Internet-Draft QUIC NAT Traversal March 2024
Seemann & Kinnear Expires 5 September 2024 [Page]
Workgroup:
QUIC
Internet-Draft:
draft-seemann-quic-nat-traversal-02
Published:
Intended Status:
Standards Track
Expires:
Authors:
M. Seemann
E. Kinnear
Apple Inc.

Using QUIC to traverse NATs

Abstract

QUIC is well-suited to various NAT traversal techniques. As it operates over UDP, and because the QUIC header was designed to be demultipexed from other protocols, STUN can be used on the same UDP socket, enabling ICE to be used with QUIC. Furthermore, QUIC’s path validation mechanism can be used to test the viability of an address candidate pair while at the same time creating the NAT bindings required for a direction connection, after which QUIC connection migration can be used to migrate the connection to a direct path.

Discussion Venues

This note is to be removed before publishing as an RFC.

Discussion of this document takes place on the QUIC Working Group mailing list (quic@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/quic/.

Source for this draft and an issue tracker can be found at https://github.com/marten-seemann/draft-seemann-quic-nat-traversal.

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."

This Internet-Draft will expire on 5 September 2024.

Table of Contents

1. Introduction

This document describes two ways to use QUIC ([RFC9000]) to traverse NATs:

  1. Using ICE ([RFC8445]) with an external signaling channel to select a pair of UDP addresses. Once candidate nomination is completed, a new QUIC connection between the two endpoints can be established.

  2. Using a (proxied) QUIC connection as the signaling channel. QUIC's path validation logic is used to test connectivity of possible paths.

The first option documents how NAT traversal can be achieved using unmodified QUIC and ICE stacks. The only requirement is the ability to send and receive non-QUIC (STUN ([RFC5389])) packets on the UDP socket that a QUIC server is listening on. However, it necessitates running a separate signaling channel for the communication between the two ICE agents.

The second option doesn't use ICE at all, although it makes use of some of the concepts, in particular the address matching logic described in [RFC8445]. It is assumed that the nodes are connected via a proxied QUIC connection, for example using [CONNECT-UDP-LISTEN]. Using the QUIC extension defined in this document, the nodes coordinate QUIC path validation attempts that create the necessary NAT bindings to achieve traversal of the NAT. This mechanism makes extensive use of the path validation mechanism described in [RFC9000]. In addition, the QUIC server needs the capability to initiate path validation, which, as per [RFC9000], is initiated by the client. Starting with a proxied QUIC connection allows the nodes to start exchanging application data right away and switch to the direct connection once it has been established and deemed suitable for the application's needs.

2. Conventions 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.

3. NAT Traversal Using an External Signaling Channel

When an external signaling channel is used, the QUIC connection is established after the two ICE agents have agreed on a candidate pair. This mode doesn't require any modification to existing QUIC stacks. In particular, it does not necessitate the negotiation of the extension defined in this document.

For address discovery to work, QUIC and ICE need to use the same UDP socket. Since this requires demultiplexing of QUIC and STUN packets, the QUIC bit cannot be greased as described in [RFC9287].

Once ICE has completed, the client immediately initiates a normal QUIC handshake using the server's address from the nominated address pair. The ICE connectivity checks should have created the necessary NAT bindings for the client's first flight to reach the server and for the server's first flight to reach the client.

4. NAT Traversal using the NAT Traversal QUIC Extension

QUIC's path validation mechanism can be used to establish the required NAT mappings that allow for a direct connection. Once the NAT mappings are established, QUIC's connection migration can be used to migrate the connection to a direct path. During the path validation phase, multiple different paths might be established in parallel. When using QUIC Multipath [MULTIPATH], these paths may be used at the some time, however, the mechanism described in this document does not require the use of QUIC multipath.

Although ICE is not directly used, the logic run on the client makes use of ICE's candidate pairing logic (see especially Section 6.1.2.2 of [RFC8445]). Implementations are free to implement different algorithms as they see fit.

This mode needs be negotiated during the handshake, see Section 4.5.

4.1. Gathering Address Candidates

The gathering of address candidates is out of scope for this document. Endpoints MAY use the logic described in Sections 5.1.1 and 5.2 of [RFC8445], or they MAY use address candidates provided by the application.

