Internet Engineering Task Force (IETF)                       K. Fujiwara
Request for Comments: 8198                                          JPRS
Updates: 4035                                                    A. Kato
Category: Standards Track                                      Keio/WIDE
ISSN: 2070-1721                                                W. Kumari
                                                                  Google
                                                               July 2017

                Aggressive Use of DNSSEC-Validated Cache

Abstract

   The DNS relies upon caching to scale; however, the cache lookup
   generally requires an exact match.  This document specifies the use
   of NSEC/NSEC3 resource records to allow DNSSEC-validating resolvers
   to generate negative answers within a range and positive answers from
   wildcards.  This increases performance (i.e., performance, decreases latency), latency, decreases
   resource utilization on both authoritative and recursive servers, and
   increases privacy.  Also, it may help increase resilience to certain
   DoS attacks in some circumstances.

   This document updates RFC 4035 by allowing validating resolvers to
   generate negative answers based upon NSEC/NSEC3 records and positive
   answers in the presence of wildcards.

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
   http://www.rfc-editor.org/info/rfc8198.

Copyright Notice

   Copyright (c) 2017 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
   (http://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  . . . . . . . . . . . . . . . . . . . . . . . .   2
   2.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   3
   3.  Problem Statement . . . . . . . . . . . . . . . . . . . . . .   3
   4.  Background  . . . . . . . . . . . . . . . . . . . . . . . . .   4
   5.  Aggressive Use of DNSSEC-Validated Cache  . . . . . . . . . . . . . . . . . . .   5
     5.1.  NSEC  . . . . . . . . . . . . . . . . . . . . . . . . . .   6
     5.2.  NSEC3 . . . . . . . . . . . . . . . . . . . . . . . . . .   6
     5.3.  Wildcards . . . . . . . . . . . . . . . . . . . . . . . .   6
     5.4.  Consideration on TTL  . . . . . . . . . . . . . . . . . .   6
   6.  Benefits  . . . . . . . . . . . . . . . . . . . . . . . . . .   7
   7.  Update to RFC 4035  . . . . . . . . . . . . . . . . . . . . .   8
   8.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .   8
   9.  Security Considerations . . . . . . . . . . . . . . . . . . .   8
   10. References  . . . . . . . . . . . . . . . . . . . . . . . . .   9
     10.1.  Normative References . . . . . . . . . . . . . . . . . .   9
     10.2.  Informative References . . . . . . . . . . . . . . . . .  10
   Appendix A.  Detailed Implementation Notes  . . . . . . . . . . .  10
   Appendix B.  Procedure for Determining ENT vs. NXDOMAIN with NSEC  11
   Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . .  11
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  12

1.  Introduction

   A DNS negative cache exists, and is used to cache the fact that an
   RRset does not exist.  This method of negative caching requires exact
   matching; this leads to unnecessary additional lookups, increases
   latency, leads to extra resource utilization on both authoritative
   and recursive servers, and decreases privacy by leaking queries.

   This document updates RFC 4035 to allow resolvers to use NSEC/NSEC3
   resource records to synthesize negative answers from the information
   they have in the cache.  This allows validating resolvers to respond
   with a negative answer immediately if the name in question falls into
   a range expressed by an NSEC/NSEC3 resource record already in the
   cache.  It also allows the synthesis of positive answers in the
   presence of wildcard records.

   Aggressive negative caching was first proposed in Section 6 of DNSSEC
   Lookaside Validation (DLV) [RFC5074] in order to find covering NSEC
   records efficiently.

   [RFC8020] and [RES-IMPROVE] propose steps to using NXDOMAIN
   information for more effective caching.  This document takes this
   technique further.

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.

   Many of the specialized terms used in this document are defined in
   DNS Terminology [RFC7719].

   The key words "source of synthesis" in this document are to be
   interpreted as described in [RFC4592].

3.  Problem Statement

   The DNS negative cache caches negative (non-existent) information,
   and requires an exact match in most instances [RFC2308].

   Assume that the (DNSSEC-signed) "example.com" zone contains:

   albatross.example.com

   albatross.example.com. IN A 192.0.2.1
   elephant.example.com
   elephant.example.com.  IN A 192.0.2.2
   zebra.example.com
   zebra.example.com.     IN A 192.0.2.3

   If a validating resolver receives a query for cat.example.com, it
   contacts its resolver (which may be itself) to query the example.com
   servers and will get back an NSEC record stating that there are no
   records (alphabetically) between albatross and elephant, or an NSEC3
   record stating there is nothing between two hashed names.  The
   resolver then knows that cat.example.com does not exist; however, it
   does not use the fact that the proof covers a range (albatross to
   elephant) to suppress queries for other labels that fall within this
   range.  This means that if the validating resolver gets a query for
   ball.example.com (or dog.example.com) it will once again go off and
   query the example.com servers for these names.

