Network Time Protocol Suggest REFID Extension Field
draft-stenn-ntp-suggest-refid-02

Abstract

NTP has been widely used through several revisions, with the latest being RFC 5905. A core component of the protocol and the algorithms is the Reference ID, or REFID, which is used to identify the source of time used for synchronization. Traditionally, when the source of time was another system the REFID was the IPv4 address of that other system. The purpose of the REFID is to prevent a one-degree timing loop, where if A has several timing sources that include B, if B decides to get its time from A we don't want A then deciding to get its time from B. The REFID is considered to be "public data" and is a vital core-component of the base NTP packet. If a system's REFID is the IPv4 address of its system peer, an attacker can try to use that information to send spoofed time packets to either or both the target or the target's server, attempting to cause a disruption in time service. This proposal is a backward-compatible way for a time source to tell its peers or clients "If you use me as your system peer, use this nonce as your REFID." This nonce SHOULD be untraceable to the original system, and if it is used as the REFID this type of attack is prevented.

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 June 3, 2018.

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 (https://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
• 1.1. Requirements Language
• 2. The REFID
• 3. The Suggested REFID Extension Field
• 4. Generating and Sending the Suggested REFID Extension Field
• 5. Receiving a Suggested REFID Extension Field
• 6. Acknowledgements
• 7. IANA Considerations
• 8. Security Considerations
• 9. Normative References
• Author's Address

1. Introduction

NTP has been widely used through several revisions, with the latest being RFC 5905. A core component of the protocol and the algorithms is the Reference ID, or REFID, which is used to identify the source of time used for synchronization. Traditionally, when the source of time was another system, the REFID was the IPv4 address of that other system. If the remote system was using IPv6 for its connection, a 4 octet digest value of the IPv6 address was used. The purpose of the REFID is to prevent a one-degree timing loop, where if A has several timing sources that include B, if B decides to get its time from A we don't want A then deciding to get its time from B. The REFID is considered to be "public data" and is a vital core-component of the base NTP packet. If a system's REFID is the IPv4 address of its system peer, an attacker can try to use that information to send spoofed time packets to either or both the target or the target's server, attempting to cause a disruption in time service. This proposal is a backward-compatible way for a time source to tell its peers or clients "If you use me as your system peer, use this nonce as your REFID." This nonce, a Suggested REFID, SHOULD be untraceable to the sending system. If the receiving system uses this Suggested REFID nonce instead of the IPv4 address as its REFID, this type of attack and information disclosure is prevented.

The NTP protocol was designed with a mechanism that allowed for a depth-1 loop detection to avoid a simple "time loop". Recently, this mechanism was discovered to be a potential vulnerability exploit. The best way to mitigate this vulnerability is to decouple the IPv4 address of the server from its REFID. But there is no current way for a potential time source to tell the other party any other alternative to use as the REFID. This proposal creates an extension field to accomplish this.

1.1. Requirements Language

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.

2. The REFID

The core reason for the REFID in the NTP Protocol is to prevent a timing loop of degree 1. Put another way, if servers A and B are exchanging time with each other and server B decides to follow A as its system peer, the REFID that B will use must be able to identify server A. The interpretation of a REFID is based on the stratum, as documented in RFC 5905, section 7.3, "Packet Header Variables". At Stratum 2+, which will be the case if servers A and B are exchanging packets over IPv4, if server B follows A, then B will have A's IPv4 address as its REFID. When A asks B for its time, A will see that B is synchronized to A because B will tell A that its REFID is A's IPv4 address, so when A sees its IP address as B's REFID, A knows that if it were to follow B for its time then there would be a timing loop. In this case, A will not select B as a potential source of time.

3. The Suggested REFID Extension Field

Since there is no way in the base NTP packet for "this" instance of an NTP server to tell the "other" instance what REFID it should use if the "other" instance decides to use "this" instance as its system peer, the best available way to convey this information is via an extension field.

 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
+---------------+---------------+-------------------------------+ 
|          Field Type           |        Field Length           | 
+-------------------------------+-------------------------------+ 
|                        Suggested REFID                        | 
+---------------------------------------------------------------+

NTP Extension Field: REFID Suggestion

Field Type: TBD (Recommendation for IANA: 0x0006 (Suggested REFID, MAC required), 0x2006 (Suggested REFID, MAC OPTIONAL))

Field Length: 0x0008

Suggested REFID: The 4 octets of the suggested REFID. This value SHOULD be 0xFDxxxxxx, where the bottom 3 octets SHOULD be random values.

Examples: When decoded as an IPv4 address, suggested REFIDs would decode as 253.0.0.0 thru 253.255.255.255.

4. Generating and Sending the Suggested REFID Extension Field

A system that decides to send a Suggested REFID extension field SHOULD generate a new Suggested REFID for each new association. It MAY generate a new Suggested REFID for any association in any response. In addition to remembering the IP-based REFID, the sender MUST also remember its most-recent Suggested REFID.

Since the core NTPv4 and earlier protocols do not contain any way to tell the recipient what to use as a REFID and RFC 5905 uses the IPv4 address of the sender as the REFID if the association is effected over an IPv4 connection, this means that an attacker can simply send an NTP client request to a server knowing that server's system peer will be returned as the REFID in the response packet. At this point, an attacker can, if that REFID is an IPv4 address, begin to launch attacks at the target forging the putative IP of the target's time source, or the attacker can start forging packets to the putative time server claiming to be from the target, in an attempt to cause the time server to limit or deny time service to the target.

Using a nonce for the REFID that is only recognized by the sending machine effectively prevents this type of attack.

If servers S1, S2, and S3 are all exchanging time with each other and are all using the Suggested REFID mechanism, there is a 3 in 16,777,216 (2^24) chance that two different servers in the same group will happen to choose the same nonce, and that would produce a false-positive timing loop detection. If the Suggested REFID is never changed, this false-positive condition will occur for potentially a long time. This small risk can be reduced by periodically generating a new Suggested REFID.

5. Receiving a Suggested REFID Extension Field

An NTP server keeps track of the IP address it uses to talk to a client. If an NTP server chooses to send a Suggested REFID to an association, it MUST remember this value. When checking for a timing loop, the Suggested REFID must also be included in the list of tested values.

6. Acknowledgements

The author wishes to acknowledge the contributions of Martin Burnicki and Sam Weiler.

7. IANA Considerations

This memo requests IANA to allocate NTP Extension Field Types 0x0006 (Suggested REFID, MAC required) and 0x2006 (Suggested REFID, MAC OPTIONAL) for this purpose.

8. Security Considerations

Additional information TBD

9. Normative References

Author's Address