | DHCP: DHCP configuration only (see |
| | | | Section 7.3.1) |
| | | | |
| 5 | [ADN + IP] | [ADN + IP] | DANE: DNSSEC chain obtained via |
| | | TLSA record | Opportunistic Privacy meta-queries |
| | | | to NP DNS server (see Section |
| | | | 8.2.1) |
| | | | |
| 6 | [ADN + IP] | [ADN + IP] | TLS extension: DNSSEC chain |
| | | TLSA record | provided by PE DNS server in TLS |
| | | | DNSSEC chain extension (see |
| | | | Section 8.2.2) |
+---+------------+-------------+------------------------------------+
SPKI == SPKI pin set(s); IP == IP Address;
ADN == Authentication Domain Name; NP == Network-Provided;
PE == Privacy-Enabling; [ ] == Data may be obtained either
statically or dynamically
Table 2: Overview of Authentication Mechanisms
The following summary attempts to present some key attributes of each
of the mechanisms (using the "Short name" from Table 2), indicating
attractive attributes with a "+" and undesirable attributes
with a "-".
1. SPKI
+ Minimal leakage (note that the ADN is always leaked in the
Server Name Indication (SNI) field in the ClientHello in TLS
when communicating with a privacy-enabling DNS server)
- Overhead of ongoing key management required
2. ADN
+ Minimal leakage
+ One-off direct configuration only
3. ADN only
+ Minimal one-off direct configuration; only a human-recognizable
domain name needed
- A/AAAA meta-queries leaked to network-provided DNS server that
may be subject to active attack (attack can be mitigated by
DNSSEC validation)
4. DHCP
+ No static config
- Requires a non-standard or future DHCP option in order to
provide the ADN
- Requires secure and trustworthy connection to DHCP server if
used with a Strict Privacy profile
5. DANE
The ADN and/or IP may be obtained statically or dynamically, and
the relevant attributes of that method apply.
+ DANE options (e.g., matching on entire certificate)
- Requires a DNSSEC-validating stub implementation (the
deployment of which is limited at the time of this writing)
- DNSSEC chain meta-queries leaked to network-provided DNS server
that may be subject to active attack
6. TLS extension
The ADN and/or IP may be obtained statically or dynamically, and
the relevant attributes of that method apply.
+ Reduced latency compared with DANE
+ No network-provided DNS server required if ADN and IP
statically configured
+ DANE options (e.g., matching on entire certificate)
- Requires a DNSSEC-validating stub implementation
6.4. Combining Authentication Mechanisms
This document does not make explicit recommendations about how an
authentication mechanism based on SPKI pin sets should be combined
with a domain-based mechanism from an operator perspective. However,
it can be envisaged that a DNS server operator may wish to make both
an SPKI pin set and an authentication domain name available to allow
clients to choose which mechanism to use. Therefore, the following
text provides guidance on how clients ought to behave if they choose
to configure both, as is possible in HPKP [RFC7469].
A DNS client that is configured with both an authentication domain
name and an SPKI pin set for a DNS server SHOULD match on both a
valid credential for the authentication domain name and a valid SPKI
pin set (if both are available) when connecting to that DNS server.
In this case, the client SHOULD treat individual SPKI pins as
specified in Section 2.6 of [RFC7469] with regard to user-defined
trust anchors. The overall authentication result SHOULD only be
considered successful if both authentication mechanisms are
successful.
6.5. Authentication in Opportunistic Privacy
An Opportunistic Privacy Profile (based on Opportunistic Security
[RFC7435]) that MAY be used for DNS over (D)TLS is described in
[RFC7858] and is further specified in this document.
DNS clients that issue queries under an Opportunistic Privacy profile
and that know authentication information for a given privacy-enabling
DNS server SHOULD try to authenticate the server using the mechanisms
described here. This is useful for detecting (but not preventing)
active attacks, since the fact that authentication information is
available indicates that the server in question is a privacy-enabling
DNS server to which it should be possible to establish an
authenticated and encrypted connection. In this case, whilst a
client cannot know the reason for an authentication failure, from a
security standpoint the client should consider an active attack in
progress and proceed under that assumption. For example, a client
that implements a nameserver selection algorithm that preferentially
uses nameservers that successfully authenticated (see Section 5)
might not continue to use the failing server if there were
alternative servers available.
Attempting authentication is also useful for debugging or diagnostic
purposes if there are means to report the result. This information
can provide a basis for a DNS client to switch to (preferred) Strict
Privacy where it is viable, e.g., where all the configured servers
support DNS over (D)TLS and successfully authenticate.
6.6. Authentication in Strict Privacy
To authenticate a privacy-enabling DNS server, a DNS client needs to
know authentication information for each server it is willing to
contact. This is necessary to protect against active attacks that
attempt to redirect clients to rogue DNS servers.
A DNS client requiring Strict Privacy MUST use either (1) one of the
sources listed in Section 7, to obtain an authentication domain name
for the server it contacts or (2) an SPKI pin set as described in
[RFC7858].
