RFC 9018
Interoperable Domain Name System (DNS) Server Cookies
RFC 9018: Interoperable Domain Name System (DNS) Server Cookies 🦋
Okay bestie, so this is RFC 9018 🎀. It's an IETF document (Internet Engineering Task Force) on the Standards Track, which means it's a real Internet Standard-track spec 👑. It came out in April 2021 and its ISSN is 2070-1721. The authors are O. Sury (Internet Systems Consortium), W. Toorop (NLnet Labs), D. Eastlake 3rd (Futurewei Technologies), and M. Andrews (Internet Systems Consortium) 💜. It updates RFC 7873, which is the original DNS Cookies spec 🍓. It doesn't obsolete anything.
Abstract ✨
So DNS Cookies, the stuff from RFC 7873, are this lightweight transaction security thing for DNS 🧚. They give limited protection to DNS servers and clients against denial-of-service amplification, forgery, and cache-poisoning attacks from off-path attackers 🔥.
This document updates RFC 7873 with precise instructions for making Server Cookies, so an anycast server set running different implementations can still work with standard clients 🫶. It also has suggestions for building Client Cookies in a privacy-preserving way, plus tips on how to update a Server Secret 🌈. There's an IANA registry listing the methods and pseudorandom functions that are suitable for making DNS Server Cookies, and the method described here is the first entry and, at publication time, the only one 💅.
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 went through public review and got approved for publication by the Internet Engineering Steering Group (IESG). More info on Internet Standards is in Section 2 of RFC 7841 ✨.
You can find info about the current status of this document, any errata, and how to send feedback at https://www.rfc-editor.org/info/rfc9018 🌸.
Copyright Notice 💜
Copyright (c) 2021 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) as they stand on the publication date of this document. Please read these carefully, because they explain your rights and restrictions for this document. Any Code Components pulled out of this document must include Simplified BSD License text, as described in Section 4.e of the Trust Legal Provisions, and they come without warranty, as described in the Simplified BSD License 🦋.
Table of Contents 🌈
1. Introduction
1.1. Terminology and Definitions
2. Changes to RFC 7873
3. Constructing a Client Cookie
4. Constructing a Server Cookie
4.1. The Version Sub-Field
4.2. The Reserved Sub-Field
4.3. The Timestamp Sub-Field
4.4. The Hash Sub-Field
5. Updating the Server Secret
6. Cookie Algorithms
7. IANA Considerations
8. Security and Privacy Considerations
8.1. Client Cookie Construction
8.2. Server Cookie Construction
9. References
9.1. Normative References
9.2. Informative References
Appendix A. Test Vectors
A.1. Learning a New Server Cookie
A.2. The Same Client Learning a Renewed (Fresh) Server Cookie
A.3. Another Client Learning a Renewed Server Cookie
A.4. IPv6 Query with Rolled Over Secret
Appendix B. Implementation Status
Acknowledgements
Authors' Addresses1. Introduction 🌸
Okay so DNS Cookies, as specified in [RFC7873], are a lightweight DNS transaction security mechanism 🧁. They give limited protection to DNS servers and clients against denial-of-service amplification, forgery, and cache-poisoning attacks by off-path attackers 💀. This document specifies a way to produce interoperable cookies, so an anycast server set running diverse implementations can be set up to work with standard clients without drama 🫶. Also, single-implementation or non-anycast services can benefit from a well-studied standardized algorithm, whose behavioral and security characteristics are more widely known 👑.
The threats considered for DNS Cookies, and the properties of the other DNS Security features besides DNS Cookies, are covered in [RFC7873] ✨.
In Section 6 of [RFC7873], to keep things simple, it says "RECOMMENDED that the same Server Secret be used by each DNS server in a set of anycast servers." 🙈 But how exactly a Server Cookie gets calculated from that Server Secret is left up to each implementation 🤷♀️.
That guidance led to a total gallimaufry of DNS Cookie implementations, each one calculating the Server Cookie in its own way 😭. So DNS Cookies are impractical to deploy on multi-vendor anycast networks. Even when all the DNS software shares the same secret, as RECOMMENDED in Section 6 of [RFC7873], a Server Cookie built by one implementation generally can't be validated by another 💔.
DNS client (resolver) Cookies don't need to be interoperable across implementations. Each client only needs to recognize its own cookies 🧚. That said, this document does include recommendations for building Client Cookies in a client-protecting way 🦋.
