Internet Research Task Force (IRTF) J. Ahrenholz Request for Comments: 6537 The Boeing Company Category: Experimental February 2012 ISSN: 2070-1721 Host Identity Protocol Distributed Hash Table Interface Abstract This document specifies a common interface for using the Host Identity Protocol (HIP) with a Distributed Hash Table (DHT) service to provide a name-to-Host-Identity-Tag lookup service and a Host- Identity-Tag-to-address lookup service. Status of This Memo This document is not an Internet Standards Track specification; it is published for examination, experimental implementation, and evaluation. This document defines an Experimental Protocol for the Internet community. This document is a product of the Internet Research Task Force (IRTF). The IRTF publishes the results of Internet-related research and development activities. These results might not be suitable for deployment. This RFC represents the consensus of the HIP Research Group of the Internet Research Task Force (IRTF). Documents approved for publication by the IRSG are not a candidate for any level of Internet Standard; see Section 2 of RFC 5741. Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc6537. Copyright Notice Copyright (c) 2012 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. Ahrenholz Experimental [Page 1] RFC 6537 HIP DHT Interface February 2012 Table of Contents 1. Introduction ....................................................2 2. The OpenDHT Interface ...........................................3 3. HDRR - The HIP DHT Resource Record ..............................6 4. HIP Lookup Services .............................................8 4.1. HIP Name to HIT Lookup .....................................9 4.2. HIP Address Lookup ........................................12 5. Use Cases ......................................................15 6. Issues with DHT Support ........................................16 7. Security Considerations ........................................17 8. IANA Considerations ............................................18 9. Acknowledgments ................................................18 10. References ....................................................19 10.1. Normative References .....................................19 10.2. Informative References ...................................19 1. Introduction The Host Identity Protocol (HIP) [RFC5201] may benefit from a lookup service based on Distributed Hash Tables (DHTs). The Host Identity namespace is flat, consisting of public keys, in contrast to the hierarchical Domain Name System (DNS). These keys are hashed and prefixed to form Host Identity Tags (HITs), which appear as large random numbers. As the current DNS system has been heavily optimized for address lookup, it may be worthwhile to experiment with other services such as those defined here. DHTs manage such data well by applying a hash function that distributes data across a number of servers. DHTs are also designed to be updated more frequently than a DNS-based approach. For an alternative method of using HITs to look up IP addresses using DNS, see [HIT2IP]. One freely available implementation of a DHT is the Bamboo DHT, which is Java-based software that has been deployed on PlanetLab servers to form a free service named OpenDHT. OpenDHT was available via the Internet for any program to store and retrieve arbitrary data. OpenDHT used a well-defined Extensible Markup Language-Remote Procedure Calling (XML-RPC) interface, featuring put, get, and remove operations. OpenLookup, while not implemented as a DHT, is another deployment of open source software compatible with this OpenDHT interface. This document discusses a common way for HIP to use this OpenDHT interface, so that various HIP experimenters may employ lookup services in an interoperable fashion. This document is a product of the HIP research group (RG) of the IRTF. The HIP research group reached consensus that this interface specification should be published as an Experimental RFC, based on Ahrenholz Experimental [Page 2] RFC 6537 HIP DHT Interface February 2012 document review by at least six RG members including the chairs, and based on implementation experience. This document is not an IETF product and is not a standard. 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 [RFC2119]. 2. The OpenDHT Interface OpenDHT [OPENDHT] was a public deployment of Bamboo DHT servers that ran on about 150 PlanetLab nodes, and was retired in July 2009. While the Bamboo project provided the actual software running on the servers, here we will refer only to OpenDHT, which uses a certain defined interface for the XML-RPC calls. Another service compatible with this interface is OpenLookup. One can run their own Bamboo nodes to set up a private ring of servers. OpenDHT was chosen because it was a well-known, publicly available DHT used within the research community. Its interface features a simple, standards-based protocol that can be easily implemented by HIP developers. This document does not aim to dictate that only the services and servers described here should be used, but is rather meant to act as a starting point to gain experience with these services, choosing tools that are readily available. OpenDHT stores values and indexes those values by using (hash) keys. Keys are limited to 20 bytes in length, and values can be up to 1024 bytes. Values are stored for a certain number of seconds, up to a maximum of 604,800 seconds (one week.) For more information, see the OpenDHT website: . Three RPC operations are supported: put, get, and rm (remove). Put is called with key and value parameters, causing the value to be stored using the key as its hash index. Get is called with the key parameter, when you have a key and want to retrieve the value. Rm is called with a hash of the value to be removed along with a secret value, a hash of which was included in the put operation. The definitions below are taken from the OpenDHT users guide at . Ahrenholz Experimental [Page 3] RFC 6537 HIP DHT Interface February 2012 The put operation takes the following arguments: +----------------+--------------------------------------+ | field | type | +----------------+--------------------------------------+ | application | string | | | | | client_library | string | | | | | key | byte array, 20 bytes max. | | | | | value | byte array, 1024 bytes max. | | | | | ttl_sec | four-byte integer, max. value 604800 | | | | | secret_hash | optional SHA-1 hash of secret value | +----------------+--------------------------------------+ The server replies with an integer -- 0 for "success", 1 if it is "over capacity", and 2 indicating "try again". The return code 3 indicates "failure" and is used for a modified OpenDHT server that performs signature and HIT verification, see Section 3. The get operation takes the following arguments: +----------------+---------------------------------------------+ | field | type | +----------------+---------------------------------------------+ | application | string | | | | | client_library | string | | | | | key | byte array, 20 bytes max. | | | | | maxvals | four-byte singed integer, max. value 2^31-1 | | | | | placemark | byte array, 100 bytes max. | +----------------+---------------------------------------------+ The server replies with an array of values, and a placemark that can be used for fetching additional values. Ahrenholz Experimental [Page 4] RFC 6537 HIP DHT Interface February 2012 The rm operation takes the following arguments: +----------------+----------------------------------------------+ | field | type | +----------------+----------------------------------------------+ | application | string | | | | | client_library | string | | | | | key | byte array, 20 bytes max. | | | | | value_hash | SHA-1 hash of value to remove | | | | | ttl_sec | four-byte integer, max. value 604800 | | | | | secret | secret value (SHA-1 of this was used in put) | +----------------+----------------------------------------------+ The server replies with an integer -- 0 for "success", 1 if it is "over capacity", and 2 indicating "try again". This is the basic XML-RPC interface provided by OpenDHT. Each "field" from the above tables are XML tags that enclose their corresponding values. The key is a byte array used to index the record for storage and retrieval from the DHT. The value is a byte array of the data being stored in the DHT. The application and client_library fields are metadata used only for logging purposes. The ttl_sec field specifies the number of seconds that the DHT should store the value. The secret_hash field allows values to be later removed from the DHT. The maxvals and placemark fields are for retrieving a maximum number of values and for iterating get results. The return code of 0 "success" indicates a successful put or remove operation. The return code of 1 "over capacity" means that a client is using too much storage space on the server. The return value of 2 "try again" indicates that the client should retry the put operation because a temporary problem prevented the server from accepting the put. In the sections that follow, specific uses for these DHT operations and their XML fields are suggested for use with HIP. Ahrenholz Experimental [Page 5] RFC 6537 HIP DHT Interface February 2012 3. HDRR - The HIP DHT Resource Record The two lookup services described in this document use a HIP DHT Resource Record (HDRR) defined in this section. This record is a wrapper around data contained in TLVs, similar to a HIP control packet. The data contained in each HDRR differs between the two services. The HDRR uses the same binary format as HIP packets (defined in [RFC5201].) This packet encoding is used as a convenience, even though this data is actually a resource record stored and retrieved by the DHT servers, not a packet sent on the wire by a HIP protocol daemon. Note that this HDRR format is different than the HIP RR used by the Domain Name System as defined in [RFC5205]. The reason it is different is that it is a different record from a functional point of view: in DNS, the query key is a Fully Qualified Domain Name (FQDN), and the return value is a HIT, while here, the query key is a HIT. HIP header values for the HDRR: HIP Header: Packet Type = 20 DHT Resource Record SRC HIT = Sender's HIT DST HIT = NULL HDRR used with HIT lookup: HIP ( [CERT] ) HDRR used with address lookup: HIP ( LOCATOR, SEQ, HOST_ID, [CERT], HIP_SIGNATURE ) The Initiator HIT (Sender's HIT, SRC HIT) MUST be set to the HIT that the host wishes to make available using the lookup service. With the HIT lookup service, this is the main piece of information returned by a get operation. For the address lookup service, this HIT MUST be the same one used to derive the HIT_KEY used as the DHT key. The Responder HIT (Receiver's HIT, DST HIT) MUST be NULL (all zeroes) since the data is intended for any host. The only other TLV used with the HIT lookup service is an optional CERT parameter containing a certificate for validating the name that is used as the DHT key. The CERT parameter is defined in [RFC6253]. The DHT server SHOULD use the certificate to verify that the client is authorized to use the name used for the DHT key, using the hash of the name found in the certificate. The Common Name (CN) field from the Distinguished Name (DN) of the X.509.v3 certificate MUST be used. Which certificates are considered trusted is a local policy issue. Ahrenholz Experimental [Page 6] RFC 6537 HIP DHT Interface February 2012 The remaining parameters described here are used with the address lookup service. The LOCATOR parameter contains the addresses that the host wishes to make available using the lookup service. A host MAY publish its current preferred IPv4 and IPv6 locators, for example. The SEQ parameter contains an unsigned 32-bit sequence number, the Update ID. This is typically initialized to zero and incremented by one for each new HDRR that is published by the host. The host SHOULD retain the last Update ID value it used for each HIT across reboots, or perform a self lookup in the DHT. The Update ID value may be retained in the DHT records and will determine the preferred address used by peers. The HOST_ID parameter contains the Host Identity that corresponds with the Sender's HIT. (The encoding of this parameter is defined in Section 5.2.8 of [RFC5201].) The HOST_ID parameter and HIP_SIGNATURE parameter MUST be used with the HDRR so that HIP clients receiving the record can validate the sender and the included LOCATOR parameter. The HIT_KEY used for the DHT key will also be verified against the Host Identity. The client that receives the HDRR from the DHT response MUST perform the signature and HIT_KEY verification. If the signature is invalid for the given Host Identity or the HIT_KEY used to retrieve the record does not match the Host Identity, the DHT record retrieved MUST be ignored. Note that for client-only verification, the DHT server does not need to be modified. The Sender's HIT in the HDRR MUST correspond with the key used for the lookup and Host Identity verification. The Receiver's HIT MUST be NULL (all zeroes) in the HDRR header. When several HDRR records are returned by the server, the client SHOULD pick the most recent record as indicated by the Update ID in the SEQ TLV of the HDRR and perform verification on that record. The order in which records are returned should not be considered. The DHT server MAY also verify the SIGNATURE and HOST_ID, with some modifications to the Bamboo DHT software and a new return code with the OpenDHT interface. The signature in the put MUST be verified using the given Host Identity (public key), and the HIT_KEY provided as the lookup key MUST match this Host Identity according to the Overlay Routable Cryptographic Hash Identifiers (ORCHID) generation method defined by [RFC4843]. If either signature or HIT verification Ahrenholz Experimental [Page 7] RFC 6537 HIP DHT Interface February 2012 fails, the put MUST not be recorded into the DHT, and the server returns a failure code. The failure code is an additional return code not defined by OpenDHT, with a value of 3. This server-side verification of records could introduce a source of a denial-of-service attack. The server policy could require clients to have an active HIP association. See Section 7 for further discussion. 4. HIP Lookup Services This document defines a HIT lookup and address lookup service for use with HIP. The HIT lookup uses a text name to discover a peer's HIT. The address lookup uses a peer's HIT to discover its current addresses. The two lookups are defined below. The abbreviated notation refers to the HIP parameter types; for example, HIP_SIG is the HIP signature parameter defined by [RFC5201]. HDRR([CERT]) = get(SHA-1("name")) HDRR(LOCATOR, SEQ, HOST_ID, [CERT], HIP_SIG) = get(HIT_KEY) The HIT lookup service returns the Host Identity Tag of a peer given a name. The name SHOULD be the FQDN, hostname, or some other alias. This HIT is found in the Sender's HIT field of the HDRR. The HIT is the hash of the public-key-based Host Identity as described in [RFC5201]. There are no security properties of the name, unlike the HIT. An optional certificate MAY be included in the record, for validating the name, providing some measure of security. Which certificates are considered trusted is a local policy issue. This service is intended for use when legacy DNS servers do not support HIP resource records, or when hosts do not have administrative access to publish their own DNS records. Such an unmanaged naming service may help facilitate experimentation. The address lookup returns a locator and other validation data in the HDRR for a given HIT. Before a HIP association can be initiated (not in opportunistic mode), a HIP host needs to know the peer's HIT and the current address at which the peer is reachable. Often the HIT will be pre-configured, available via DNS lookup using a hostname lookup [RFC5205], or retrieved using the HIT lookup service defined in this document. With HIP mobility [RFC5206], IP addresses may be used as locators and may often change. The Host Identity and the HIT remain relatively constant and can be used to securely identify a host, so the HIT serves as a suitable DHT key for storing and retrieving addresses. Ahrenholz Experimental [Page 8] RFC 6537 HIP DHT Interface February 2012 The address lookup service includes the peer's Host Identity and a signature over the locators. This allows the DHT client or server to validate the address information stored in the DHT. These two separate lookups are defined instead of one because the address record is expected to change more frequently, while the name- to-HIT binding should remain relatively constant. For example, local policy may specify checking the name-to-HIT binding on a daily basis, while the address record is updated hourly for active peers. Also, the client and server validation of the two records is different, with the HIT lookup using certificates verifying the name and the address lookup using a signature produced by the bearer of a particular Host Identity/HIT. These services reduce the amount of pre-configuration required at each HIP host. The address of each peer no longer needs to be known ahead of time, if peers also participate by publishing their addresses. If peers choose to publish their HITs with a name, peer HITs also no longer require pre-configuration. However, discovering an available DHT server for servicing these lookups will require some additional configuration. 4.1. HIP Name to HIT Lookup Given the SHA-1 hash of a name, a lookup returns the HIT of the peer. The hash of a name is used because OpenDHT keys are limited to 20 bytes, so this allows for longer names. Publish, lookup, and remove operations are defined below. HDRR([CERT]) = get(SHA-1("name")) put(SHA-1("name"), HDRR([CERT]), [SHA-1(secret)]) rm(SHA-1("name"), SHA-1(HDRR), secret) Ahrenholz Experimental [Page 9] RFC 6537 HIP DHT Interface February 2012 HIT publish +----------------+----------------------------------------+---------+ | field | value | data | | | | type | +----------------+----------------------------------------+---------+ | application | "hip-name-hit" | string | | | | | | client_library | (implementation dependent) | string | | | | | | key | SHA-1 hash of a name | base64 | | | | encoded | | | | | | value | HDRR([CERT]), with the HIT to be | base64 | | | published contained in the Sender's | encoded | | | HIT field of the HDRR, and an optional | | | | certificate for validating the name | | | | used as the key | | | | | | | ttl_sec | lifetime for this record, value from | numeric | | | 0-604800 seconds | string | | | | | | secret_hash | optional SHA-1 hash of secret value | base64 | | | | encoded | +----------------+----------------------------------------+---------+ HIT lookup +----------------+---------------------------------+----------------+ | field | value | data type | +----------------+---------------------------------+----------------+ | application | "hip-name-hit" | string | | | | | | client_library | (implementation dependent) | string | | | | | | key | SHA-1 hash of a name | base64 encoded | | | | | | maxvals | (implementation dependent) | numeric string | | | | | | placemark | (NULL, or used from server | base64 encoded | | | reply) | | +----------------+---------------------------------+----------------+ Ahrenholz Experimental [Page 10] RFC 6537 HIP DHT Interface February 2012 HIT remove (optional) +----------------+----------------------------------------+---------+ | field | value | data | | | | type | +----------------+----------------------------------------+---------+ | application | "hip-name-hit" | string | | | | | | client_library | (implementation dependent) | string | | | | | | key | SHA-1 hash of a name | base64 | | | | encoded | | | | | | value_hash | SHA-1 hash of HDRR (value used during | base64 | | | publish) to remove | encoded | | | | | | ttl_sec | lifetime for the remove should be | numeric | | | greater than or equal to the amount of | string | | | time remaining for the record | | | | | | | secret | secret value (SHA-1 of this was used | base64 | | | in put) | encoded | +----------------+----------------------------------------+---------+ The key for both HIT publish and lookup is the SHA-1 hash of the name. The name does not necessarily need to be associated with a valid DNS or host name. It does not need to be related to the Domain Identifier found in the HI TLV. OpenDHT limits the keys to 20 bytes in length, so the SHA-1 hash is used to allow arbitrary name lengths. The value used in the publish and lookup response MUST be the base64- encoded HDRR containing the HIT, and MAY include an optional certificate. The HIT MUST be stored in the Sender's HIT field in the HDRR header and is a 128-bit value that can be identified as a HIT both by its length and by the ORCHID prefix [RFC4843] that it starts with. If a certificate is included in this HIT record, the name used for the DHT key MUST be listed in the certificate. The CERT parameter is defined in [RFC6253]. The Common Name (CN) field from the Distinguished Name (DN) of the X.509.v3 certificate MUST be used. The server can hash this name to verify it matches the DHT key. The ttl_sec field specifies the number of seconds requested by the client that the entry should be stored by the DHT server, which is implementation or policy dependent. Ahrenholz Experimental [Page 11] RFC 6537 HIP DHT Interface February 2012 The secret_hash is an optional field used with HIT publish if the value will later be removed with an rm operation. It is RECOMMENDED that clients support these rm operations for the values they publish. The secret_hash contains the base64-encoded SHA-1 hash of some secret value known only to the publishing host. A different secret value SHOULD be used for each put because rm requests are visible on the network. The max_vals and placemark fields used with the HIT lookup are defined by the get XML-RPC interface. 4.2. HIP Address Lookup Given a HIT, a lookup returns the IP address of the peer. The address is contained in a LOCATOR TLV inside the HDRR, along with other validation data. This interface has publish, lookup, and remove operations. A summary of these three operations is listed below. The abbreviated notation refers to the HIP parameter types; for example, HIP_SIG is the HIP signature parameter defined by [RFC5201]. The details of these DHT operations is then described in greater detail. HDRR(LOCATOR, SEQ, HOST_ID, [CERT], HIP_SIG) = get(HIT_KEY) put(HIT_KEY, HDRR(LOCATOR, SEQ, HOST_ID, [CERT], HIP_SIG), [SHA-1(secret)]) rm(HIT_KEY, SHA-1(HDRR), secret) The HDRR is defined in Section 3. It contains one or more locators that the peer wants to publish, a sequence number, the peer's Host Identity, an optional certificate, and a signature over the contents. The HIT_KEY is comprised of the last 100 bits of the HIT appended with 60 zero bits. This is the portion of the HIT used as a DHT key. The last 100 bits are used to avoid uneven distribution of the stored values across the DHT servers. The HIT's ORCHID Prefix (defined by [RFC4843]) is comprised of the first 28 bits, and this prefix is dropped because it is the same for all HITs, which would cause this uneven distribution. Zero padding is appended to this 100-bit value to fill the length required by the DHT, 160 bits total. Ahrenholz Experimental [Page 12] RFC 6537 HIP DHT Interface February 2012 Address publish +----------------+----------------------------------------+---------+ | field | value | data | | | | type | +----------------+----------------------------------------+---------+ | application | "hip-addr" | string | | | | | | client_library | (implementation dependent) | string | | | | | | key | HIT_KEY | base64 | | | | encoded | | | | | | value | HDRR(LOCATOR, SEQ, HOST_ID, [CERT], | base64 | | | HIP_SIG), with the IP address to be | encoded | | | published contained in the LOCATOR TLV | | | | in the HDRR, along with other | | | | validation data | | | | | | | ttl_sec | amount of time HDRR should be valid, | numeric | | | or the lifetime of the preferred | string | | | address, a value from 0-604800 seconds | | | | | | | secret_hash | optional SHA-1 hash of secret value | base64 | | | | encoded | +----------------+----------------------------------------+---------+ Address lookup +----------------+---------------------------------+----------------+ | field | value | data type | +----------------+---------------------------------+----------------+ | application | "hip-addr" | string | | | | | | client_library | (implementation dependent) | string | | | | | | key | HIT_KEY | base64 encoded | | | | | | maxvals | (implementation dependent) | numeric string | | | | | | placemark | (NULL, or used from server | base64 encoded | | | reply) | | +----------------+---------------------------------+----------------+ Ahrenholz Experimental [Page 13] RFC 6537 HIP DHT Interface February 2012 Address remove (optional) +----------------+-------------------------------------+------------+ | field | value | data type | +----------------+-------------------------------------+------------+ | application | "hip-addr" | string | | | | | | client_library | (implementation dependent) | string | | | | | | key | HIT_KEY | base64 | | | | encoded | | | | | | value_hash | SHA-1 hash of HDRR (value used | base64 | | | during publish) to remove | encoded | | | | | | ttl_sec | old address lifetime | numeric | | | | string | | | | | | secret | secret value (SHA-1 of this was | base64 | | | used in put) | encoded | +----------------+-------------------------------------+------------+ The application and client_library fields are used for logging in OpenDHT. The client_library may vary between different implementations, specifying the name of the XML-RPC library used or the application that directly makes XML-RPC calls. The key used with the address lookup and with publishing the address is the HIT_KEY as defined above, 160 bits base64 encoded [RFC2045]. The value used in the publish and lookup response is the base64- encoded HDRR containing one or more LOCATORs. The ttl_sec field used with address publish indicates the time-to- live (TTL). This is the number of seconds for which the entry will be stored by the DHT. The TTL SHOULD be set to the number of seconds remaining in the address lifetime. The secret_hash is an optional field that MAY be used with address publish if the value will later be removed with an rm operation. The secret_hash contains the base64-encoded SHA-1 hash of some secret value that MUST be known only to the publishing host. Clients SHOULD include the secret_hash and remove outdated values to reduce the amount of data the peer needs to handle. A different secret value SHOULD be used for each put because rm requests are visible on the network. Ahrenholz Experimental [Page 14] RFC 6537 HIP DHT Interface February 2012 The max_vals and placemark fields used with address lookup are defined by the get XML-RPC interface. The get operation needs to know the maximum number of values to retrieve. The placemark is a value found in the server reply that causes the get to continue to retrieve values starting where it left off. 5. Use Cases Below are some suggestions of when a HIP implementation MAY want to use the HIT and address lookup services. To learn of a peer's HIT, a host might first consult DNS using the peer's hostname if the DNS server supports the HIP resource record defined by [RFC5205]. Sometimes hosts do not have administrative authority over their DNS entries and/or the DNS server is not able to support HIP resource records. Hosts may want to associate other non- DNS names with their HITs. For these and other reasons, a host MAY use the HIT publish service defined in Section 4.1. The peer HIT may be learned by performing a DHT lookup of such a name. Once a peer HIT is learned or configured, an address lookup MAY be performed so that the LOCATORs can be cached and immediately available for when an association is requested. Implementations might load a list of peer HITs on startup, resulting in several lookups that can take some time to complete. However, cached LOCATORs may quickly become obsolete, depending on how often the peer changes its preferred address. Performing an address lookup before sending the I1 may be needed. At this time, the latency of a lookup may be intolerable, and a lookup could instead be performed after the I1 retransmission timer fires -- when no R1 reply has been received -- to detect any change in address. A HIP host SHOULD publish its preferred LOCATORs upon startup, so other hosts may determine where it is reachable. The host SHOULD periodically refresh its HDRR entry because each entry carries a TTL and will eventually expire. Also, when there is a change in the preferred address, usually associated with sending UPDATE packets with included locator parameters, the host SHOULD update its HDRR with the DHT. The old HDRR SHOULD be removed using the rm operation, if a secret value was used in the put. Addresses from the private address space SHOULD NOT be published to the DHT. If the host is located behind a NAT, for example, the host could publish the address of its Rendezvous Server (RVS, from [RFC5204]) to the DHT if that is how it is reachable. In this case, however, a peer could instead simply use the RVS field of the NATed host's HIP DNS record, which would eliminate a separate DHT lookup. Ahrenholz Experimental [Page 15] RFC 6537 HIP DHT Interface February 2012 A HIP host SHOULD also publish its HIT upon startup or whenever a new HIT is configured, for use with the HIT lookup service, if desired. The host SHOULD first check if the name already exists in the DHT by performing a lookup, to avoid interfering with an existing name-to- HIT mapping. The name-to-HIT binding needs to be refreshed periodically before the TTL expires. When publishing data to the DHT server, care should be taken to check the response from the server. The server may respond with an "over capacity" code, indicating that its resources are too burdened to honor the given size and TTL. The host SHOULD then select another server for publishing or reduce the TTL and retry the put operation. 6. Issues with DHT Support The DHT put operation does not replace existing values. If a host does not remove its old HDRR before adding another, several entries may be present. A client performing a lookup SHOULD determine the most recent address based on the Update ID from the SEQ TLV of the HDRR. The order of values returned in the server's response may not be guaranteed. Before performing each put, a host SHOULD remove its old HDRR data using the rm operation. In the case of the HIT lookup service, there is nothing preventing different hosts from publishing the same name. A lookup performed on this name will return multiple HITs that belong to different devices. The server may enforce a policy that requires clients to include a certificate when publishing a HIT, and only store HITs with a name that has been authorized by some trusted certificate. Otherwise, this is an unmanaged free-for-all service, and it is RECOMMENDED that a host simply pick another name. Selecting an appropriate DHT server to use is not covered here. If a particular server becomes unavailable, the connect will timeout and some server selection algorithm SHOULD be performed, such as trying the next server in a configured list. OpenDHT formerly provided a DNS-based anycast service; when one performed a lookup of "opendht.nyuld.net", it returned the two nearest OpenDHT servers. The latency involved with the DHT put and get operations should be considered when using these services with HIP. The calls rely on servers that may be located across the Internet and may trigger communications between servers that add delay. The DHT operations themselves may be slow to produce a response. The maximum size of 1024 bytes for the value field will limit the maximum size of the Host Identity and certificates that may be used within the HDRR. Ahrenholz Experimental [Page 16] RFC 6537 HIP DHT Interface February 2012 7. Security Considerations There are two classes of attacks on this information exchange between the host and DHT server: attacks on the validity of the information provided by the DHT to the host (such as a spoofed DHT response) and attacks on the DHT records themselves (such as polluted records for a given key). Without the server performing some measure