   Apart from wasting bandwidth, this also wastes resources on the
   recursive server (it needs to keep state for outstanding queries),
   wastes resources on the authoritative server (it has to answer
   additional questions), increases latency (the end user has to wait
   longer than necessary to get back an NXDOMAIN answer), can be used by
   attackers to cause a DoS (see additional resources), DoS, and also has privacy implications (e.g.,
   typos leak out further than necessary).

   Another example: assume that the (DNSSEC-signed) "example.org" zone
   contains:

   avocado.example.org

   avocado.example.org.   IN A 192.0.2.1
   *.example.org
   *.example.org.         IN A 192.0.2.2
   zucchini.example.org
   zucchini.example.org.  IN A 192.0.2.3

   If a query is received for leek.example.org, the system contacts its
   resolver (which may be itself) to query the example.org servers and
   will get back an NSEC record stating that there are no records
   (alphabetically) between avocado and zucchini (or an NSEC3 record
   stating there is nothing between two hashed names), as well as an
   answer for leek.example.org, with the label count of the signature
   set to two (see [RFC7129], Section 5.3 for more details).

   If the validating resolver gets a query for banana.example.org, it
   will once again go off and query the example.org servers for
   banana.example.org (even though it already has proof that there is a
   wildcard record) -- just like above, this has privacy implications,
   wastes resources, can be used to contribute to a DoS, etc.

4.  Background

   DNSSEC [RFC4035] and [RFC5155] both provide "authenticated denial of
   existence"; this is a cryptographic proof that the queried-for name
   does not exist or the type does not exist.  Proof that a name does
   not exist is accomplished by providing a (DNSSEC-secured) record
   containing the names that appear alphabetically before and after the
   queried-for name.  In the first example above, if the (DNSSEC-
   validating) recursive server were to query for dog.example.com, it
   would receive a (signed) NSEC record stating that there are no labels
   between "albatross" and "elephant" (or, for NSEC3, a similar pair of
   hashed names).  This is a signed, cryptographic proof that these
   names are the ones before and after the queried-for label.  As
   dog.example.com falls within this range, the recursive server knows
   that dog.example.com really does not exist.  Proof that a type does
   not exist is accomplished by providing a (DNSSEC-secured) record
   containing the queried-for name, and a type bitmap that does not
   include the requested type.

   This document specifies that this NSEC/NSEC3 record should be used to
   generate negative answers for any queries that the validating server
   receives that fall within the range covered by the record (for the
   TTL for the record).  This document also specifies that a positive
   answer should be generated for any queries that the validating server
   receives that are proven to be covered by a wildcard record.

   Section 4.5 of [RFC4035] says:

      In theory, a resolver could use wildcards or NSEC RRs to generate
      positive and negative responses (respectively) until the TTL or
      signatures on the records in question expire.  However, it seems
      prudent for resolvers to avoid blocking new authoritative data or
      synthesizing new data on their own.  Resolvers that follow this
      recommendation will have a more consistent view of the namespace.

   And, earlier, Section 4.5 of [RFC4035] says:

      The reason for these recommendations is that, between the initial
      query and the expiration of the data from the cache, the
      authoritative data might have been changed (for example, via
      dynamic update).

   In other words, if a resolver generates negative answers from an NSEC
   record, it will not send any queries for names within that NSEC range
   (for the TTL).  If a new name is added to the zone during this
   interval, the resolver will not know this.  Similarly, if the
   resolver is generating responses from a wildcard record, it will
   continue to do so (for the TTL).

   We believe that this recommendation can be relaxed because, in the
   absence of this technique, a lookup for the exact name could have
   come in during this interval, and so a negative answer could already
   be cached (see [RFC2308] for more background).  This means that zone
   operators should have no expectation that an added name would work
   immediately.  With DNSSEC and Aggressive NSEC, aggressive use of DNSSEC-validated
   cache, the TTL of the NSEC/
   NSEC3 NSEC/NSEC3 record and the SOA.MINIMUM field are
   the authoritative statement of how quickly a name can start working
   within a zone.

5.  Aggressive Use of DNSSEC-Validated Cache

   This document relaxes the restriction given in Section 4.5 of
   [RFC4035].  See Section 7 for more detail.