4.2. Sending Address Candidates to the Client

The server sends its address candidates to the client using ADD_ADDRESS frames. It SHOULD NOT wait until address candidate discovery has finished, instead, it SHOULD send address candidates as soon as they become available. This speeds up the NAT traversal and is similar to Trickle ICE ([RFC8838]).

Addresses sent to the client can be removed using the REMOVE_ADDRESS frame if the address candidate becomes stale, e.g. because the network interface becomes unavailable.

Since address matching is run on the client side, address candidates are only sent from the server to the client. The client does not send any addresses to the server.

4.3. Address Matching

The client matches the address candidates sent by the server with its own address candidates, forming candidate pairs. Section 5.1 of [RFC8445] describes an algorithm for pairing address candidates. Since the pairing algorithm is only run on the client side, the endpoints do not need to agree on the algorithm used, and the client is free to use a different algorithm.

4.4. Probing Paths

The client sends candidate pairs to the server using PUNCH_ME_NOW frames. The client SHOULD start path validation (see Section 8.2 of [RFC9000]) for the respective path immediately after sending the PUNCH_ME_NOW frame.

The server SHOULD start path validation immediately upon receipt of a PUNCH_ME_NOW frame. This document introduces the concept of path validation on the server side, since [RFC9000] assumes that any QUIC server is able to receive packets on a path without creating a NAT binding first. Path validation on the server works as described in Section 8.2.1 of [RFC9000], with additional rate-limiting (see Section 4.4.2) to prevent amplification attacks.

Path probing happens in rounds, allowing the peers to limit the bandwidth consumed by sending path validation packets. For every round, the client MUST NOT send more PUNCH_ME_NOW frames than allowed by the server's transport parameter. A new round is started when a PUNCH_ME_NOW frame with a higher Round value is received. This immediately cancels all path probes in progress.

To speed up NAT traversal, the client SHOULD send address pairs as soon as they become available. However, for small concurrency limits, it MAY delay sending of address pairs in order rank them first and only initiate path validation for the highest-priority candidate pairs.

4.4.1. Interaction with active_connection_id_limit

The active_connection_id_limit limits the number of connection IDs that are active at any given time. Both endpoints need to use a previously unused connection ID when validating a new path in order to avoid linkability. Therefore, the active_connection_id_limit effectively places a limit on the number of concurrent path validations.

Endpoints SHOULD set an active_connection_id_limit that is high enough to allow for the desired number of concurrent path validation attempts.

4.4.2. Amplification Attack Mitigation

TODO describe exactly how to migitate amplification attacks

4.5. Negotiating Extension Use

Endpoints advertise their support of the extension by sending the nat_traversal (0x3d7e9f0bca12fea6) transport parameter (Section 7.4 of [RFC9000]).

The client MUST send this transport parameter with an empty value. A server implementation that understands this transport parameter MUST treat the receipt of a non-empty value as a connection error of type TRANSPORT_PARAMETER_ERROR.

For the server, the value of this transport parameter is a variable-length integer, the concurrency limit. The concurrency limit limits the amount of concurrent NAT traversal attempts and can be used to limit the bandwith required to execute the path validation. Any value larger than 0 is valid. A client implementation that understands this transport parameter MUST treat the receipt of a value that is not a variable-length integer, or the receipt of the value 0, as a connection error of type TRANSPORT_PARAMETER_ERROR.

In order to the use of this extension in 0-RTT packets, the client MUST remember the value of this transport parameter. If 0-RTT data is accepted by the server, the server MUST not disable this extension on the resumed connection.

4.6. Frames

4.6.1. ADD_ADDRESS Frame

ADD_ADDRESS Frame {
    Type (i) = 0x3d7e90..0x3d7e91,
    Sequence Number (i),
    [ IPv4 (32) ],
    [ IPv6 (128) ],
    Port (16),
}

The ADD_ADDRESS frame contains the following fields:

Sequence Number:

A variable-length integer encoding the sequence number of this address advertisement.

IPv4:

The IPv4 address. Only present if the least significant bit of the frame type is 0.

IPv6:

The IPv6 address. Only present if the least significant bit of the frame type is 1.

Port:

The port number.

ADD_ADDRESS frames are ack-eliciting. When lost, they SHOULD be retransmitted, unless the address is not active anymore.

This frame is only sent from the server to the client. Servers MUST treat receipt of an ADD_ADDRESS frame as a connection error of type PROTOCOL_VIOLATION.

