]> Google LLC

bemasc@google.com

BT

chris.box@bt.com

General dprive Internet-Draft This draft describes how the Discovery of Designated Resolvers (DDR) standard interacts with legacy DNS forwarders, including potential incompatibilities and relevant mitigations. Discussion Venues Discussion of this document takes place on the mailing list (add@ietf.org), which is archived at . Source for this draft and an issue tracker can be found at .

Conventions and Definitions Private IP Address - Any IP address reserved for loopback , link-local , private , local , or Carrier-Grade NAT use. Legacy DNS Forwarder - An apparent DNS resolver, known to the client only by a private IP address, that forwards the client's queries to an upstream resolver, and has not been updated with any knowledge of DDR. Cross-Forwarder Upgrade - Establishment of a direct, encrypted connection between the client and the upstream resolver.

Introduction

Background The Discovery of Designated Resolvers specification describes a mechanism for clients to learn about the encrypted protocols supported by a DNS server. It also describes a conservative client validation policy that has strong security properties and is unlikely to create compatibility problems. On the topic of client validation of encrypted DNS transports, the DDR specification says:

• If the IP address of a Designated Resolver differs from that of an Unencrypted Resolver, clients MUST validate that the IP address of the Unencrypted Resolver is covered by the SubjectAlternativeName of the Encrypted Resolver's TLS certificate

As TLS certificates cannot cover private IP addresses, this prevents clients that are behind a legacy DNS forwarder from connecting directly to the upstream resolver ("cross-forwarder upgrade"). Recent estimates suggest that a large fraction, perhaps a majority, of residential internet users in the United States and Europe rely on local DNS forwarders that are not compatible with DDR.

Scope This informational document describes the interaction between DDR and legacy DNS forwarders. It discusses possible client policies, problems that might arise, and relevant mitigations. DNS forwarders and resolvers that are implemented with awareness of DDR are out of scope, as they are not affected by this discussion (although see Security Considerations, ). IPv6-only networks whose default DNS server has a Global Unicast Address are out of scope, even if this server is actually a simple forwarder. If the DNS server does not use a private IP address, it is not a "legacy DNS forwarder" under this draft's definition.

Relaxed Validation client policy We define a "relaxed validation" client policy as a client behavior that removes the certificate validation requirement when the Unencrypted Resolver is identified by a private IP address, regardless of the Designated Resolver's IP address. Instead, under this condition, the client connects using the Opportunistic Privacy Profile of encrypted DNS (). The Opportunistic Privacy Profile is a broad category, including clients that "might or might not validate" the TLS certificate chain even though there is no authentication identity for the server. This kind of validation can be valuable when combined with a reputation system or a user approval step (see and ). This client policy is otherwise identical to the one described in .

Naturally compatible behaviors The following system behaviors are naturally compatible with relaxed validation.

Compatible behaviors in the local network

Malware and threat domain filtering Certain DNS forwarders block access to domains associated with malware and other threats. Such threats rely on frequently changing domains, so these forwarders necessarily maintain an actively curated list of domains to block. To ensure that this service is not lost due to a cross-forwarder upgrade, the maintainers can simply add "resolver.arpa" to the list. This pattern has been deployed by Mozilla, with the domain "use-application-dns.net" .

Service category restrictions Certain DNS forwarders may block access to domains based on the category of service provided by those domains, e.g. domains hosting services that are not appropriate for a work or school environment. As in the previous section, this requires an actively curated list of domains, because the set of domains that offer a given type of service is constantly changing. An actively managed blocking list can easily be revised to include "resolver.arpa".

Time of use restrictions Certain networks may impose restrictions on the time or duration of use by certain users. This behavior is necessarily implemented below the DNS layer, because DNS-based blocking would be ineffective due to stub resolver caching, so it is not affected by changes in the DNS resolver.

Upstream resolver services The forwarder's upstream resolver might provide additional services, such as filtering. These services are generally independent of cross-forwarder upgrade, and hence naturally compatible. In special cases where the upstream resolver requires cooperation from a legacy forwarder (e.g. for marking certain queries), one solution is for the upstream resolver to choose not to deploy DDR until all cooperating forwarders have been upgraded. Alternatively, each legacy forwarder can block "resolver.arpa" as described above.

Privacy Considerations The conservative validation policy results in no encryption when a legacy DNS forwarder is present. This leaves the user's query activity vulnerable to passive monitoring , either on the local network or between the user and the upstream resolver. The relaxed validation policy allows the use of encrypted transport in these configurations, reducing exposure to a passive surveillance adversary.