A DNS client requiring Strict Privacy MUST only attempt to connect to
DNS servers for which at least one piece of authentication
information is known. The client MUST use the available verification
mechanisms described in Section 8 to authenticate the server and MUST
abort connections to a server when no verification mechanism
succeeds.
With Strict Privacy, the DNS client MUST NOT commence sending DNS
queries until at least one of the privacy-enabling DNS servers
becomes available.
A privacy-enabling DNS server may be temporarily unavailable when
configuring a network. For example, for clients on networks that
require registration through web-based login (a.k.a. "captive
portals"), such registration may rely on DNS interception and
spoofing. Techniques such as those used by dnssec-trigger
[dnssec-trigger] MAY be used during network configuration, with the
intent to transition to the designated privacy-enabling DNS servers
after captive-portal registration. If using a Strict Privacy
profile, the system MUST alert by some means that the DNS is not
private during such a bootstrap operation.
6.7. Implementation Guidance
Section 9 describes the (D)TLS profile for DNS over (D)TLS.
Additional considerations relating to general implementation
guidelines are discussed in both Section 11 and Appendix A.
7. Sources of Authentication Domain Names
7.1. Full Direct Configuration
DNS clients may be directly and securely provisioned with the
authentication domain name of each privacy-enabling DNS server -- for
example, using a client-specific configuration file or API.
In this case, direct configuration for a DNS client would consist of
both an IP address and an authentication domain name for each DNS
server that were obtained via an out-of-band mechanism.
7.2. Direct Configuration of ADN Only
A DNS client may be configured directly and securely with only the
authentication domain name of each of its privacy-enabling DNS
servers -- for example, using a client-specific configuration file
or API.
A DNS client might learn of a default recursive DNS resolver from an
untrusted source (such as DHCP's DNS Recursive Name Server option
[RFC3646]). It can then use meta-queries performed using an
Opportunistic Privacy profile to an untrusted recursive DNS resolver
to establish the IP address of the intended privacy-enabling DNS
resolver by doing a lookup of A/AAAA records. A DNSSEC-validating
client SHOULD apply the same validation policy to the A/AAAA
meta-queries as it does to other queries. A client that does not
validate DNSSEC SHOULD apply the same policy (if any) to the A/AAAA
meta-queries as it does to other queries. Private DNS resolution can
now be done by the DNS client against the pre-configured privacy-
enabling DNS resolver, using the IP address obtained from the
untrusted DNS resolver.
A DNS client so configured that successfully connects to a privacy-
enabling DNS server MAY choose to locally cache the server host IP
addresses in order to not have to repeat the meta-query.
7.3. Dynamic Discovery of ADN
This section discusses the general case of a DNS client discovering
both the authentication domain name and IP address dynamically. At
the time of this writing, this is not possible by any standard means.
However, since, for example, a future DHCP extension could (in
principle) provide this mechanism, the required security properties
of such mechanisms are outlined here.
When using a Strict Privacy profile, the dynamic discovery technique
used as a source of authentication domain names MUST be considered
secure and trustworthy. This requirement does not apply when using
an Opportunistic Privacy profile, given the security expectation of
that profile.
7.3.1. DHCP
In the typical case today, a DHCP server [RFC2131] [RFC3315] provides
a list of IP addresses for DNS resolvers (see Section 3.8 of
[RFC2132]) but does not provide an authentication domain name for the
DNS resolver, thus preventing the use of most of the authentication
methods described here (all of those that are based on a mechanism
with ADN; see Table 2).
This document does not specify or request any DHCP extension to
provide authentication domain names. However, if one is developed in
future work, the issues outlined in Section 8 of [RFC7227] should be
taken into account, as should the security considerations discussed
in Section 23 of [RFC3315].
This document does not attempt to describe secured and trusted
relationships to DHCP servers, as this is purely a DHCP issue (and
still open, at the time of this writing). Whilst some implementation
work is in progress to secure IPv6 connections for DHCP, IPv4
connections have received little or no implementation attention in
this area.
8. Credential Verification Based on Authentication Domain Name
8.1. Authentication Based on PKIX Certificate
When a DNS client configured with an authentication domain name
connects to its configured DNS server over (D)TLS, the server may
present it with a PKIX certificate. In order to ensure proper
authentication, DNS clients MUST verify the entire certification path
per [RFC5280]. The DNS client additionally uses validation
techniques as described in [RFC6125] to compare the domain name to
the certificate provided.
A DNS client constructs one reference identifier for the server based
on the authentication domain name: a DNS-ID, which is simply the
authentication domain name itself.
If the reference identifier is found (as described in Section 6 of
[RFC6125]) in the PKIX certificate's subjectAltName extension, the
DNS client should accept the certificate for the server.
A compliant DNS client MUST only inspect the certificate's
subjectAltName extension for the reference identifier. In
particular, it MUST NOT inspect the Subject field itself.
8.2. DANE
DANE [RFC6698] provides various mechanisms using DNSSEC to anchor
trust for certificates and raw public keys. However, this requires
the DNS client to have an authentication domain name (which must be
obtained via a trusted source) for the DNS privacy server.