1.1. Terminology and Definitions 💅
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 📖.
Note: "IP address" is used here as a length-independent term that covers both IPv4 and IPv6 addresses 🌈.
2. Changes to RFC 7873 🍓
Appendices A.1 and B.1 of [RFC7873] give example "simple" algorithms for computing Client and Server Cookies, respectively. These algorithms MUST NOT be used, because the cookies they produce are too weak by modern security standards 🚫💀.
Appendix B.2 of [RFC7873] gives an example "more complex" server algorithm. That algorithm gets replaced by the interoperable specification in Section 4 of this document, and Server Cookie implementations MUST use that specification 👑✨.
This document has suggestions on Client Cookie construction in Section 3. The previous example in Appendix A.2 of [RFC7873] is NOT RECOMMENDED 🙅♀️.
3. Constructing a Client Cookie 🎀
The Client Cookie acts as an identifier for a given client and its IP address, and it needs to be unguessable 🔐. To give minimal authentication of the targeted server, a client MUST use a different Client Cookie for each different Server IP address 🦋. This makes it harder for a server to spoof answers for other DNS servers 🫣. The Client Cookie SHOULD have 64 bits of entropy 💜.
When a server does not support DNS Cookies, the client MUST NOT send the same Client Cookie to that same server again 🙅♀️. Instead, it is recommended that the client does not send a Client Cookie to that server for a certain period (for example, five minutes) before it retries with a new Client Cookie ⏳.
When a server does support DNS Cookies, the client should store the Client Cookie alongside the Server Cookie it registered for that server 🧁.
Unless the Client IP address changes, there's no real need to change the Client Cookie often 🌸. It's reasonable to change it only if it has been compromised, or after a relatively long implementation-defined period of time. That period should be no longer than a year, and either way, Client Cookies aren't expected to survive a program restart 🔄.
Here's the ABNF, which is just the 64 bits of entropy thing:
Client-Cookie = 64 bits of entropyPreviously, the recommended algorithm for computing the Client Cookie included the Client IP address as an input to a hashing function 🧐. But when DNS vendors were implementing DNS Cookies, several of them found it impractical to include the Client IP, because the Client Cookie is typically computed before the Client IP address is known 😩. So the requirement to put the Client IP address in as an input got removed ✂️.
That said, for privacy reasons, clients MUST NOT reuse a Client or Server Cookie after the Client IP address has changed 🕵️♀️. This keeps devices from being tracked across links, and it means nobody can get around IPv6 Privacy Extensions [RFC8981] 🛡️✨.
One way to meet this non-reuse requirement is to register the Client IP address alongside the Server Cookie when the client receives that Server Cookie 📝. In later queries to the server using that Server Cookie, the socket MUST be bound to the Client IP address that was also used, and registered, when the Server Cookie was received. If the binding fails, a new Client Cookie MUST then result 💅🔥.
4. Constructing a Server Cookie ✨
okay bestie, so the Server Cookie is basically a Message Authentication Code (MAC) 🔐💖 when it shows up in a COOKIE option in a request, it's meant to weakly assure the server that the request came from a client that is both at the request's source IP address AND using the Client Cookie in that option. that assurance comes from the Server Cookie the server (or any other server in the anycast set) gave that client in an earlier response, which then appears as the Server Cookie field in the weakly authenticated request (see Section 5.2 of [RFC7873]). it's a little bit of trust, not a whole fortress 🫶
DNS Cookies do not provide protection against "on-path" adversaries (see Section 9 of [RFC7873]) 💀 an on-path observer that has seen a Server Cookie for a client can abuse that Server Cookie to spoof request for that client within the time span a Server Cookie is valid (see Section 4.3). not me watching a stranger steal the cookie 🙈
The Server Cookie is calculated from the Client Cookie, a series of Sub-Fields specified below, the Client IP address, and a Server Secret that is known only to the server or only to the set of servers at the same anycast address 🔑
For calculation of the Server Cookie, a pseudorandom function is RECOMMENDED with the property that an attacker that does not know the Server Secret cannot find (any information about) the Server Secret, and cannot create a Server Cookie for any combination of the Client Cookie, the series of Sub-Fields specified below, and the client IP address, for which it has not seen a Server Cookie before. the vibe is "unguessable or bust" 🌈 Because DNS servers need to use the pseudorandom function in order to verify Server Cookies, it is RECOMMENDED that it be efficient to calculate ⚡ The pseudorandom function described in [SipHash-2-4] and introduced in Section 4.4 of this document fits these recommendations. slay, that's our girl 👑
Changing the Server Secret regularly is RECOMMENDED, but when a secure pseudorandom function is used, it doesn't need to change too frequently. Once a month, for example, would be adequate 🗓️ See Section 5 on operator and implementation guidelines for updating a Server Secret.
