DNS Security Extension Clarification on Zone Status

Status of this Memo

This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.

Copyright Notice

Copyright © The Internet Society (2001). All Rights Reserved.

Abstract

The definition of a secured zone is presented, clarifying and updating sections of RFC 2535. RFC 2535 defines a zone to be secured based on a per algorithm basis, e.g., a zone can be secured with RSA keys, and not secured with DSA keys. This document changes this to define a zone to be secured or not secured regardless of the key algorithm used (or not used). To further simplify the determination of a zone's status, "experimentally secure" status is deprecated.

1 Introduction

Whether a DNS zone is "secured" or not is a question asked in at least four contexts. A zone administrator asks the question when configuring a zone to use DNSSEC. A dynamic update server asks the question when an update request arrives, which may require DNSSEC processing. A delegating zone asks the question of a child zone when the parent enters data indicating the status the child. A resolver asks the question upon receipt of data belonging to the zone.

1.1 When a Zone's Status is Important

A zone administrator needs to be able to determine what steps are needed to make the zone as secure as it can be. Realizing that due to the distributed nature of DNS and its administration, any single zone is at the mercy of other zones when it comes to the appearance of security. This document will define what makes a zone qualify as secure.

A name server performing dynamic updates needs to know whether a zone being updated is to have signatures added to the updated data, NXT records applied, and other required processing. In this case, it is conceivable that the name server is configured with the knowledge, but being able to determine the status of a zone by examining the data is a desirable alternative to configuration parameters.

A delegating zone is required to indicate whether a child zone is secured. The reason for this requirement lies in the way in which a resolver makes its own determination about a zone (next paragraph). To shorten a long story, a parent needs to know whether a child should be considered secured. This is a two part question. Under what circumstances does a parent consider a child zone to be secure, and how does a parent know if the child conforms?

A resolver needs to know if a zone is secured when the resolver is processing data from the zone. Ultimately, a resolver needs to know whether or not to expect a usable signature covering the data. How this determination is done is out of the scope of this document, except that, in some cases, the resolver will need to contact the parent of the zone to see if the parent states that the child is secured.

1.2 Islands of Security

The goal of DNSSEC is to have each zone secured, from the root zone and the top-level domains down the hierarchy to the leaf zones. Transitioning from an unsecured DNS, as we have now, to a fully secured - or "as much as will be secured" - tree will take some time. During this time, DNSSEC will be applied in various locations in the tree, not necessarily "top down."

For example, at a particular instant, the root zone and the "test." TLD might be secured, but region1.test. might not be. (For reference, let's assume that region2.test. is secured.) However, subarea1.region1.test. may have gone through the process of becoming secured, along with its delegations. The dilemma here is that subarea1 cannot get its zone keys properly signed as its parent zone, region1, is not secured.

The colloquial phrase describing the collection of contiguous secured zones at or below subarea1.region1.test. is an "island of security." The only way in which a DNSSEC resolver will come to trust any data from this island is if the resolver is pre-configured with the zone key(s) for subarea1.region1.test., i.e., the root of the island of security. Other resolvers (not so configured) will recognize this island as unsecured.

An island of security begins with one zone whose public key is pre- configured in resolvers. Within this island are subzones which are also secured. The "bottom" of the island is defined by delegations to unsecured zones. One island may also be on top of another - meaning that there is at least one unsecured zone between the bottom of the upper island and the root of the lower secured island.

Although both subarea1.region1.test. and region2.test. have both been properly brought to a secured state by the administering staff, only the latter of the two is actually "globally" secured - in the sense that all DNSSEC resolvers can and will verify its data. The former, subarea1, will be seen as secured by a subset of those resolvers, just those appropriately configured. This document refers to such zones as being "locally" secured.

In RFC 2535, there is a provision for "certification authorities," entities that will sign public keys for zones such as subarea1. There is another document, [RFC3008], that restricts this activity. Regardless of the other document, resolvers would still need proper configuration to be able to use the certification authority to verify the data for the subarea1 island.