This section assumes a solid understanding of both DANE [RFC6698] and
DANE operations [RFC7671]. A few pertinent issues covered in these
documents are outlined here as useful pointers, but familiarity with
both of these documents in their entirety is expected.
Note that [RFC6698] says
Clients that validate the DNSSEC signatures themselves MUST use
standard DNSSEC validation procedures. Clients that rely on
another entity to perform the DNSSEC signature validation MUST use
a secure mechanism between themselves and the validator.
Note that [RFC7671] covers the following topics:
o Sections 4.1 ("Opportunistic Security and PKIX Usages") and 14
("Security Considerations") of [RFC7671], which both discuss the
use of schemes based on trust anchors and end entities (PKIX-TA(0)
and PKIX-EE(1), respectively) for Opportunistic Security.
o Section 5 ("Certificate-Usage-Specific DANE Updates and
Guidelines") of [RFC7671] -- specifically, Section 5.1 of
[RFC7671], which outlines the combination of certificate usage
DANE-EE(3) and selector SPKI(1) with raw public keys [RFC7250].
Section 5.1 of [RFC7671] also discusses the security implications
of this mode; for example, it discusses key lifetimes and
specifies that validity period enforcement is based solely on the
TLSA RRset properties for this case.
o Section 13 ("Operational Considerations") of [RFC7671], which
discusses TLSA TTLs and signature validity periods.
The specific DANE record for a DNS privacy server would take the form
_853._tcp.[authentication-domain-name] for TLS
_853._udp.[authentication-domain-name] for DTLS
8.2.1. Direct DNS Meta-Queries
The DNS client MAY choose to perform the DNS meta-queries to retrieve
the required DANE records itself. The DNS meta-queries for such DANE
records MAY use the Opportunistic Privacy profile or be in the clear
to avoid trust recursion. The records MUST be validated using DNSSEC
as described in [RFC6698].
8.2.2. TLS DNSSEC Chain Extension
The DNS client MAY offer the TLS extension described in
[TLS-DNSSEC-Chain-Ext]. If the DNS server supports this extension,
it can provide the full chain to the client in the handshake.
If the DNS client offers the TLS DNSSEC chain extension, it MUST be
capable of validating the full DNSSEC authentication chain down to
the leaf. If the supplied DNSSEC chain does not validate, the client
MUST ignore the DNSSEC chain and validate only via other supplied
credentials.
9. (D)TLS Protocol Profile
This section defines the (D)TLS protocol profile of DNS over (D)TLS.
Clients and servers MUST adhere to the (D)TLS implementation
recommendations and security considerations of [RFC7525], except with
respect to the (D)TLS version.
Since encryption of DNS using (D)TLS is a greenfield deployment, DNS
clients and servers MUST implement only (D)TLS 1.2 or later. For
example, implementing (D)TLS 1.3 [TLS-1.3] [DTLS-1.3] is also an
option.
Implementations MUST NOT offer or provide TLS compression, since
compression can leak significant amounts of information, especially
to a network observer capable of forcing the user to do an arbitrary
DNS lookup in the style of the Compression Ratio Info-leak Made Easy
(CRIME) attacks [CRIME].
Implementations compliant with this profile MUST implement the
following items:
o TLS session resumption without server-side state [RFC5077], which
eliminates the need for the server to retain cryptographic state
for longer than necessary. (This statement updates [RFC7858].)
o Raw public keys [RFC7250], which reduce the size of the
ServerHello and can be used by servers that cannot obtain
certificates (e.g., DNS servers on private networks). A client
MUST only indicate support for raw public keys if it has an SPKI
pin set pre-configured (for interoperability reasons).
Implementations compliant with this profile SHOULD implement the
following items:
o TLS False Start [RFC7918], which reduces round trips by allowing
the TLS second flight of messages (ChangeCipherSpec) to also
contain the (encrypted) DNS query.
o The Cached Information Extension [RFC7924], which avoids
transmitting the server's certificate and certificate chain if the
client has cached that information from a previous TLS handshake.
Guidance specific to TLS is provided in [RFC7858], and guidance
specific to DTLS is provided in [RFC8094].
10. IANA Considerations
This document does not require any IANA actions.
11. Security Considerations
Security considerations discussed in [RFC7525], [RFC8094], and
[RFC7858] apply to this document.
DNS clients SHOULD implement (1) support for the mechanisms described
in Section 8.2 and (2) offering a configuration option that limits
authentication to using only those mechanisms (i.e., with no fallback
to pure PKIX-based authentication) such that authenticating solely
via the PKIX infrastructure can be avoided.
11.1. Countermeasures to DNS Traffic Analysis
This section makes suggestions for measures that can reduce the
ability of attackers to infer information pertaining to encrypted
client queries by other means (e.g., via an analysis of encrypted
traffic size or via monitoring of the unencrypted traffic from a DNS
recursive resolver to an authoritative server).
DNS-over-(D)TLS clients and servers SHOULD implement the following
relevant DNS extensions:
o Extension Mechanisms for DNS (EDNS(0)) padding [RFC7830], which
allows encrypted queries and responses to hide their size, making
analysis of encrypted traffic harder.