The 128-bit Server Cookie consists of the following Sub-Fields: a 1-octet Version Sub-Field, a 3-octet Reserved Sub-Field, a 4-octet Timestamp Sub-Field, and an 8-octet Hash Sub-Field. here's the layout 🎀
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Version | Reserved |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Timestamp |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Hash |
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+4.1. The Version Sub-Field
The Version Sub-Field prescribes the structure and Hash calculation formula. This document defines Version 1 to be the structure and way to calculate the Hash Sub-Field as defined in this section. the version byte is basically the cookie's little ID badge 🪪
4.2. The Reserved Sub-Field
The value of the Reserved Sub-Field is reserved for future versions of server-side cookie construction. On construction, it MUST be set to zero octets. On Server Cookie verification, the server MUST NOT enforce those fields to be zero, and the Hash should be computed with the received value as described in Section 4.4. so the zeros are for building it, but the checker won't be picky about them 🧚
4.3. The Timestamp Sub-Field
The Timestamp value prevents Replay Attacks and MUST be checked by the server to be within a defined period of time. The DNS server SHOULD allow cookies within a 1-hour period in the past and a 5-minute period into the future, to allow operation of low-volume clients and some limited time skew between the DNS servers in the anycast set ⏰🍓
The Timestamp value specifies a date and time in the form of a 32-bit unsigned number of seconds elapsed since 1 January 1970 00:00:00 UTC, ignoring leap seconds, in network byte order. All comparisons involving these fields MUST use "Serial number arithmetic", as defined in [RFC1982]. [RFC1982] specifies how the differences should be handled. This handles any relative time window less than 68 years, at any time in the future (2038, 2106, 2256, 22209, or later.) the math is giving time-travel-proof energy 🌸
The DNS server SHOULD generate a new Server Cookie at least if the received Server Cookie from the client is more than half an hour old, but it MAY generate a new cookie more often than that 🦋
4.4. The Hash Sub-Field
It's important that all the DNS servers use the same algorithm for computing the Server Cookie. This document defines the Version 1 of the server-side algorithm to be:
Hash = SipHash-2-4(
Client Cookie | Version | Reserved | Timestamp | Client-IP,
Server Secret )where "|" indicates concatenation.
Notice that Client-IP is used for hash generation even though it is not included in the cookie value itself 👀 Client-IP can be either 4 bytes for IPv4 or 16 bytes for IPv6. The length of all the concatenated elements (the input into [SipHash-2-4]) MUST be either precisely 20 bytes in case of an IPv4 Client-IP or precisely 32 bytes in case of an IPv6 Client-IP.
When a DNS server receives a Server Cookie version 1 for validation, the length of the received COOKIE option MUST be precisely 24 bytes: 8 bytes for the Client Cookie plus 16 bytes for the Server Cookie. Verification of the length of the received COOKIE option is REQUIRED to guarantee the length of the input into [SipHash-2-4] to be precisely 20 bytes in the case of an IPv4 Client-IP and precisely 32 bytes in the case of an IPv6 Client-IP. This ensures that the input into [SipHash-2-4] is an injective function of the elements making up the input, and thereby prevents data substitution attacks. More specifically, this prevents a 36-byte COOKIE option coming from an IPv4 Client-IP from being validated as if it were coming from an IPv6 Client-IP. length checks are the bouncer at the club 🚪💅
The Server Secret MUST be configurable to make sure that servers in an anycast network return consistent results 🌐
5. Updating the Server Secret
Changing the Server Secret regularly is RECOMMENDED. All servers in an anycast set must be able to verify the Server Cookies constructed by all other servers in that anycast set at all times. Therefore, it is vital that the Server Secret is shared among all servers before it is used to generate Server Cookies on any server 🤝
Also, to maximize maintaining established relationships between clients and servers, an old Server Secret should be valid for verification purposes for a specific period 🫶
To facilitate this, deployment of a new Server Secret MUST be done in three stages. okay so here's the rollout, step by step 🎀
Stage 1
The new Server Secret is deployed on all the servers in an anycast set by the operator.