1.2.1 Determining the closest security root

Given a domain, in order to determine whether it is secure or not, the first step is to determine the closest security root. The closest security root is the top of an island of security whose name has the most matching (in order from the root) right-most labels to the given domain.

For example, given a name "sub.domain.testing.signed.exp.test.", and given the secure roots "exp.test.", "testing.signed.exp.test." and "not-the-same.xy.", the middle one is the closest. The first secure root shares 2 labels, the middle 4, and the last 0.

The reason why the closest is desired is to eliminate false senses of insecurity because of a NULL key. Continuing with the example, the reason both "testing..." and "exp.test." are listed as secure root is presumably because "signed.exp.test." is unsecured (has a NULL key). If we started to descend from "exp.test." to our given domain (sub...), we would encounter a NULL key and conclude that sub... was unsigned. However, if we descend from "testing..." and find keys "domain...." then we can conclude that "sub..." is secured.

Note that this example assumes one-label deep zones, and assumes that we do not configure overlapping islands of security. To be clear, the definition given should exclude "short.xy.test." from being a closest security root for "short.xy." even though 2 labels match.

Overlapping islands of security introduce no conceptually interesting ideas and do not impact the protocol in anyway. However, protocol implementers are advised to make sure their code is not thrown for a loop by overlaps. Overlaps are sure to be configuration problems as islands of security grow to encompass larger regions of the name space.

1.3 Parent Statement of Child Security

   In 1.1 of this document, there is the comment "the parent states that
   the child is secured."  This has caused quite a bit of confusion.

The need to have the parent "state" the status of a child is derived from the following observation. If you are looking to see if an answer is secured, that it comes from an "island of security" and is properly signed, you must begin at the (appropriate) root of the island of security.

To find the answer you are inspecting, you may have to descend through zones within the island of security. Beginning with the trusted root of the island, you descend into the next zone down. As you trust the upper zone, you need to get data from it about the next zone down, otherwise there is a vulnerable point in which a zone can be hijacked. When or if you reach a point of traversing from a secured zone to an unsecured zone, you have left the island of security and should conclude that the answer is unsecured.

However, in RFC 2535, section 2.3.4, these words seem to conflict with the need to have the parent "state" something about a child:

There MUST be a zone KEY RR, signed by its superzone, for every subzone if the superzone is secure. This will normally appear in the subzone and may also be included in the superzone. But, in the case of an unsecured subzone which can not or will not be modified to add any security RRs, a KEY declaring the subzone to be unsecured MUST appear with the superzone signature in the superzone, if the superzone is secure.

   The confusion here is that in RFC 2535, a secured parent states that
   a child is secured by SAYING NOTHING ("may also be" as opposed to
   "MUST also be").  This is counter intuitive, the fact that an absence
   of data means something is "secured."  This notion, while acceptable
   in a theoretic setting has met with some discomfort in an operation
   setting.  However, the use of "silence" to state something does
   indeed work in this case, so there hasn't been sufficient need
   demonstrated to change the definition.

1.4 Impact on RFC 2535

This document updates sections of RFC 2535. The definition of a secured zone is an update to section 3.4 of the RFC. Section 3.4 is updated to eliminate the definition of experimental keys and illustrate a way to still achieve the functionality they were designed to provide. Section 3.1.3 is updated by the specifying the value of the protocol octet in a zone key.

1.5 "MUST" and other key words

The key words "MUST", "REQUIRED", "SHOULD", "RECOMMENDED", and "MAY" in this document are to be interpreted as described in [RFC 2119]. Currently, only "MUST" is used in this document.

2 Status of a Zone

In this section, rules governing a zone's DNSSEC status are presented. There are three levels of security defined: global, local, and unsecured. A zone is globally secure when it complies with the strictest set of DNSSEC processing rules. A zone is locally secured when it is configured in such a way that only resolvers that are appropriately configured see the zone as secured. All other zones are unsecured.

Note: there currently is no document completely defining DNSSEC verification rules. For the purposes of this document, the strictest rules are assumed to state that the verification chain of zone keys parallels the delegation tree up to the root zone. (See 2.b below.) This is not intended to disallow alternate verification paths, just to establish a baseline definition.