Guidance on padding policies for EDNS(0) is provided in
[EDNS0-Pad-Policies].
DNS-over-(D)TLS clients SHOULD implement the following relevant DNS
extensions:
o Privacy election per [RFC7871] ("Client Subnet in DNS Queries").
If a DNS client does not include an edns-client-subnet EDNS0
option with SOURCE PREFIX-LENGTH set to 0 in a query, the DNS
server may potentially leak client address information to the
upstream authoritative DNS servers. A DNS client ought to be able
to inform the DNS resolver that it does not want any address
information leaked, and the DNS resolver should honor that
request.
12. References
12.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>.
[RFC5077] Salowey, J., Zhou, H., Eronen, P., and H. Tschofenig,
"Transport Layer Security (TLS) Session Resumption without
Server-Side State", RFC 5077, DOI 10.17487/RFC5077,
January 2008, <https://www.rfc-editor.org/info/rfc5077>.
[RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security
(TLS) Protocol Version 1.2", RFC 5246,
DOI 10.17487/RFC5246, August 2008,
<https://www.rfc-editor.org/info/rfc5246>.
[RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S.,
Housley, R., and W. Polk, "Internet X.509 Public Key
Infrastructure Certificate and Certificate Revocation List
(CRL) Profile", RFC 5280, DOI 10.17487/RFC5280, May 2008,
<https://www.rfc-editor.org/info/rfc5280>.
[RFC6125] Saint-Andre, P. and J. Hodges, "Representation and
Verification of Domain-Based Application Service Identity
within Internet Public Key Infrastructure Using X.509
(PKIX) Certificates in the Context of Transport Layer
Security (TLS)", RFC 6125, DOI 10.17487/RFC6125,
March 2011, <https://www.rfc-editor.org/info/rfc6125>.
[RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer
Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347,
January 2012, <https://www.rfc-editor.org/info/rfc6347>.
[RFC6698] Hoffman, P. and J. Schlyter, "The DNS-Based Authentication
of Named Entities (DANE) Transport Layer Security (TLS)
Protocol: TLSA", RFC 6698, DOI 10.17487/RFC6698,
August 2012, <https://www.rfc-editor.org/info/rfc6698>.
[RFC7250] Wouters, P., Ed., Tschofenig, H., Ed., Gilmore, J.,
Weiler, S., and T. Kivinen, "Using Raw Public Keys in
Transport Layer Security (TLS) and Datagram Transport
Layer Security (DTLS)", RFC 7250, DOI 10.17487/RFC7250,
June 2014, <https://www.rfc-editor.org/info/rfc7250>.
[RFC7525] Sheffer, Y., Holz, R., and P. Saint-Andre,
"Recommendations for Secure Use of Transport Layer
Security (TLS) and Datagram Transport Layer Security
(DTLS)", BCP 195, RFC 7525, DOI 10.17487/RFC7525,
May 2015, <https://www.rfc-editor.org/info/rfc7525>.
[RFC7671] Dukhovni, V. and W. Hardaker, "The DNS-Based
Authentication of Named Entities (DANE) Protocol: Updates
and Operational Guidance", RFC 7671, DOI 10.17487/RFC7671,
October 2015, <https://www.rfc-editor.org/info/rfc7671>.
[RFC7830] Mayrhofer, A., "The EDNS(0) Padding Option", RFC 7830,
DOI 10.17487/RFC7830, May 2016,
<https://www.rfc-editor.org/info/rfc7830>.
[RFC7858] Hu, Z., Zhu, L., Heidemann, J., Mankin, A., Wessels, D.,
and P. Hoffman, "Specification for DNS over Transport
Layer Security (TLS)", RFC 7858, DOI 10.17487/RFC7858,
May 2016, <https://www.rfc-editor.org/info/rfc7858>.
[RFC7918] Langley, A., Modadugu, N., and B. Moeller, "Transport
Layer Security (TLS) False Start", RFC 7918,
DOI 10.17487/RFC7918, August 2016,
<https://www.rfc-editor.org/info/rfc7918>.
[RFC7924] Santesson, S. and H. Tschofenig, "Transport Layer Security
(TLS) Cached Information Extension", RFC 7924,
DOI 10.17487/RFC7924, July 2016,
<https://www.rfc-editor.org/info/rfc7924>.
[RFC8094] Reddy, T., Wing, D., and P. Patil, "DNS over Datagram
Transport Layer Security (DTLS)", RFC 8094,
DOI 10.17487/RFC8094, February 2017,
<https://www.rfc-editor.org/info/rfc8094>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in
RFC 2119 Key Words", BCP 14, RFC 8174,
DOI 10.17487/RFC8174, May 2017,
<https://www.rfc-editor.org/info/rfc8174>.
12.2. Informative References
[CRIME] Rizzo, J. and T. Duong, "The CRIME Attack", Ekoparty
Security Conference, 2012,
<https://www.ekoparty.org/archivo/2012/eko8-CRIME.pdf>.
[dnssec-trigger]
NLnetLabs, "Dnssec-Trigger", December 2017,
<https://www.nlnetlabs.nl/projects/dnssec-trigger/>.