Each server learns the new Server Secret but keeps using the previous Server Secret to generate Server Cookies.
Server Cookies constructed with both the new Server Secret and the previous Server Secret are considered valid when verifying.
After stage 1 is completed, all the servers in the anycast set have learned the new Server Secret and can verify Server Cookies constructed with it, but keep generating Server Cookies with the old Server Secret.
Stage 2
This stage is initiated by the operator after the Server Cookie is present on all members in the anycast set.
When entering Stage 2, servers start generating Server Cookies with the new Server Secret. The previous Server Secret is not yet removed/forgotten.
Server Cookies constructed with both the new Server Secret and the previous Server Secret are considered valid when verifying.
Stage 3
This stage is initiated by the operator when it can be assumed that most clients have obtained a Server Cookie derived from the new Server Secret.
With this stage, the previous Server Secret can be removed and MUST NOT be used anymore for verifying 🗑️
It is RECOMMENDED that the operator wait, after initiating Stage 2 and before initiating Stage 3, at least a period of time equal to the longest TTL in the zones served by the server plus 1 hour ⏳
The operator SHOULD wait at least longer than the period clients are allowed to use the same Server Cookie, which SHOULD be 1 hour (see Section 4.3). patience is a whole skill, bestie 💜
6. Cookie Algorithms
[SipHash-2-4] is a pseudorandom function suitable as a Message Authentication Code. It is REQUIRED that a compliant DNS server use SipHash-2-4 as a mandatory and default algorithm for DNS Cookies to ensure interoperability between the DNS Implementations 🌈
The construction method and pseudorandom function used in calculating and verifying the Server Cookies are determined by the initial version byte and by the length of the Server Cookie. Additional pseudorandom or construction algorithms for Server Cookies might be added in the future 🔮
7. IANA Considerations
IANA has created a registry under the "Domain Name System (DNS) Parameters" heading as follows:
Registry Name: DNS Server Cookie Methods
Assignment Policy: Expert Review
Reference: [RFC9018], [RFC7873]
Note: A Server Cookie method (construction and pseudorandom algorithm) is determined by the Version in the first byte of the cookie and by the cookie size. Server Cookie size is limited to the inclusive range of 8 to 32 bytes. here's the registry table 📋
+=========+=======+=================================+
| Version | Size | Method |
+=========+=======+=================================+
| 0 | 8-32 | Reserved |
+---------+-------+---------------------------------+
| 1 | 8-15 | Unassigned |
+---------+-------+---------------------------------+
| 1 | 16 | SipHash-2-4 [RFC9018] Section 4 |
+---------+-------+---------------------------------+
| 1 | 17-32 | Unassigned |
+---------+-------+---------------------------------+
| 2-239 | 8-32 | Unassigned |
+---------+-------+---------------------------------+
| 240-254 | 8-32 | Reserved for Private Use |
+---------+-------+---------------------------------+
| 255 | 8-32 | Reserved |
+---------+-------+---------------------------------+
Table 1: DNS Server Cookie Methods8. Security and Privacy Considerations
⚠️ this part is seriously important, so read it carefully, bestie 🙈
DNS Cookies provide limited protection to DNS servers and clients against a variety of denial-of-service amplification, forgery, or cache-poisoning attacks by off-path attackers. They provide no protection against on-path adversaries that can observe the plaintext DNS traffic. An on-path adversary that can observe a Server Cookie for a client and server interaction can use that Server Cookie for denial-of-service amplification, forgery, or cache-poisoning attacks directed at that client for the lifetime of the Server Cookie 💀
8.1. Client Cookie Construction 💖
Okay so, in [RFC7873] the vibe was that it was RECOMMENDED to build a Client Cookie with a pseudorandom function over three things: the Client IP address, the Server IP address, and a secret only the client knows 🔐. The Client IP address was in the mix so a client couldn't get tracked if its IP changed because of privacy mechanisms or whatever ✨.