4.6.2. PUNCH_ME_NOW Frame

PUNCH_ME_NOW Frame {
    Type (i) = 0x3d7e92..0x3d7e93,
    Round (i),
    Paired With Sequence Number (i),
    [ IPv4 (32) ],
    [ IPv6 (128) ],
    Port (16),
}

The PUNCH_ME_NOW frame contains the following fields:

Round:

The sequence number of the NAT Traversal attempts.

Paired With Sequence Number:

A variable-length integer encoding the sequence number of the address that was paired with this address.

IPv4:

The IPv4 address. Only present if the least significant bit of the frame type is 0.

IPv6:

The IPv6 address. Only present if the least significant bit of the frame type is 1.

Port:

The port number.

PUNCH_ME_NOW frames are ack-eliciting.

This frame is only sent from the client to the server. Clients MUST treat receipt of a PUNCH_ME_NOW frame as a connection error of type PROTOCOL_VIOLATION.

4.6.3. REMOVE_ADDRESS Frame

REMOVE_ADDRESS Frame {
    Type (i) = 0x3d7e94,
    Sequence Number (i),
}

The REMOVE_ADDRESS frame contains the following fields:

Sequence Number:

A variable-length integer encoding the sequence number of the address advertisement to be removed.

REMOVE_ADDRESS frames are ack-eliciting. When lost, they SHOULD be retransmitted.

This frame is only sent from the server to the client. Servers MUST treat receipt of an REMOVE_ADDRESS frame as a connection error of type PROTOCOL_VIOLATION.

5. Security Considerations

This document expands QUIC's path validation logic to QUIC servers, allowing a QUIC client to request sending of path validation packets on unverified paths. A malicious client can direct traffic to a target IP. This attack is similar to the IP address spoofing attack that address validation during connection establishment (see Section 8.1 of [RFC9000]) is designed to prevent. In practice however, IP address spoofing is often additionally mitigated by both the ingress and egress network at the IP layer, which is not possible when using this extension. The server therefore needs to carefully limit the amount of data it sends on unverified paths.

6. IANA Considerations

TODO: fill out registration request for the transport parameter and frame types

7. Normative References

[CONNECT-UDP-LISTEN]
Schinazi, D. and A. Singh, "Proxying Bound UDP in HTTP", Work in Progress, Internet-Draft, draft-ietf-masque-connect-udp-listen-02, , <https://datatracker.ietf.org/doc/html/draft-ietf-masque-connect-udp-listen-02>.
[MULTIPATH]
Liu, Y., Ma, Y., De Coninck, Q., Bonaventure, O., Huitema, C., and M. Kühlewind, "Multipath Extension for QUIC", Work in Progress, Internet-Draft, draft-ietf-quic-multipath-06, , <https://datatracker.ietf.org/doc/html/draft-ietf-quic-multipath-06>.
[RFC2119]
Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/rfc/rfc2119>.
[RFC5389]
Rosenberg, J., Mahy, R., Matthews, P., and D. Wing, "Session Traversal Utilities for NAT (STUN)", RFC 5389, DOI 10.17487/RFC5389, , <https://www.rfc-editor.org/rfc/rfc5389>.
[RFC8174]
Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/rfc/rfc8174>.
[RFC8445]
Keranen, A., Holmberg, C., and J. Rosenberg, "Interactive Connectivity Establishment (ICE): A Protocol for Network Address Translator (NAT) Traversal", RFC 8445, DOI 10.17487/RFC8445, , <https://www.rfc-editor.org/rfc/rfc8445>.
[RFC8838]
Ivov, E., Uberti, J., and P. Saint-Andre, "Trickle ICE: Incremental Provisioning of Candidates for the Interactive Connectivity Establishment (ICE) Protocol", RFC 8838, DOI 10.17487/RFC8838, , <https://www.rfc-editor.org/rfc/rfc8838>.
[RFC9000]
Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based Multiplexed and Secure Transport", RFC 9000, DOI 10.17487/RFC9000, , <https://www.rfc-editor.org/rfc/rfc9000>.
[RFC9287]
Thomson, M., "Greasing the QUIC Bit", RFC 9287, DOI 10.17487/RFC9287, , <https://www.rfc-editor.org/rfc/rfc9287>.

Acknowledgments

TODO acknowledge.

Authors' Addresses

Marten Seemann
Eric Kinnear
Apple Inc.
One Apple Park Way
Cupertino, California 95014,
United States of America