[DTLS-1.3]
Rescorla, E., Tschofenig, H., and N. Modadugu, "The
Datagram Transport Layer Security (DTLS) Protocol
Version 1.3", Work in Progress, draft-ietf-tls-dtls13-26,
March 2018.
[EDNS0-Pad-Policies]
Mayrhofer, A., "Padding Policy for EDNS(0)", Work in
Progress, draft-ietf-dprive-padding-policy-04,
February 2018.
[RFC2131] Droms, R., "Dynamic Host Configuration Protocol",
RFC 2131, DOI 10.17487/RFC2131, March 1997,
<https://www.rfc-editor.org/info/rfc2131>.
[RFC2132] Alexander, S. and R. Droms, "DHCP Options and BOOTP Vendor
Extensions", RFC 2132, DOI 10.17487/RFC2132, March 1997,
<https://www.rfc-editor.org/info/rfc2132>.
[RFC3315] Droms, R., Ed., Bound, J., Volz, B., Lemon, T., Perkins,
C., and M. Carney, "Dynamic Host Configuration Protocol
for IPv6 (DHCPv6)", RFC 3315, DOI 10.17487/RFC3315,
July 2003, <https://www.rfc-editor.org/info/rfc3315>.
[RFC3646] Droms, R., Ed., "DNS Configuration options for Dynamic
Host Configuration Protocol for IPv6 (DHCPv6)", RFC 3646,
DOI 10.17487/RFC3646, December 2003,
<https://www.rfc-editor.org/info/rfc3646>.
[RFC7227] Hankins, D., Mrugalski, T., Siodelski, M., Jiang, S., and
S. Krishnan, "Guidelines for Creating New DHCPv6 Options",
BCP 187, RFC 7227, DOI 10.17487/RFC7227, May 2014,
<https://www.rfc-editor.org/info/rfc7227>.
[RFC7435] Dukhovni, V., "Opportunistic Security: Some Protection
Most of the Time", RFC 7435, DOI 10.17487/RFC7435,
December 2014, <https://www.rfc-editor.org/info/rfc7435>.
[RFC7469] Evans, C., Palmer, C., and R. Sleevi, "Public Key Pinning
Extension for HTTP", RFC 7469, DOI 10.17487/RFC7469,
April 2015, <https://www.rfc-editor.org/info/rfc7469>.
[RFC7626] Bortzmeyer, S., "DNS Privacy Considerations", RFC 7626,
DOI 10.17487/RFC7626, August 2015,
<https://www.rfc-editor.org/info/rfc7626>.
[RFC7871] Contavalli, C., van der Gaast, W., Lawrence, D., and
W. Kumari, "Client Subnet in DNS Queries", RFC 7871,
DOI 10.17487/RFC7871, May 2016,
<https://www.rfc-editor.org/info/rfc7871>.
[TLS-1.3] Rescorla, E., "The Transport Layer Security (TLS) Protocol
Version 1.3", Work in Progress, draft-ietf-tls-tls13-27,
March 2018.
[TLS-DNSSEC-Chain-Ext]
Shore, M., Barnes, R., Huque, S., and W. Toorop, "A DANE
Record and DNSSEC Authentication Chain Extension for TLS",
Work in Progress, draft-ietf-tls-dnssec-chain-
extension-06, January 2018.
Appendix A. Server Capability Probing and Caching by DNS Clients
This section presents a non-normative discussion of how DNS clients
might probe for, and cache capabilities of, privacy-enabling DNS
servers.
Deployment of both DNS over TLS and DNS over DTLS will be gradual.
Not all servers will support one or both of these protocols, and the
well-known port might be blocked by some middleboxes. Clients will
be expected to keep track of servers that support DNS over TLS and/or
DNS over DTLS, as well as those that have been previously
authenticated.
If no server capability information is available, then (unless
otherwise specified by the configuration of the DNS client) DNS
clients that implement both TLS and DTLS should try to authenticate
using both protocols before failing or falling back to an
unauthenticated or cleartext connection. DNS clients using an
Opportunistic Privacy profile should try all available servers
(possibly in parallel) in order to obtain an authenticated and
encrypted connection before falling back. (RATIONALE: This approach
can increase latency while discovering server capabilities but
maximizes the chance of sending the query over an authenticated and
encrypted connection.)
Acknowledgments
Thanks to the authors of both [RFC8094] and [RFC7858] for laying the
groundwork for this document and for reviewing the contents. The
authors would also like to thank John Dickinson, Shumon Huque,
Melinda Shore, Gowri Visweswaran, Ray Bellis, Stephane Bortzmeyer,
Jinmei Tatuya, Paul Hoffman, Christian Huitema, and John Levine for
review and discussion of the ideas presented here.
Authors' Addresses
Sara Dickinson
Sinodun Internet Technologies
Magdalen Centre
Oxford Science Park
Oxford OX4 4GA
United Kingdom
Email: sara@sinodun.com
URI: https://www.sinodun.com/
Daniel Kahn Gillmor
ACLU
125 Broad Street, 18th Floor
New York, NY 10004
United States of America
Email: dkg@fifthhorseman.net
Tirumaleswar Reddy
McAfee, Inc.