Here, we changed the game 🦋. A Client Cookie is now just 64 bits of brand-new entropy, created fresh for each new upstream server the client connects to. That means we have to be extra careful about tracking, bestie 🙈. So a new Client Cookie for a server MUST be created whenever the Client IP address changes.
Sadly, keeping tabs on Client IP changes is impractical with servers that don't support DNS Cookies 😭. So to stop tracking clients of those servers, a client MUST NOT send a previously sent Client Cookie to a server not known to support DNS Cookies. And to stop a brand-new Client Cookie from being made for every single query to a non-DNS Cookie-supporting server, it is RECOMMENDED to not send a Client Cookie to that server for a certain period, for example five minutes ⏳.
Here's the tl;dr 💅:
- In order to provide minimal authentication, a client MUST use a different Client Cookie for each different Server IP address.
- To prevent tracking of clients, a new Client Cookie MUST be created when the Client IP address changes.
- To prevent tracking of clients by a non-DNS Cookie-supporting server, a client MUST NOT send a previously sent Client Cookie to a server in the absence of an associated Server Cookie.
Also, it is infeasible for a client to detect a change in the public IP address when the client is behind a routing device performing Network Address Translation (NAT) 🌸. A server may track the public IP address of that NAT-performing routing device. Preventing tracking of the public IP of a NAT-performing routing device is beyond the scope of this document, so that's a different story 🎀.
8.2. Server Cookie Construction 🔥
[RFC7873] didn't hand us a precise recipe for Server Cookies, but it did point toward using a pseudorandom function strong enough that nobody can guess cookies. In this document, SipHash-2-4 is assigned as the pseudorandom function for version 1 Server Cookies 👑. SipHash-2-4 is considered sufficiently strong for the immediate future, but predicting future developments in cryptography and cryptanalysis is beyond the scope of this document, sorry not sorry 💜.
The precise structure of version 1 Server Cookies is defined right here in this document 📖. Part of that structure is unhashed data elements that are exposed in cleartext to an on-path observer 👀. Those unhashed data elements are fed as input to the SipHash-2-4 function, and its result is the other portion of the Server Cookie. So the unhashed portion can't be changed by an on-path attacker without also recalculating the hashed portion, and that requires knowing the Server Secret 🔑. Iconic design, honestly.
One of the elements in the unhashed portion of version 1 Server Cookies is a Timestamp, which is used to prevent Replay Attacks 🚫. Servers verifying version 1 Server Cookies need access to a reliable time value that an attacker can't alter, so they can compare it with the Timestamp value ⏰. Also, all servers in an anycast set that validate version 1 Server Cookies need their clocks synchronized. Everyone has to be on the same time, bestie 🫶.
For an on-path adversary observing a Server Cookie (as mentioned in the first paragraph of Section 8), the cleartext Timestamp data element reveals the lifetime during which the observed Server Cookie can be used to attack the client 😬. Scary, but now you know.
In addition to the Security Considerations in this section, the Security Considerations section of [RFC7873] still applies 🛡️.
9. References 📚
9.1. Normative References
[RFC1982] Elz, R. and R. Bush, "Serial Number Arithmetic", RFC 1982, DOI 10.17487/RFC1982, August 1996, <https://www.rfc-editor.org/info/rfc1982>.
[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>.
[RFC3339] Klyne, G. and C. Newman, "Date and Time on the Internet: Timestamps", RFC 3339, DOI 10.17487/RFC3339, July 2002, <https://www.rfc-editor.org/info/rfc3339>.
[RFC7873] Eastlake 3rd, D. and M. Andrews, "Domain Name System (DNS) Cookies", RFC 7873, DOI 10.17487/RFC7873, May 2016, <https://www.rfc-editor.org/info/rfc7873>.
[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>.
[SipHash-2-4] Aumasson, J. and D. J. Bernstein, "SipHash: A Fast Short-Input PRF", Progress in Cryptology - INDOCRYPT 2012, Lecture Notes in Computer Science, vol. 7668, December 2012, <https://doi.org/10.1007/978-3-642-34931-7_28>.
9.2. Informative References
[RFC8981] Gont, F., Krishnan, S., Narten, T., and R. Draves, "Temporary Address Extensions for Stateless Address Autoconfiguration in IPv6", RFC 8981, DOI 10.17487/RFC8981, February 2021, <https://www.rfc-editor.org/info/rfc8981>.