   If the negative cache of the validating resolver has sufficient
   information to validate the query, the resolver SHOULD use NSEC,
   NSEC3, and wildcard records to synthesize answers as described in
   this document.  Otherwise, it MUST fall back to send the query to the
   authoritative DNS servers.

5.1.  NSEC

   The validating resolver needs to check the existence of an NSEC RR
   matching/covering the source of synthesis and an NSEC RR covering the
   query name.

   If denial of existence can be determined according to the rules set
   out in Section 5.4 of [RFC4035], using NSEC records in the cache,
   then the resolver can immediately return an NXDOMAIN or NODATA (as
   appropriate) response.

5.2.  NSEC3

   NSEC3 aggressive negative caching is more difficult than NSEC
   aggressive caching.  If the zone is signed with NSEC3, the validating
   resolver needs to check the existence of non-terminals and wildcards
   that derive from query names.

   If denial of existence can be determined according to the rules set
   out in [RFC5155], Sections 8.4, 8.5, 8.6, and 8.7, using NSEC3
   records in the cache, then the resolver can immediately return an
   NXDOMAIN or NODATA response (as appropriate).

   If a covering NSEC3 RR has an Opt-Out flag, the covering NSEC3 RR
   does not prove the non-existence of the domain name and the
   aggressive negative caching is not possible for the domain name.

5.3.  Wildcards

   The last paragraph of [RFC4035], Section 4.5 also discusses the use
   of wildcards and NSEC RRs to generate positive responses and
   recommends that it not be relied upon.  Just like the case for the
   aggressive use of NSEC/NSEC3 for negative answers, we revise this
   recommendation.

   As long as the validating resolver can determine that a name would
   not exist without the wildcard match, determined according to the
   rules set out in Section 5.3.4 of [RFC4035] (NSEC), or in Section 8.8
   of [RFC5155], it SHOULD synthesize an answer (or NODATA response) for
   that name using the cached deduced cache-deduced wildcard.  If the corresponding
   wildcard record is not in the cache, it MUST fall back to send the
   query to the authoritative DNS servers.

5.4.  Consideration on TTL

   The TTL value of negative information is especially important,
   because newly added domain names cannot be used while the negative
   information is effective.

   Section 5 of [RFC2308] suggests a maximum default negative cache TTL
   value of 3 hours (10800).  It is RECOMMENDED that validating
   resolvers limit the maximum effective TTL value of negative responses
   (NSEC/NSEC3 RRs) to this same value.

   Section 5 of [RFC2308] also states that a negative cache entry TTL is
   taken from the minimum of the SOA.MINIMUM field and SOA's TTL.  This
   can be less than the TTL of an NSEC or NSEC3 record, since their TTL
   is equal to the SOA.MINIMUM field (see [RFC4035], Section 2.3 and
   [RFC5155], Section 3).

   A resolver that supports aggressive use of NSEC and NSEC3 SHOULD
   reduce the TTL of NSEC and NSEC3 records to match the SOA.MINIMUM
   field in the authority section of a negative response, if SOA.MINIMUM
   is smaller.

6.  Benefits

   The techniques described in this document provide a number of
   benefits, including (in no specific order):

   Reduced latency:  By answering directly from cache, validating
      resolvers can immediately inform clients that the name they are
      looking for does not exist, improving the user experience.

   Decreased recursive server load:  By answering queries from the cache
      by synthesizing answers, validating servers avoid having to send a
      query and wait for a response.  In addition to decreasing the
      bandwidth used, it also means that the server does not need to
      allocate and maintain state, thereby decreasing memory and CPU
      load.

   Decreased authoritative server load:  Because recursive servers can
      answer queries without asking the authoritative server, the
      authoritative servers receive fewer queries.  This decreases the
      authoritative server bandwidth, queries per second, and CPU
      utilization.

   The scale of the benefit depends upon multiple factors, including the
   query distribution.  For example, at the time of this writing, around
   65% of queries to root name servers result in NXDOMAIN responses (see
   statistics from [ROOT-SERVERS]); this technique will eliminate a
   sizable quantity of these.

   The technique described in this document may also mitigate so-called
   "random QNAME attacks", in which attackers send many queries for
   random subdomains to resolvers.  As the resolver will not have the
   answers cached, it has to ask external servers for each random query,
   leading to a DoS on the authoritative servers (and often resolvers).
   Aggressive NSEC
   The technique may help mitigate these attacks by allowing the
   resolver to answer directly from the cache for any random queries
   that fall within already requested ranges.  It will not always work
   as an effective defense, not least because not many zones are DNSSEC
   signed at all -- but it will still provide an additional layer of
   defense.