Embassy Golf Link Business Park
Bangalore, Karnataka 560071
India
Email: TirumaleswarReddy_Konda@McAfee.com
ISO Plain Language
On this page
§1Why DNS Privacy Needs More Than Encryption
DNS privacy issues are discussed in RFC 7626. The two issues most relevant here are passive attacks that eavesdrop on cleartext DNS transactions on the wire, and active attacks that redirect clients to rogue servers to monitor DNS traffic.
Mitigating these attacks increases the privacy of DNS transactions, although many of the other issues raised in RFC 7626 still apply. Two documents already provide ways to increase DNS privacy between DNS clients and DNS servers: DNS over TLS (RFC 7858) and DNS over DTLS (RFC 8094), the latter being an Experimental specification. Both are limited in scope to communication between stub clients and recursive resolvers, and the same scope applies to this document.
Encrypting the query and the response is one way to stop a passive attacker reading them, and it usually brings integrity protection as a side effect, so an on-path attacker cannot simply inject a bogus DNS response. To also stop an active attacker pretending to be the server, the client has to authenticate the (D)TLS connection to the server.
The document also defines a (D)TLS protocol profile for DNS, covering the configuration options and protocol extensions needed by both parties to optimize connection establishment and session resumption, and to support all the authentication mechanisms it specifies.
DNS queries normally travel in the open, where anyone on the path can read them or answer in place of the real server. This document describes two ways to protect them, and the checks a client can make to be sure it is talking to the server it expects.
§2The Terms the Document Introduces
Several terms are used in a specific way. A DNS client is a DNS stub resolver or forwarder, and for a forwarder the term refers to the side that sends queries. A DNS server is a DNS recursive resolver or forwarder, and for a forwarder it refers to the side that responds to queries; the term does not apply to authoritative servers. A privacy-enabling DNS server implements DNS over TLS, may implement DNS over DTLS, should offer at least one of the credentials the document describes, and implements the (D)TLS profile.
An authentication domain name is a domain name that can be used to authenticate a privacy-enabling DNS server. SPKI pin sets are collections of cryptographic digests that pin public key information, as described in RFC 7858, and constrain a DNS server. Authentication information is what a client may use as the basis of an authentication mechanism: either an SPKI pin set or an authentication domain name.
A reference identifier is one constructed by the DNS client when performing TLS authentication, as described in RFC 6125. A credential is information a server has that proves its identity: a PKIX certificate or a DNSSEC-validated chain to a TLSA record, and it may also include SPKI pin sets.
The document needs a name for the domain a client uses to check the server. That name is the authentication domain name, and it is the thing the server has to prove.
§5The Two Usage Profiles
A DNS client has a choice of usage profiles. The Strict Privacy profile requires both an encrypted and an authenticated connection to a privacy-enabling DNS server, and a hard failure occurs if this is not available. It mitigates both passive and active attacks, and therefore gives the client the best available privacy for DNS, at the cost of no DNS service if an encrypted, authenticated connection cannot be had.
The Opportunistic Privacy profile uses Opportunistic Security as described in RFC 7435, where cleartext is the baseline and encryption and authentication are applied when available. It may result in an encrypted and authenticated connection, an encrypted connection, or a cleartext connection, depending on the fallback logic of the client, the available authentication information and the capabilities of the DNS server. In all cases the client is willing to continue and resolve queries. Its purpose is to support incremental deployment of increased privacy with a view to widespread adoption of the Strict Privacy profile.
Both profiles can include an initial meta-query performed using Opportunistic Privacy, to obtain the IP address of the privacy-enabling server the client then connects to. Allowing this in the Strict profile avoids significant deployment obstacles, but it means an active attack on the meta-query is possible. Such an attack could cause a Strict Privacy client to connect to a server it cannot authenticate and so get no DNS service, or an Opportunistic Privacy client to connect to a server the attacker controls. DNSSEC validation can detect the attack, which may result in the client not obtaining DNS service at all, depending on its validation policy.
A DNS client SHOULD select a particular usage profile when resolving a query, and MUST NOT fall back from Strict Privacy to Opportunistic Privacy during the resolution of a given query, since that could invalidate the protection against attackers. The Strict Privacy profile provides the best attack mitigation and SHOULD always be implemented in clients that implement the Opportunistic Privacy profile. A client that implements DNS over (D)TLS SHOULD NOT be configured by default to use only cleartext.
The strict profile refuses to send a query unless the connection is both encrypted and proven. The opportunistic profile sends the query anyway, with whatever protection it can get. That gives a choice between better privacy and always getting an answer.
| Aspect | Strict Privacy | Opportunistic Privacy |
|---|---|---|
| Encrypted connection | Required | Tried, not required |
| Authenticated server | Required | Tried, not required |
| If that is not available | Hard failure, no DNS service | Falls back, possibly to cleartext |
| Passive eavesdropping | Prevented | Prevented only when encrypted |
| Redirection to a rogue server | Prevented | Prevented only when authenticated |
| Purpose THIS RFC | Best available privacy | Incremental deployment and the maximum chance of service |
§6.3Authentication Mechanisms and How to Combine Them
The document summarises six mechanisms for authenticating a DNS server, and whether the information each needs is configured statically or obtained dynamically. The SPKI mechanism uses an SPKI pin set and an IP address obtained out of band; it leaks the least, but needs ongoing key management. ADN uses an authentication domain name and an IP address obtained out of band, so it needs only a one-off direct configuration. ADN only configures just the domain name and obtains the IP address from A/AAAA meta-queries to a network-provided resolver, which leaks those queries to a server that may be under active attack, an attack that DNSSEC validation can mitigate. DHCP discovers both the domain name and the IP address dynamically, but needs a non-standard or future DHCP option and a secure and trustworthy connection to the DHCP server when used with the Strict profile.
DANE obtains a DNSSEC chain from a DNSSEC-validated lookup, so the client needs a DNSSEC-validating stub implementation, and the chain meta-queries leak to a network-provided server that may be under active attack. The TLS extension takes the DNSSEC chain from the privacy-enabling server itself, in the TLS DNSSEC chain extension, which reduces latency compared with DANE and needs no network-provided DNS server when the domain name and IP address are configured statically; it still needs a DNSSEC-validating stub.
The document gives no recommendation on how an operator should combine an SPKI pin set with a domain-based mechanism. A client configured with both SHOULD match on both a valid credential for the authentication domain name and a valid SPKI pin set, when both are available, and the overall authentication result SHOULD only be considered successful if both mechanisms succeed. It also notes that the authentication domain name is always leaked in the Server Name Indication field of the ClientHello in TLS when communicating with a privacy-enabling DNS server.
There are six ways to check the server. The client has to know something about the server in advance, or get it from somewhere trustworthy, and then match what the server presents against it. If the client has two kinds of information, it should insist that both match.
- SPKIPin set and IP address obtained out of band; minimal leakage, but ongoing key management is required.
- ADNAuthentication domain name and IP address obtained out of band; one-off direct configuration.
- ADN onlyOnly the domain name is configured; the address comes from meta-queries that can be attacked.
- DHCPBoth are discovered dynamically, which needs a DHCP option that does not exist yet.
- DANEThe TLSA record comes from a DNSSEC-validated lookup; needs a DNSSEC-validating stub.
- TLS extensionThe DNSSEC chain arrives in the TLS handshake, which lowers latency but still needs a DNSSEC-validating stub.
§7Where the Authentication Domain Name Comes From
Many (D)TLS clients use PKIX authentication based on an authentication domain name for the server they contact, and they typically look up the server's network address in the DNS before making the connection. Such a client has a bootstrap problem, because it typically only knows the IP address of its DNS server. Before connecting, it has to learn the authentication domain name to associate with that IP address.
The document describes three sources. Full direct configuration means the client is directly and securely provisioned with the authentication domain name of each privacy-enabling DNS server, for example through a client-specific configuration file or API, together with the IP address, both obtained out of band.
Direct configuration of the domain name only means the client is configured directly and securely with only the authentication domain name. The client may learn of a default recursive resolver from an untrusted source, such as the DHCP DNS Recursive Name Server option (RFC 3646), and can then use meta-queries performed with the Opportunistic Privacy profile to an untrusted recursive resolver to establish the IP address of the intended privacy-enabling resolver by looking up A/AAAA records. A DNSSEC-validating client SHOULD apply the same validation policy to those meta-queries as to its other queries, and a client that does not validate DNSSEC SHOULD apply the same policy, if any, to them as well. The client may then cache the server host IP addresses locally so that it need not repeat the meta-query.
Dynamic discovery of both the domain name and the IP address was not possible by any standard means at the time of writing, though a future DHCP extension could provide it. When the Strict Privacy profile is used, the dynamic discovery technique MUST be considered secure and trustworthy; this requirement does not apply to the Opportunistic Privacy profile. A typical DHCP server provides a list of IP addresses for DNS resolvers but no authentication domain name, which prevents the use of most of the ADN-based mechanisms, and this document neither specifies nor requests a DHCP extension to provide one.
The client usually starts with only an IP address, so it has to learn the name to check the server against. It can be told the name directly, told just the name and looked up the address, or get both from the network, which today needs an option that does not exist.
- A/AAAA meta-querySent with the Opportunistic Privacy profile to find the address.
- IP address
- (D)TLS handshake and DNS queryAuthenticated against the configured authentication domain name, then encrypted.
§8Checking the Server's Credential
When a DNS client configured with an authentication domain name connects to its configured DNS server over (D)TLS, the server may present it with a PKIX certificate. To ensure proper authentication, the DNS client MUST verify the entire certification path per RFC 5280, and additionally use the validation techniques of RFC 6125 to compare the domain name to the certificate provided.
The client constructs one reference identifier for the server based on the authentication domain name: a DNS-ID, which is simply the authentication domain name itself. If that reference identifier is found in the PKIX certificate's subjectAltName extension, as described in Section 6 of RFC 6125, the client should accept the certificate for the server. A compliant DNS client MUST only inspect the certificate's subjectAltName extension for the reference identifier, and in particular MUST NOT inspect the Subject field itself.