Appendix A. Test Vectors 🧚
A.1. Learning a New Server Cookie 🌈
Picture a resolver (client) sending from IPv4 address 198.51.100.100. It asks for "example.com" from an authoritative server listening on 192.0.2.53, and it hasn't learned that server's Server Cookie yet 🍓.
The DNS requests and replies in this appendix are in a "dig"-like format 💅. The content of the DNS COOKIE Option is shown in hexadecimal format after "; COOKIE:".
;; Sending:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 57406
;; flags:; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: 2464c4abcf10c957
;; QUESTION SECTION:
;example.com. IN A
;; QUERY SIZE: 52The authoritative nameserver (server) is configured with the following secret: e5e973e5a6b2a43f48e7dc849e37bfcf (as hex data) 🔑.
It receives the query on Wed Jun 5 10:53:05 UTC 2019 ⏰.
The content of the DNS COOKIE Option that the server returns is shown below in hexadecimal format after "; COOKIE:" ✨.
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 57406
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: 2464c4abcf10c957010000005cf79f111f8130c3eee29480 (good)
;; QUESTION SECTION:
;example.com. IN A
;; ANSWER SECTION:
example.com. 86400 IN A 192.0.2.34
;; Query time: 6 msec
;; SERVER: 192.0.2.53#53(192.0.2.53)
;; WHEN: Wed Jun 5 10:53:05 UTC 2019
;; MSD SIZE rcvd: 84The Timestamp field Section 4.3 in the returned Server Cookie has value 1559731985. In the format described in [RFC3339], this is 2019-06-05 10:53:05+00:00 🎀.
A.2. The Same Client Learning a Renewed (Fresh) Server Cookie 🦋
40 minutes later, the same resolver (client) queries the same server for "example.org" 🕐. It reuses the Server Cookie it learned in the previous query.
The Timestamp field in that previously learned Server Cookie, which is now sent along in the request, was and is 1559731985. In the format of [RFC3339], this is 2019-06-05 10:53:05+00:00 🌸.
;; Sending:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 50939
;; flags:; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: 2464c4abcf10c957010000005cf79f111f8130c3eee29480
;; QUESTION SECTION:
;example.org. IN A
;; QUERY SIZE: 52The authoritative nameserver (server) now generates a new Server Cookie. The server SHOULD do this, because it can see the Server Cookie the client sent is older than half an hour (Section 4.3) 👀. It's also fine for a server to generate a new Server Cookie sooner, or even for every answer 💖.
The Timestamp field in the returned new Server Cookie has value 1559734385, which, in the format of [RFC3339], is 2019-06-05 11:33:05+00:00 ⭐.
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 50939
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: 2464c4abcf10c957010000005cf7a871d4a564a1442aca77 (good)
;; QUESTION SECTION:
;example.org. IN A
;; ANSWER SECTION:
example.org. 86400 IN A 192.0.2.34
;; Query time: 6 msec
;; SERVER: 192.0.2.53#53(192.0.2.53)
;; WHEN: Wed Jun 5 11:33:05 UTC 2019
;; MSD SIZE rcvd: 84A.3. Another Client Learning a Renewed Server Cookie 💜
Another resolver (client) with IPv4 address 203.0.113.203 sends a request to the same server with a valid Server Cookie it learned before (on Wed Jun 5 09:46:25 UTC 2019) 🧚.
The Timestamp field of the Server Cookie in the request has value 1559727985, which, in the format of [RFC3339], is 2019-06-05 09:46:25+00:00 ⏰.
Heads up 🚨: the Server Cookie has Reserved bytes set but is still valid with the configured secret. The Hash part is calculated taking along the Reserved bytes.
;; Sending:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 34736
;; flags:; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: fc93fc62807ddb8601abcdef5cf78f71a314227b6679ebf5
;; QUESTION SECTION:
;example.com. IN A
;; QUERY SIZE: 52The authoritative nameserver (server) replies with a freshly generated Server Cookie for this client, conformant with this specification, i.e., with the Reserved bits set to zero 🙌.
The Timestamp field in the returned new Server Cookie has value 1559734700, which, in the format of [RFC3339], is 2019-06-05 11:38:20+00:00 🎀.
