SI6 Networks / UTN-FRH

Evaristo Carriego 2644 1706Haedo Provincia de Buenos Aires Argentina +54 11 4650 8472 fgont@si6networks.com http://www.si6networks.com

Huawei Technologies

Bantian, Longgang District Shenzhen 518129 China liushucheng@huawei.com

Alcatel-Lucent

Copernicuslaan 50 2018 Antwerp Antwerp Belgium +32 476 476 022 gunter.van_de_velde@alcatel-lucent.com

Operations and Management opsec This document specifies a mechanism for protecting hosts connected to a switched network against rogue DHCPv6 servers. It is based on DHCPv6 packet filtering at the layer 2 device at which the packets are received. A similar mechanism has been widely deployed in IPv4 networks ('DHCP snooping'); hence, it is desirable that similar functionality be provided for IPv6 networks. This document specifies a Best Current Practice for the implementation of DHCPv6-Shield.

This document specifies DHCPv6-Shield, a mechanism for protecting hosts connected to a switched network against rogue DHCPv6 servers . The basic concept behind DHCPv6-Shield is that a layer 2 device filters DHCPv6 messages intended for DHCPv6 clients (henceforth, "DHCPv6-server messages"), according to a number of different criteria. The most basic filtering criterion is that DHCPv6-server messages are discarded by the layer 2 device unless they are received on specific ports of the layer 2 device. Before the DHCPv6-Shield device is deployed, the administrator specifies the layer 2 port(s) on which DHCPv6-server messages are to be allowed. Only those ports to which a DHCPv6 server or relay is to be connected should be specified as such. Once deployed, the DHCPv6-Shield device inspects received packets and allows (i.e., passes) DHCPv6-server messages only if they are received on layer 2 ports that have been explicitly configured for such purpose. DHCPv6-Shield is analogous to the Router Advertisement Guard (RA-Guard) mechanism , intended for protection against rogue Router Advertisement messages. We note that DHCPv6-Shield mitigates only DHCPv6-based attacks against hosts. Attack vectors based on other messages meant for network configuration (such as ICMPv6 Router Advertisements) are not addressed by DHCPv6-Shield itself. In a similar vein, DHCPv6-Shield does not mitigate attacks against DHCPv6 servers (e.g., Denial of Service).

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 .

The set of filtering rules specified in this document, meant to mitigate attacks that employ DHCPv6-server packets. A layer 2 device (typically a layer 2 switch) that enforces the filtering policy specified in this document. For the purposes of this document, the terms "IPv6 Extension Header", "First Fragment", "IPv6 Header Chain", and "Upper-Layer Header" are used as specified in : IPv6 Extension Headers are defined in Section 4 of . As a result of , provides a list of assigned Internet Protocol Numbers and designates which of those protocol numbers also represent IPv6 Extension Headers. An IPv6 fragment with a Fragment Offset equal to 0. The IPv6 Header Chain contains an initial IPv6 header, zero or more IPv6 Extension Headers, and optionally, a single Upper-Layer Header. If an Upper-Layer Header is present, it terminates the IPv6 Header Chain; otherwise, the "No Next Header" value (Next Header = 59) terminates it. The first member of the IPv6 Header Chain is always an IPv6 header. For a subsequent header to qualify as a member of the IPv6 Header Chain, it must be referenced by the "Next Header" field of the previous member of the IPv6 Header Chain. However, if a second IPv6 header appears in the IPv6 Header Chain, as is the case when IPv6 is tunneled over IPv6, the second IPv6 header is considered to be an Upper-Layer Header and terminates the IPv6 Header Chain. Likewise, if an Encapsulating Security Payload (ESP) header appears in the IPv6 Header Chain, it is considered to be an Upper-Layer Header, and it terminates the IPv6 Header Chain. In the general case, the Upper-Layer Header is the first member of the Header Chain that is neither an IPv6 header nor an IPv6 Extension Header. However, if either an ESP header or a second IPv6 header occurs in the IPv6 Header Chain, it is considered to be an Upper-Layer Header, and it terminates the IPv6 Header Chain. Neither the upper-layer payload nor any protocol data following the upper-layer payload is considered to be part of the IPv6 Header Chain. In a simple example, if the Upper-Layer Header is a TCP header, the TCP payload is not part of the IPv6 Header Chain. In a more complex example, if the Upper-Layer Header is an ESP header, neither the payload data nor any of the fields that follow the payload data in the ESP header are part of the IPv6 Header Chain.