DANE (RFC 6698) provides mechanisms that use DNSSEC to anchor trust for certificates and raw public keys, but this requires the client to have an authentication domain name obtained from a trusted source. The specific DANE record for a DNS privacy server takes the form of _853._tcp.[authentication-domain-name] for TLS and _853._udp.[authentication-domain-name] for DTLS.
The client MAY perform the DNS meta-queries for those DANE records itself, using the Opportunistic Privacy profile or in the clear to avoid trust recursion, and the records MUST be validated using DNSSEC. The client MAY also offer the TLS DNSSEC chain extension, in which case the server can provide the full chain in the handshake. If the client offers that extension, it MUST be capable of validating the full DNSSEC authentication chain down to the leaf, and if the supplied chain does not validate, the client MUST ignore it and validate only via other supplied credentials.
With a normal certificate, the client checks the chain of trust and then looks for the server's domain name in the certificate's alternative name field. With DANE, the client looks up a TLSA record protected by DNSSEC and matches the certificate or key against it.
| Aspect | PKIX certificate | DANE |
|---|---|---|
| Where the trust comes from | The certification path, verified per RFC 5280 | A DNSSEC-validated TLSA record |
| What is compared | The reference identifier in the certificate's subjectAltName | The TLSA record, which can match on the whole certificate |
| Extra requirement on the client | None beyond the certificate path | The client must be able to get validated DNSSEC results |
| Record name for a DNS privacy server | Not applicable | _853._tcp for TLS, _853._udp for DTLS |
§9The (D)TLS Protocol Profile
Clients and servers MUST adhere to the (D)TLS implementation recommendations and security considerations of RFC 7525, except with respect to the (D)TLS version. Since encryption of DNS using (D)TLS is a greenfield deployment, DNS clients and servers MUST implement only (D)TLS 1.2 or later; implementing (D)TLS 1.3 is also an option. Implementations MUST NOT offer or provide TLS compression, since compression can leak significant amounts of information, especially to a network observer able to force the user to do an arbitrary DNS lookup, as in CRIME-style attacks.
Implementations compliant with the profile MUST implement TLS session resumption without server-side state (RFC 5077), which eliminates the need for the server to retain cryptographic state for longer than necessary. This statement updates RFC 7858. They MUST also implement raw public keys (RFC 7250), which reduce the size of the ServerHello and can be used by servers that cannot obtain certificates, such as DNS servers on private networks; a client MUST only indicate support for raw public keys if it has an SPKI pin set pre-configured, for interoperability reasons.
Implementations compliant with the profile SHOULD implement TLS False Start (RFC 7918), which reduces round trips by allowing the second flight of messages to also carry the encrypted DNS query, and the Cached Information Extension (RFC 7924), which avoids transmitting the server's certificate and certificate chain if the client has cached that information from a previous handshake.
The profile sets the rules for the encrypted connection itself: which version of TLS to use, which features to turn on, and which to leave off. Compression is banned because it can leak what was sent.
- Version 1.2 or laterOnly (D)TLS 1.2 or later may be implemented, since DNS encryption is a greenfield deployment.
- No compressionCompression must not be offered, because it can leak information in CRIME-style attacks.
- Stateless resumptionSession resumption without server-side state, so the server need not keep cryptographic state.
- Raw public keysRaw public keys shrink the ServerHello and let servers without certificates work.
- False StartRecommended, so the DNS query can travel in the second flight of messages.
- Cached informationRecommended, so the certificate chain need not be sent again.
§11Security Considerations and Traffic Analysis
The security considerations discussed in RFC 7525, RFC 8094, and RFC 7858 apply. DNS clients SHOULD implement support for the mechanisms in Section 8.2, and SHOULD offer a configuration option that limits authentication to those mechanisms, with no fallback to pure PKIX-based authentication, so that authenticating solely through the PKIX infrastructure can be avoided.
Encryption does not stop an attacker from learning things by looking at the traffic. The document suggests measures that reduce the ability of attackers to infer information about encrypted client queries by other means, such as analysis of encrypted traffic size or monitoring of the unencrypted traffic from a recursive resolver to an authoritative server. Clients and servers SHOULD implement EDNS(0) padding (RFC 7830), which allows encrypted queries and responses to hide their size and makes analysis of encrypted traffic harder. Clients SHOULD implement the privacy election of RFC 7871: if a client does not include an edns-client-subnet option with SOURCE PREFIX-LENGTH set to 0 in a query, the DNS server may potentially leak client address information to upstream authoritative servers, so the client ought to be able to tell the resolver that it does not want any address information leaked, and the resolver should honour that request.
Appendix A adds a non-normative note on server capability probing. Deployment of DNS over TLS and DTLS will be gradual, not all servers will support either protocol, and the well-known port may be blocked by some middleboxes, so clients are expected to keep track of which servers support which protocol and which have already authenticated.
Even with encryption, the size and timing of traffic can give things away. Padding the queries helps hide that. The document also asks clients to be able to say that they do not want their address passed upstream.