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 34736
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: fc93fc62807ddb86010000005cf7a9acf73a7810aca2381e (good)
;; QUESTION SECTION:
;example.com. IN A
;; ANSWER SECTION:
example.com. 86400 IN A 192.0.2.34
;; Query time: 6 msec
;; SERVER: 192.0.2.53#53(192.0.2.53)
;; WHEN: Wed Jun 5 11:38:20 UTC 2019
;; MSD SIZE rcvd: 84A.4. IPv6 Query with Rolled Over Secret 🌈
okay so this one is the IPv6 glow-up 🦋 A client at 2001:db8:220:1:59de:d0f4:8769:82b8 is talking to a server at 2001:db8:8f::53. The client already got a valid Server Cookie back on Wed Jun 5 13:36:57 UTC 2019, back when the server was using the secret dd3bdf9344b678b185a6f5cb60fca715. Then the server swapped to a brand new secret, but it can still check the cookie the client sends because the old secret hasn't expired yet. no drama, just a smooth secret transition 💅
The Timestamp field in the Server Cookie in the request is 1559741817. In [RFC3339] format that's 2019-06-05 13:36:57+00:00 ⏰
Here's the query going out:
;; Sending:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 6774
;; flags:; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: 22681ab97d52c298010000005cf7c57926556bd0934c72f8
;; QUESTION SECTION:
;example.net. IN A
;; QUERY SIZE: 52The authoritative nameserver (the server) replies with a freshly generated Server Cookie for this same client, made with its new secret: 445536bcd2513298075a5d379663c962 🫶
The Timestamp field in that new Server Cookie is 1559741961, which in [RFC3339] format is 2019-06-05 13:39:21+00:00 🎀
And here's the answer that comes back:
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 6774
;; flags: qr aa; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
; COOKIE: 22681ab97d52c298010000005cf7c609a6bb79d16625507a (good)
;; QUESTION SECTION:
;example.net. IN A
;; ANSWER SECTION:
example.net. 86400 IN A 192.0.2.34
;; Query time: 6 msec
;; SERVER: 2001:db8:8f::53#53(2001:db8:8f::53)
;; WHEN: Wed Jun 5 13:36:57 UTC 2019
;; MSD SIZE rcvd: 84Look at that (good) tag on the COOKIE line ✨ The answer is correct and the cookie checks out, even though the secret rolled over. That's the whole point of the example, bestie 👑
Appendix B. Implementation Status 🍓
okay real talk, let's see who's actually shipping this 🔥
At the time of writing, BIND from version 9.16 and Knot DNS from version 2.9.0 create Server Cookies according to the recipe in this document. Iconic behaviour from both 💖
Unbound and NSD have a Proof-of-Concept implementation that was tested for interoperability during the hackathon at IETF 104 in Prague 🧚 They weren't fully there yet, but they showed up and tested 🫶
Construction of privacy-maintaining Client Cookies according to the directions in this document has been implemented in the getdns library. It will be in the upcoming getdns-1.6.1 release, and in Stubby version 0.3.1 🌸
Acknowledgements 🙈
Shoutout to Witold Krecicki and Pieter Lexis for the valuable input, suggestions, text, and, honestly, for the biggest flex of all: implementing a prototype of an interoperable DNS Cookie in Bind9, Knot, and PowerDNS during the hackathon at IETF 104 in Prague. Absolute legends 👑
Thanks also go to Ralph Dolmans, Bob Harold, Daniel Salzman, Martin Hoffmann, Mukund Sivaraman, Petr Spacek, Loganaden Velvindron, Bob Harold, Philip Homburg, Tim Wicinski, and Brian Dickson for valuable input and suggestions 💜 (Bob Harold's name shows up twice in the original, and yes, we're keeping it that way 🙃)
Authors' Addresses 🦋
Ondrej Sury
Internet Systems Consortium
Czechia
Email: ondrej@isc.org
Willem Toorop
NLnet Labs
Science Park 400
1098 XH Amsterdam
Netherlands
Email: willem@nlnetlabs.nl
Donald E. Eastlake 3rd
Futurewei Technologies
2386 Panoramic Circle
Apopka, FL 32703
United States of America
Phone: +1-508-333-2270
Email: d3e3e3@gmail.com
Mark Andrews
Internet Systems Consortium
950 Charter Street
Redwood City, CA 94063
United States of America
Email: marka@isc.org📦 where this page comes from
- ietf: rfc9018 2021-04
- 💖 girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms