MLD Security
draft-vyncke-pim-mld-security-00

Abstract

The latest version of Multicast Listener Discovery protocol is defined in RFC 3810, dated back in 2004. New security research has exhibited new vulnerabilities in MLD, both remote and local attack vectors. This document describes those vulnerabilities and proposes specific mitigation techniques.

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 http://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 December 27, 2015.

Copyright Notice

Copyright (c) 2015 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 (http://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
• 2. Local Vulnerabilities
• 2.1. Downgrading to MLDv1
• 2.2. Queries sent to unicast address
• 2.3. Win the election
• 2.4. Host enumeration
• 2.5. Flooding of MLD messages
• 2.6. Amplification
• 3. Remote Vulnerabilities
• 4. Mitigations
• 5. IANA Considerations
• 6. Security Considerations
• 7. Acknowledgements
• 8. References
• 8.1. Normative References
• 8.2. Informative References
• Authors' Addresses

1. Introduction

The Multicast Listener Discovery protocol version 2 (MLDv2) RFC3810 [RFC3810] has a security section but it was not exhaustive and the focus was only on local forged MLD packets. This document goes beyond those attacks.

For the reader who is not familiar with MLDv2, here are the main points:[Troopers2015].

• Multicast routers send MLD queries which are either generic (query about all multicast group) sent to ff02::1 (link-scope all nodes) or specific (query about a specific group) sent to this multicast group. Query message can also be sent to a unicast address.
• Multicast members reply to MLDv2 queries with reports sent to ff02::16 (link-scope all MDLDv2 routers). In version 1 of MLD RFC2710 [RFC2710], the reports are sent to the multicast group being reported. Reports can be transmitted twice or more in order to ensure that the MLD router gets at least one report.
• All MLD packets are ICMPv6 RFC443 [RFC4443] messages sent with a hop-limit of 1, from a link-local address and there is no authentication
• MLD messages received with a hop-limit greater than 1 should be discarded
• Neighbor Discovery Protocol RFC4861 [RFC4861] requires nodes to become member of the respective solicited-node multicast groups for all their link-scope and global-scope addresses
• Switches are assumed to implement MLD snooping RFC4541 [RFC4541] to learn where to forward multicast packets. It must be noted though that implementations of MLD snooping do not act on link-local multicast groups such as solicited-node multicast group: they simply forward all packets destined to a link-local multicast group to all port in the same layer-2 network
• Every IPv6 node must support MLD

This document is heavily based on previous research:

2. Local Vulnerabilities

2.1. Downgrading to MLDv1

A single MLDv1 report message is enough to downgrade all MLD nodes (hosts and routers) to the version 1 protocol. This could be used to force a MLD host to reply with MLDv1 reports sent to the multicast group rather than to ff02::16. This downgrade to MLDv1 could also be used to transmit the MLDv1 report with a 'done' operation to remove a listener (stopping the multicast traffic on the subnet). Another consequence of downgrading to MLDv1 can be the fact that an attacker can also used “Host Suppression” feature as part of a DoS attack.

2.2. Queries sent to unicast address

Section 5.1.15 of RFC3810 [RFC3810], specifies that for debugging purposes, nodes must accept and process queries sent to any of their addresses (including unicast). Lab testing, described in [Troopers2015], cleary show that all implementations except FreeBSD accept and process MLD queries sent to a unicast global address. An attacker can then completely bypass the MLD router.

2.3. Win the election

When there are multiple MLD routers in a layer-2 domain, the one with the lowest IPv6 address wins the election and becomes the designated MLD router. An hostile node can then send from a lower link-local address a MLD message and become the MLD router. This could be leveraged to mount a denial of service attack.

2.4. Host enumeration

Some hosts try to prevent host enumeration by not responding to ICMPv6 echo request messages sent to any multicast group. But, the same hosts must reply to any MLD queries including the generic one sent to ff02::1, this allows for MLD host enumeration. As hosts join different groups based on their operating system (specific groups for Microsoft Windows for example), the MLD report can also help for OS fingerprinting.

2.5. Flooding of MLD messages

If an implementation does not rate limit in hardware the rate of processed MLD messages, then they are vulnerable to a CPU exhaustion denial of services. If a node does not limit the number of states associated to MLD, then this node is vulnerable to a memory exhaustion denial of services.

2.6. Amplification

Nodes usually join multiple groups (for example, Microsoft Windows 8.1 joins 4 groups). Therefore a forged generic MLDv1 query will force those nodes to transmit MLDv1 reports for each of their groups (in our example 4); furthermore, many implementations send MLD reports twice (in our example 8 in total). MLDv2 is a little better because reports are sent to ff02::16 and not to the multicast group.

3. Remote Vulnerabilities

MLD messages with hop-limit different than 1 should be discarded but nothing prevent a hostile party located n hops away from the victim to send any MLD messages with a hop-limit set to n+1. Therefore, a remote hostile party can mount attacks against MLD (especially because implementations process MLD quaries sent to a global unicast address).

4. Mitigations

This section proposes some mitigation techniques that could be used to prevent the above attacks. This section is not a specification of any kind, the words 'should' is plain English and is not related to RFC2119 [RFC2119].

• Similar to RA-guard RFC6105 [RFC6105], there should be a MLD-guard function in layer-2 switches; MLD queries received on ports attached to non multicast routers should be discarded. Switches could also block all MLDv1 packets in order to prevent the downgrading of MLD version.
• MLD queries should not be accepted and processed when sent to a unicast address (either link-local or global scope).
• All nodes should be able to disabled MLDv1.
• Control plane policing should also be implemented in order to avoid denial of services attacks.
• To mitigate the remote attacks, the hop-limit should have been set to 255 and MLD nodes should discard packets with a hop-limit different than 255.

5. IANA Considerations

This document contains no IANA considerations.

6. Security Considerations

Thi document describes multiple vulnerabilities that have been described above.

7. Acknowledgements

The author would like to thank Stig Venaas for some discussions on this topic.

8. References

8.1. Normative References

8.2. Informative References

Authors' Addresses