Before being deployed for production, the DHCPv6-Shield device is explicitly configured with respect to which layer 2 ports are allowed to receive DHCPv6 packets destined to DHCPv6 clients (i.e., DHCPv6-server messages). Only those layer 2 ports explicitly configured for such purpose are allowed to receive DHCPv6 packets to pass to DHCPv6 clients.

Following are the filtering rules that are enforced as part of a DHCPv6-Shield implementation on those ports that are not allowed to receive DHCPv6 packets to DHCPv6 clients: DHCPv6-Shield implementations MUST parse the entire IPv6 Header Chain present in the packet to identify whether or not it is a DHCPv6 packet meant for a DHCPv6 client (i.e., a DHCPv6-server message). RATIONALE: DHCPv6-Shield implementations MUST NOT enforce a limit on the number of bytes they can inspect (starting from the beginning of the IPv6 packet), since this could introduce false negatives: DHCP6-server packets received on ports not allowed to receive such packets could be allowed simply because the DHCPv6-Shield device does not parse the entire IPv6 Header Chain present in the packet. When parsing the IPv6 Header Chain, if the packet is a First Fragment (i.e., a packet containing a Fragment Header with the Fragment Offset set to 0) and it fails to contain the entire IPv6 Header Chain (i.e., all the headers starting from the IPv6 header up to, and including, the Upper-Layer Header), DHCPv6-Shield MUST drop the packet and ought to log the packet drop event in an implementation-specific manner as a security fault. RATIONALE: Packets that fail to contain the entire IPv6 Header Chain could otherwise be leveraged for circumventing DHCPv6-Shield. requires that the First Fragment (i.e., the fragment with the Fragment Offset set to 0) contain the entire IPv6 Header Chain. also allows intermediate systems such as routers to drop packets that fail to comply with this requirement. NOTE: This rule should only be applied to IPv6 fragments with a Fragment Offset of 0 (non-First Fragments can be safely passed, since they will never reassemble into a complete datagram if they are part of a DHCPv6 packet meant for a DHCPv6 client received on a port where such packets are not allowed). DHCPv6-Shield MUST provide a configuration knob that controls whether or not packets with unrecognized Next Header values are dropped; this configuration knob MUST default to "drop". When parsing the IPv6 Header Chain, if the packet contains an unrecognized Next Header value and the configuration knob is configured to "drop", DHCPv6-Shield MUST drop the packet and ought to log the packet drop event in an implementation-specific manner as a security fault. RATIONALE: An unrecognized Next Header value could possibly identify an IPv6 Extension Header and thus be leveraged to conceal a DHCPv6-server packet (since there is no way for DHCPv6-Shield to parse past unrecognized Next Header values ). requires that nodes be configurable with respect to whether or not packets with unrecognized headers are forwarded and allows the default behavior to be that such packets be dropped. When parsing the IPv6 Header Chain, if the packet is identified to be a DHCPv6 packet meant for a DHCPv6 client, DHCPv6-Shield MUST drop the packet and SHOULD log the packet drop event in an implementation-specific manner as a security alert. RATIONALE: Ultimately, the goal of DHCPv6-Shield is to drop DHCPv6 packets destined to DHCPv6 clients (i.e., DHCPv6-server messages) that are received on ports that have not been explicitly configured to allow the receipt of such packets. In all other cases, DHCPv6-Shield MUST pass the packet as usual. NOTE: For the purpose of enforcing the DHCPv6-Shield filtering policy, an ESP header should be considered to be an "upper-layer protocol" (that is, it should be considered the last header in the IPv6 Header Chain). This means that packets employing ESP would be passed by the DHCPv6-Shield device to the intended destination. If the destination host does not have a security association with the sender of the aforementioned IPv6 packet, the packet would be dropped. Otherwise, if the packet is considered valid by the IPsec implementation at the receiving host and encapsulates a DHCPv6 message, what to do with such a packet is up to the receiving host. The rules above indicate that if a packet is dropped due to this filtering policy, the packet drop event should be logged in an implementation-specific manner as a security fault. It is useful for the logging mechanism to include a per-port drop counter dedicated to DHCPv6-Shield packet drops. In order to protect current end-node IPv6 implementations, Rule #2 has been defined such that the default is for packets that cannot be positively identified as not being DHCPv6-server packets (because the packet is a fragment that fails to include the entire IPv6 Header Chain) to be dropped. This means that, at least in theory, DHCPv6-Shield could result in false-positive blocking of some legitimate (non-DHCPv6-server) packets. However, as noted in , IPv6 packets that fail to include the entire IPv6 Header Chain are virtually impossible to police with stateless filters and firewalls; hence, they are unlikely to survive in real networks. requires that hosts employing fragmentation include the entire IPv6 Header Chain in the First Fragment (the fragment with the Fragment Offset set to 0), thus eliminating the aforementioned false positives. The aforementioned filtering rules implicitly handle the case of fragmented packets: if the DHCPv6-Shield device fails to identify the upper-layer protocol as a result of the use of fragmentation, the corresponding packets would be dropped. Finally, we note that IPv6 implementations that allow overlapping fragments (i.e., that do not comply with ) might still be subject of DHCPv6-based attacks. However, a recent assessment of IPv6 implementations with respect to their fragment reassembly policy seems to indicate that most current implementations comply with .

The recommendations in this document represent the ideal behavior of a DHCPv6-Shield device. However, in order to implement DHCPv6-Shield on the fast path, it may be necessary to limit the depth into the packet that can be scanned before giving up. In circumstances where there is such a limitation, it is recommended that implementations drop packets after attempting to find a protocol header up to that limit, whatever it is. Ideally, such devices should be configurable with a list of protocol header identifiers so that if new transport protocols are standardized after the device is released, they can be added to the list of protocol header types that the device recognizes. Since any protocol header that is not a UDP header would be passed by the DHCPv6-Shield algorithm, this would allow such devices to avoid blocking the use of new transport protocols. When an implementation must stop searching for recognizable header types in a packet due to such limitations, the device SHOULD be configurable to either pass or drop that packet. The mechanism specified in this document can be used to mitigate DHCPv6-based attacks against hosts. Attack vectors based on other messages meant for network configuration (such as ICMPv6 Router Advertisements) are out of the scope of this document. Additionally, the mechanism specified in this document does not mitigate attacks against DHCPv6 servers (e.g., Denial of Service). If deployed in a layer 2 domain with several cascading switches, there will be an ingress port on the host's local switch that will need to be enabled for receiving DHCPv6-server messages. However, this local switch will be reliant on the upstream devices filtering out rogue DHCPv6-server messages, as the local switch has no way of determining which upstream DHCP-server messages are valid. Therefore, in order to be effective, DHCPv6-Shield should be deployed and enabled on all layer 2 switches of a given layer 2 domain. As noted in , IPv6 implementations that allow overlapping fragments (i.e., that do not comply with ) might still be subject to DHCPv6-based attacks. However, most current implementations seem to comply with and hence forbid IPv6 overlapping fragments. We note that if an attacker sends a fragmented DHCPv6 packet on a port not allowed to receive such packets, the First Fragment would be dropped, and the rest of the fragments would be passed. This means that the victim node would tie memory buffers for the aforementioned fragments, which would never reassemble into a complete datagram. If a large number of such packets were sent by an attacker, and the victim node failed to implement proper resource management for the fragment reassembly buffer, this could lead to a Denial of Service (DoS). However, this does not really introduce a new attack vector, since an attacker could always perform the same attack by sending a forged fragmented datagram in which at least one of the fragments is missing. discusses some resource management strategies that could be implemented for the fragment reassembly buffer. Additionally, we note that the security of a site employing DHCPv6-Shield could be further improved by deploying to mitigate IPv6 address spoofing attacks. Finally, we note that other mechanisms for mitigating attacks based on DHCPv6-server messages are available that have different deployment considerations. For example, allows for authentication of DHCPv6-server packets if the IPv6 addresses of the DHCPv6 servers can be pre-configured at the client nodes.

IANA SI6 Networks Centre for the Protection of National Infrastructure

The authors would like to thank Mike Heard, who provided detailed feedback on earlier draft versions of this document and helped a lot in producing a technically sound document throughout the whole publication process. The authors would like to thank (in alphabetical order) Ben Campbell, Jean-Michel Combes, Sheng Jiang, Ted Lemon, Pete Resnick, Carsten Schmoll, Juergen Schoenwaelder, Robert Sleigh, Donald Smith, Mark Smith, Hannes Tschofenig, Eric Vyncke, and Qin Wu for providing valuable comments on earlier draft versions of this document. Part of of this document was borrowed from , authored by Fernando Gont, Vishwas Manral, and Ron Bonica. This document is heavily based on , authored by Fernando Gont. Thus, the authors would like to thank the following individuals for providing valuable comments on : Ran Atkinson, Karl Auer, Robert Downie, Washam Fan, David Farmer, Mike Heard, Marc Heuse, Nick Hilliard, Ray Hunter, Joel Jaeggli, Simon Perreault, Arturo Servin, Gunter Van de Velde, James Woodyatt, and Bjoern A. Zeeb. The authors would like to thank Joel Jaeggli for his advice and guidance throughout the IETF process. Fernando Gont would like to thank Diego Armando Maradona for his magic and inspiration.