Security Considerations When the client uses the conservative validation policy described in , and a DDR-enabled resolver is identified by a private IP address, the client can establish a secure DDR connection only in the absence of an active attacker. An on-path attacker can impersonate the resolver and intercept all queries, by preventing the DDR upgrade or advertising their own DDR endpoint. These basic security properties also apply if the client uses the relaxed validation policy described in . Nonetheless, there are some subtle but important differences in the security properties of these two policies.

Transient attackers With the conservative validation policy, a transient on-path attacker can only intercept queries for the duration of their active presence on the network, because the client will only send queries to the original (private) server IP address. With the relaxed validation behavior, a transient on-path attacker could implant a long-lived DDR response in the client's cache, directing its queries to an attacker-controlled server on the public internet. This would allow the attack to continue long after the attacker has left the network. Solving or mitigating this attack is of great importance for the user's security.

Solution: DNR This attack does not apply if the client and network implement support for Discovery of Network-designated Resolvers, as that mechanism takes precedence over DDR (see ).

Mitigation: Frequent refresh The client can choose to refresh the DDR record arbitrarily frequently, e.g. by limiting the TTL. For example, by limiting the TTL to 5 minutes, a client could ensure that any attacker can continue to monitor queries for at most 5 minutes after they have left the local network.

Mitigation: Resolver reputation A relaxed-validation client might choose to accept a potential cross-forwarder upgrade only if the designated encrypted resolver has sufficient reputation, according to some proprietary reputation scheme (e.g. a locally stored list of respectable resolvers). This limits the ability of a DDR forgery attack to cause harm. Major DoH client implementations already include lists of known resolvers .

Forensic logging

Network-layer logging With the conservative validation policy, a random sample of IP packets is likely sufficient for manual retrospective detection of an active attack. With the relaxed validation policy, forensic logs must capture a specific packet (the attacker's DDR designation response) to enable retrospective detection.

Mitigation: Log all DDR responses Network-layer forensic logs that are not integrated with the resolver can enable detection of these attacks by logging all DDR responses, or more generally all DNS responses. This makes retrospective attack detection straightforward, as the attacker's DDR response will indicate an unexpected server.

DNS-layer logging DNS-layer forensic logging conducted by a legacy DNS forwarder would be lost in a cross-forwarder upgrade.

Solution: Respond for resolver.arpa Forwarders that want to observe all queries from relaxed validation clients will have to synthesize their own response for resolver.arpa, either implementing DDR or disabling it.

Compatibility Considerations Using DDR with legacy DNS forwarders also raises several potential concerns related to loss of existing network services.

Split-horizon namespaces Some network resolvers contain additional names that are not resolvable in the global DNS. If these local resolvers are also legacy DNS forwarders, a client that performs a cross-forwarder upgrade might lose access to these local names.

Mitigation: NXDOMAIN Fallback In "NXDOMAIN Fallback", the client repeats a query to the unencrypted resolver if the encrypted resolver returns NXDOMAIN. This allows the resolution of local names, provided they do not collide with globally resolvable names (as required by ). This is similar to the fallback behavior currently deployed in Mozilla Firefox . NXDOMAIN Fallback results in slight changes to the security and privacy properties of encrypted DNS. Queries for nonexistent names no longer have protection against a local passive adversary, and local names are revealed to the upstream resolver. NXDOMAIN Fallback is only applicable when a legacy DNS forwarder might be present, i.e. the unencrypted resolver has a private IP address, and the encrypted resolver has a different IP address. In the other DDR configurations, any local names are expected to resolve similarly on both resolvers.

Interposable domains An "interposable domain" is a domain whose owner deliberately allows resolvers to forge certain responses. This arrangement is most common for search engines, which often support a configuration where resolvers forge a CNAME record to direct all clients to a child-appropriate instance of the search engine . Future deployments of interposable domains can instruct administrators to enable or disable DDR when adding the forged record, but forged records in legacy DNS forwarders could be lost due to a cross-forwarder upgrade.

Mitigation: Exemption list There are a small number of pre-existing interposable domains, largely of interest only to web browsers. Clients can maintain a list of relevant interposable domains and resolve them only via the network's resolver.

Caching Some legacy DNS forwarders also provide a shared cache for all network users. Cross-forwarder upgrades will bypass this cache, resulting in slower DNS resolution.