   As these benefits are only accrued by those using DNSSEC, it is hoped
   that these techniques will lead to more DNSSEC deployment.

7.  Update to RFC 4035

   Section 4.5 of [RFC4035] shows that "In theory, a resolver could use
   wildcards or NSEC RRs to generate positive and negative responses
   (respectively) until the TTL or signatures on the records in question
   expire.  However, it seems prudent for resolvers to avoid blocking
   new authoritative data or synthesizing new data on their own.
   Resolvers that follow this recommendation will have a more consistent
   view of the namespace".

   The paragraph is updated as follows:

   +-----------------------------------------------------------------+
   |  Once the records are validated, DNSSEC-enabled validating      |
   |  resolvers SHOULD use wildcards and NSEC/NSEC3 resource records |
   |  to generate positive and negative responses until the          |
   |  effective TTLs or signatures for those records expire.         |
   +-----------------------------------------------------------------+

8.  IANA Considerations

   This document does not require any IANA actions.

9.  Security Considerations

   Use of NSEC/NSEC3 resource records without DNSSEC validation may
   create serious security issues, and so this technique requires DNSSEC
   validation.

   Newly registered resource records may not be used immediately.
   However, choosing a suitable TTL value and a negative cache TTL value
   (SOA.MINIMUM field) will mitigate the delay concern, and it is not a
   security problem.

   It is also suggested to limit the maximum TTL value of NSEC/NSEC3
   resource records in the negative cache to, for example, 10800 seconds
   (3 hours), to mitigate this issue.

   Although the TTL of NSEC/NSEC3 records is typically fairly short
   (minutes or hours), their RRSIG expiration time can be much further
   in the future (weeks).  An attacker who is able to successfully spoof
   responses might poison a cache with old NSEC/NSEC3 records.  If the
   resolver is not making aggressive use of NSEC/NSEC3, the attacker has
   to repeat the attack for every query.  If the resolver is making
   aggressive use of NSEC/NSEC3, one successful attack would be able to
   suppress many queries for new names, up to the negative TTL.

10.  References

10.1.  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,
              <http://www.rfc-editor.org/info/rfc2119>.

   [RFC2308]  Andrews, M., "Negative Caching of DNS Queries (DNS
              NCACHE)", RFC 2308, DOI 10.17487/RFC2308, March 1998,
              <http://www.rfc-editor.org/info/rfc2308>.

   [RFC4035]  Arends, R., Austein, R., Larson, M., Massey, D., and S.
              Rose, "Protocol Modifications for the DNS Security
              Extensions", RFC 4035, DOI 10.17487/RFC4035, March 2005,
              <http://www.rfc-editor.org/info/rfc4035>.

   [RFC4592]  Lewis, E., "The Role of Wildcards in the Domain Name
              System", RFC 4592, DOI 10.17487/RFC4592, July 2006,
              <http://www.rfc-editor.org/info/rfc4592>.

   [RFC5155]  Laurie, B., Sisson, G., Arends, R., and D. Blacka, "DNS
              Security (DNSSEC) Hashed Authenticated Denial of
              Existence", RFC 5155, DOI 10.17487/RFC5155, March 2008,
              <http://www.rfc-editor.org/info/rfc5155>.

   [RFC7129]  Gieben, R. and W. Mekking, "Authenticated Denial of
              Existence in the DNS", RFC 7129, DOI 10.17487/RFC7129,
              February 2014, <http://www.rfc-editor.org/info/rfc7129>.

   [RFC7719]  Hoffman, P., Sullivan, A., and K. Fujiwara, "DNS
              Terminology", RFC 7719, DOI 10.17487/RFC7719, December
              2015, <http://www.rfc-editor.org/info/rfc7719>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <http://www.rfc-editor.org/info/rfc8174>.

10.2.  Informative References

   [RES-IMPROVE]
              Vixie, P., Joffe, R., and F. Neves, "Improvements to DNS
              Resolvers for Resiliency, Robustness, and Responsiveness",
              Work in Progress, draft-vixie-dnsext-resimprove-00, June
              2010.

   [RFC5074]  Weiler, S., "DNSSEC Lookaside Validation (DLV)", RFC 5074,
              DOI 10.17487/RFC5074, November 2007,
              <http://www.rfc-editor.org/info/rfc5074>.

   [RFC8020]  Bortzmeyer, S. and S. Huque, "NXDOMAIN: There Really Is
              Nothing Underneath", RFC 8020, DOI 10.17487/RFC8020,
              November 2016, <http://www.rfc-editor.org/info/rfc8020>.

   [ROOT-SERVERS]
              "Root Server Technical Operations Assn",
              <http://www.root-servers.org/>.

Appendix A.  Detailed Implementation Notes

   o  Previously, cached negative responses were indexed by QNAME,
      QCLASS, QTYPE, and the setting of the CD bit (see RFC 4035,
      Section 4.7), and only queries matching the index key would be
      answered from the cache.  With aggressive negative caching, the
      validator, in addition to checking to see if the answer is in its
      cache before sending a query, checks to see whether any cached and
      validated NSEC record denies the existence of the sought
      record(s).  Using aggressive negative caching, a validator will
      not make queries for any name covered by a cached and validated
      NSEC record.  Furthermore, a validator answering queries from
      clients will synthesize a negative answer (or NODATA response)
      whenever it has an applicable validated NSEC in its cache unless
      the CD bit was set on the incoming query.  (Imported from
      Section 6 of [RFC5074].)

   o  Implementing aggressive negative caching suggests that a validator
      will need to build an ordered data structure of NSEC and NSEC3
      records for each signer domain name of NSEC/NSEC3 records in order
      to efficiently find covering NSEC/NSEC3 records.  Call the table
      as "NSEC_TABLE".  (Imported from Section 6.1 of [RFC5074] and
      expanded.)

   o  The aggressive negative caching may be inserted at the cache
      lookup part of the recursive resolvers.

   o  If errors happen in an aggressive negative caching algorithm,
      resolvers MUST fall back to resolve the query as usual.  "Resolve
      the query as usual" means that the resolver must process the query
      as though it does not implement aggressive negative caching.

Appendix B.  Procedure for Determining ENT vs. NXDOMAIN with NSEC

   This procedure outlines how to determine if a given name does not
   exist, or is an ENT (empty non-terminal; see [RFC5155], Section 1.3)
   with NSEC.

   If the NSEC record has not been verified as secure, discard it.

   If the given name sorts before or matches the NSEC owner name,
   discard it as it does not prove the NXDOMAIN or ENT.

   If the given name is a subdomain of the NSEC owner name and the NS
   bit is present and the SOA bit is absent, then discard the NSEC as it
   is from a parent zone.

   If the next domain name sorts after the NSEC owner name and the given
   name sorts after or matches next domain name, then discard the NSEC
   record as it does not prove the NXDOMAIN or ENT.

   If the next domain name sorts before or matches the NSEC owner name
   and the given name is not a subdomain of the next domain name, then
   discard the NSEC as it does not prove the NXDOMAIN or ENT.

   You now have an NSEC record that proves the NXDOMAIN or ENT.

   If the next domain name is a subdomain of the given name, you have an
   ENT.  Otherwise, you have an NXDOMAIN.

Acknowledgments

   The authors gratefully acknowledge DNSSEC Lookaside Validation (DLV)
   [RFC5074] author Samuel Weiler and the Unbound developers.

   Thanks to Mark Andrews for providing the helpful notes for
   implementors provided in Appendix B.

   The authors would like to specifically thank Stephane Bortzmeyer (for
   standing next to and helping edit), Ralph Dolmans, Tony Finch, Tatuya
   JINMEI for extensive review and comments, and also Mark Andrews,
   Casey Deccio, Alexander Dupuy, Olafur Gudmundsson, Bob Harold, Shumon
   Huque, John Levine, Pieter Lexis, Matthijs Mekking (who even sent
   pull requests!), and Ondrej Sury.

Authors' Addresses

   Kazunori Fujiwara
   Japan Registry Services Co., Ltd.
   Chiyoda First Bldg. East 13F, 3-8-1 Nishi-Kanda
   Chiyoda-ku, Tokyo  101-0065
   Japan

   Phone: +81 3 5215 8451
   Email: fujiwara@jprs.co.jp

   Akira Kato
   Keio University/WIDE Project
   Graduate School of Media Design, 4-1-1 Hiyoshi
   Kohoku, Yokohama  223-8526
   Japan

   Phone: +81 45 564 2490
   Email: kato@wide.ad.jp

   Warren Kumari
   Google
   1600 Amphitheatre Parkway
   Mountain View, CA  94043
   United States of America

   Email: warren@kumari.net