To avoid repetition in the rules below, the following terms are defined.

2.a Zone signing KEY RR - A KEY RR whose flag field has the value 01 for name type (indicating a zone key) and either value 00 or value 01 for key type (indicating a key permitted to authenticate data). (See RFC 2535, section 3.1.2). The KEY RR also has a protocol octet value of DNSSEC (3) or ALL (255).

The definition updates RFC 2535's definition of a zone key. The requirement that the protocol field be either DNSSEC or ALL is a new requirement (a change to section 3.1.3.)

2.b On-tree Validation - The authorization model in which only the parent zone is recognized to supply a DNSSEC-meaningful signature that is used by a resolver to build a chain of trust from the child's keys to a recognized root of security. The term "on-tree" refers to following the DNS domain hierarchy (upwards) to reach a trusted key, presumably the root key if no other key is available. The term "validation" refers to the digital signature by the parent to prove the integrity, authentication and authorization of the child's key to sign the child's zone data.

2.c Off-tree Validation - Any authorization model that permits domain names other than the parent's to provide a signature over a child's zone keys that will enable a resolver to trust the keys.

2.1 Globally Secured

A globally secured zone, in a nutshell, is a zone that uses only mandatory to implement algorithms (RFC 2535, section 3.2) and relies on a key certification chain that parallels the delegation tree (on- tree validation). Globally secured zones are defined by the following rules.

2.1.a. The zone's apex MUST have a KEY RR set. There MUST be at least one zone signing KEY RR (2.a) of a mandatory to implement algorithm in the set.

2.1.b. The zone's apex KEY RR set MUST be signed by a private key belonging to the parent zone. The private key's public companion MUST be a zone signing KEY RR (2.a) of a mandatory to implement algorithm and owned by the parent's apex.

If a zone cannot get a conforming signature from the parent zone, the child zone cannot be considered globally secured. The only exception to this is the root zone, for which there is no parent zone.

2.1.c. NXT records MUST be deployed throughout the zone. (Clarifies RFC 2535, section 2.3.2.) Note: there is some operational discomfort with the current NXT record. This requirement is open to modification when two things happen. First, an alternate mechanism to the NXT is defined and second, a means by which a zone can indicate that it is using an alternate method.

2.1.d. Each RR set that qualifies for zone membership MUST be signed by a key that is in the apex's KEY RR set and is a zone signing KEY RR (2.a) of a mandatory to implement algorithm. (Updates 2535, section 2.3.1.)

Mentioned earlier, the root zone is a special case. The root zone will be considered to be globally secured provided that if conforms to the rules for locally secured, with the exception that rule 2.1.a. be also met (mandatory to implement requirement).

2.2 Locally Secured

The term "locally" stems from the likely hood that the only resolvers to be configured for a particular zone will be resolvers "local" to an organization.

A locally secured zone is a zone that complies with rules like those for a globally secured zone with the following exceptions. The signing keys may be of an algorithm that is not mandatory to implement and/or the verification of the zone keys in use may rely on a verification chain that is not parallel to the delegation tree (off-tree validation).

2.2.a. The zone's apex MUST have a KEY RR set. There MUST be at least one zone signing KEY RR (2.a) in the set.

2.2.b. The zone's apex KEY RR set MUST be signed by a private key and one of the following two subclauses MUST hold true.

2.2.b.1 The private key's public companion MUST be pre-configured in all the resolvers of interest.

2.2.b.2 The private key's public companion MUST be a zone signing KEY RR (2.a) authorized to provide validation of the zone's apex KEY RR set, as recognized by resolvers of interest.

The previous sentence is trying to convey the notion of using a trusted third party to provide validation of keys. If the domain name owning the validating key is not the parent zone, the domain name must represent someone the resolver trusts to provide validation.

2.2.c. NXT records MUST be deployed throughout the zone. Note: see the discussion following 2.1.c.