Mitigation: Stub caches Clients can compensate partially for any loss of shared caching by implementing local DNS caches. This mitigation is already widely deployed in browsers and operating systems.

General mitigation: User controls For these and other compatibility concerns, a possible mitigation is to provide users or administrators with the ability to control whether DDR is used with legacy forwarders. For example, this control could be provided via a general preference, or via a notification upon discovering a new upstream resolver.

Informative References n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. This RFC is an official specification for the Internet community. It incorporates by reference, amends, corrects, and supplements the primary protocol standards documents relating to hosts. [STANDARDS-TRACK] To participate in wide-area IP networking, a host needs to be configured with IP addresses for its interfaces, either manually by the user or automatically from a source on the network such as a Dynamic Host Configuration Protocol (DHCP) server. Unfortunately, such address configuration information may not always be available. It is therefore beneficial for a host to be able to depend on a useful subset of IP networking functions even when no address configuration is available. This document describes how a host may automatically configure an interface with an IPv4 address within the 169.254/16 prefix that is valid for communication with other devices connected to the same physical (or logical) link. IPv4 Link-Local addresses are not suitable for communication with devices not directly connected to the same physical (or logical) link, and are only used where stable, routable addresses are not available (such as on ad hoc or isolated networks). This document does not recommend that IPv4 Link-Local addresses and routable addresses be configured simultaneously on the same interface. [STANDARDS-TRACK] This document describes address allocation for private internets. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. This document defines an IPv6 unicast address format that is globally unique and is intended for local communications, usually inside of a site. These addresses are not expected to be routable on the global Internet. [STANDARDS-TRACK] This document requests the allocation of an IPv4 /10 address block to be used as Shared Address Space to accommodate the needs of Carrier- Grade NAT (CGN) devices. It is anticipated that Service Providers will use this Shared Address Space to number the interfaces that connect CGN devices to Customer Premises Equipment (CPE). Shared Address Space is distinct from RFC 1918 private address space because it is intended for use on Service Provider networks. However, it may be used in a manner similar to RFC 1918 private address space on routing equipment that is able to do address translation across router interfaces when the addresses are identical on two different interfaces. Details are provided in the text of this document. This document details the allocation of an additional special-use IPv4 address block and updates RFC 5735. This memo documents an Internet Best Current Practice. Apple Inc. Apple Inc. Cloudflare Fastly Microsoft This document defines Discovery of Designated Resolvers (DDR), a mechanism for DNS clients to use DNS records to discover a resolver's encrypted DNS configuration. This mechanism can be used to move from unencrypted DNS to encrypted DNS when only the IP address of a resolver is known. This mechanism is designed to be limited to cases where unencrypted resolvers and their designated resolvers are operated by the same entity or cooperating entities. It can also be used to discover support for encrypted DNS protocols when the name of an encrypted resolver is known. This document describes the use of Transport Layer Security (TLS) to provide privacy for DNS. Encryption provided by TLS eliminates opportunities for eavesdropping and on-path tampering with DNS queries in the network, such as discussed in RFC 7626. In addition, this document specifies two usage profiles for DNS over TLS and provides advice on performance considerations to minimize overhead from using TCP and TLS with DNS. This document focuses on securing stub-to-recursive traffic, as per the charter of the DPRIVE Working Group. It does not prevent future applications of the protocol to recursive-to-authoritative traffic. Pervasive monitoring is a technical attack that should be mitigated in the design of IETF protocols, where possible. Orange McAfee, Inc. Citrix Systems, Inc. Open-Xchange Microsoft The document specifies new DHCP and IPv6 Router Advertisement options to discover encrypted DNS servers (e.g., DNS-over-HTTPS, DNS-over- TLS, DNS-over-QUIC). Particularly, it allows to learn an authentication domain name together with a list of IP addresses and a set of service parameters to reach such encrypted DNS servers. Internet Architecture Board This document discusses the existence of a globally unique public name space in the Internet called the DNS (Domain Name System). This name space is a hierarchical name space derived from a single, globally unique root. It is a technical constraint inherent in the design of the DNS. One root must be supported by a set of coordinated root servers administered by a unique naming authority. It is not technically feasible for there to be more than one root in the public DNS. This memo provides information for the Internet community.

Acknowledgments Thanks to Anthony Lieuallen and Eric Orth for early reviews.