2.2.d. Each RR set that qualifies for zone membership MUST be signed by a key that is in the apex's KEY RR set and is a zone signing KEY RR (2.a). (Updates 2535, section 2.3.1.)

2.3 Unsecured

All other zones qualify as unsecured. This includes zones that are designed to be experimentally secure, as defined in a later section on that topic.

2.4 Wrap up

The designation of globally secured, locally secured, and unsecured are merely labels to apply to zones, based on their contents. Resolvers, when determining whether a signature is expected or not, will only see a zone as secured or unsecured.

Resolvers that follow the most restrictive DNSSEC verification rules will only see globally secured zones as secured, and all others as unsecured, including zones which are locally secured. Resolvers that are not as restrictive, such as those that implement algorithms in addition to the mandatory to implement algorithms, will see some locally secured zones as secured.

The intent of the labels "global" and "local" is to identify the specific attributes of a zone. The words are chosen to assist in the writing of a document recommending the actions a zone administrator take in making use of the DNS security extensions. The words are explicitly not intended to convey a state of compliance with DNS security standards.

3 Experimental Status

The purpose of an experimentally secured zone is to facilitate the migration from an unsecured zone to a secured zone. This distinction is dropped.

The objective of facilitating the migration can be achieved without a special designation of an experimentally secure status. Experimentally secured is a special case of locally secured. A zone administrator can achieve this by publishing a zone with signatures and configuring a set of test resolvers with the corresponding public keys. Even if the public key is published in a KEY RR, as long as there is no parent signature, the resolvers will need some pre- configuration to know to process the signatures. This allows a zone to be secured with in the sphere of the experiment, yet still be registered as unsecured in the general Internet.

4 IANA Considerations

This document does not request any action from an assigned number authority nor recommends any actions.

5 Security Considerations

Without a means to enforce compliance with specified protocols or recommended actions, declaring a DNS zone to be "completely" secured is impossible. Even if, assuming an omnipotent view of DNS, one can declare a zone to be properly configured for security, and all of the zones up to the root too, a misbehaving resolver could be duped into believing bad data. If a zone and resolver comply, a non-compliant or subverted parent could interrupt operations. The best that can be hoped for is that all parties are prepared to be judged secure and that security incidents can be traced to the cause in short order.

6 Acknowledgements

The need to refine the definition of a secured zone has become apparent through the efforts of the participants at two DNSSEC workshops, sponsored by the NIC-SE (.se registrar), CAIRN (a DARPA- funded research network), and other workshops. Further discussions leading to the document include Olafur Gudmundsson, Russ Mundy, Robert Watson, and Brian Wellington. Roy Arends, Ted Lindgreen and others have contributed significant input via the namedroppers mailing list.

7 References

   [RFC1034] Mockapetris, P., "Domain Names - Concepts and Facilities",
             STD 13, RFC 1034, November 1987.
   
   [RFC1035] Mockapetris, P., "Domain Names - Implementation and
             Specification", STD 13, RFC 1035, November 1987.
   
   [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
             Requirement Levels", BCP 14, RFC 2119, March 1997.
   
   [RFC2136] Vixie, P., (Ed.), Thomson, S., Rekhter, Y. and J. Bound,
             "Dynamic Updates in the Domain Name System", RFC 2136,
             April 1997.
   
   [RFC2535] Eastlake, D., "Domain Name System Security Extensions", RFC
             2535, March 1999.
   
   [RFC3007] Wellington, B., "Simple Secure Domain Name System (DNS)
             Dynamic Update", RFC 3007, November 2000.
   
   [RFC3008] Wellington, B., "Domain Name System Security (DNSSEC)
             Signing Authority", RFC 3008, November 2000.

10 Author's Address

Edward Lewis
NAI Labs
3060 Washington Road Glenwood
MD 21738

   Phone: +1 443 259 2352
   EMail: lewis@tislabs.com

11 Full Copyright Statement

Copyright © The Internet Society (2001). All Rights Reserved.

This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English.

The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns.

This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Acknowledgement

Funding for the RFC Editor function is currently provided by the Internet Society.