of verification, not much can be done to prevent these attacks. For the HIT lookup based on a name (Section 4.1), there are no guarantees on the validity of the HIT. Users concerned with the validity of HITs found in the DHT SHOULD simply exchange HITs out-of- band with peers. Including a signature will not help here because the HIT that identifies the Host Identity for signing is not known ahead of time. A certificate MAY be included with the HIT, which guarantees that the name used for the lookup has been authorized by some third-party authority. Which certificates are considered trusted is a local policy issue. For the address lookup based on HIT (Section 4.2), the validity of the DHT response MUST be checked with the HOST_ID and SIGNATURE parameters in the HDRR. A HIP initiating host SHOULD also validate the DHT response after the R1 message is received during a HIP exchange. The Host Identity provided in the R1 can be hashed to obtain a HIT that MUST be checked against the original HIT. However, a legacy OpenDHT service without server modifications does not prevent an attacker from polluting the DHT records for a known HIT, thereby causing a denial-of-service attack, since server validation is not performed. Relying solely on client validation may be harmful. An attacker can replay the put packets containing the signed HDRR, possibly causing stale or invalid information to exist in the DHT. If an attacker replays the signed put message and changes some aspect each time, and if the server is not performing signature and HIT validation, there could be a multitude of invalid entries stored in the DHT. When a client retrieves these records, it would need to perform signature and HIT verification on each one, which could cause unacceptable amounts of delay or computation. To protect against this type of attack, the DHT server SHOULD perform signature and HIT verification of each put operation as described in Section 3. Another option would be the server running HIP itself and requiring client authentication with a HIP association before accepting HDRR puts. Further validation would be only accepting HIT and address records from the association bound to the same HIT. Ahrenholz Experimental [Page 17] RFC 6537 HIP DHT Interface February 2012 Performing server-side verification adds to the processing burden of the DHT server and may be a source for a denial-of-service attack. Requiring a HIP association before accepting HDRR puts may help here. The HIT verification is less computationally intensive by design, using a hash algorithm. Certificate validation (for name lookups) and signature verification (for HDRRs) may cause unacceptable amounts of computation. A server may rate limit the number of puts that it allows. The SHA-1 message digest algorithm is used in two ways in this document, and the security of using this algorithm should be considered within the context of [RFC6194]. The first use is with the OpenDHT put and remove operations, described in Section 2, and the second is to reduce the size of the name string for the HIT lookup service in Section 4.1. The first use is intended to protect the secret values used to store records in the DHT as described by the OpenDHT interface. An attacker would be able to remove a record, after capturing the plaintext put, if a secret value could be found that produces the same secret hash. The purpose of this document is to maintain interoperable compatibility with that interface, which prescribes the use of SHA-1. Future revisions of that interface should consider hash algorithm agility. The OpenDHT FAQ states that future support for other hash algorithms is planned. The second use of the SHA-1 algorithm is to reduce the arbitrarily sized name strings to fit the fixed OpenDHT key size. No security properties of the SHA-1 algorithm are used in this context. 8. IANA Considerations This document defines a new HIP Packet Type, the "HIP Distributed Hash Table Resource Record (HDRR)". This packet type is defined in Section 3 with a value of 20. 9. Acknowledgments Thanks to Tom Henderson, Samu Varjonen, Andrei Gurtov, Miika Komu, Kristian Slavov, Ken Rimey, Ari Keranen, and Martin Stiemerling for providing comments. Samu most notably contributed the resolver packet and its suggested parameters, which became the HDRR here. Ahrenholz Experimental [Page 18] RFC 6537 HIP DHT Interface February 2012 10. References 10.1. Normative References [OPENDHT] Rhea, S., Godfrey, B., Karp, B., Kubiatowicz, J., Ratnasamy, S., Shenker, S., Stocia, I., and H. Yu, "OpenDHT: A Public DHT Service and Its Uses", Proceedings of ACM SIGCOMM 2005, August 2005. [RFC2045] Freed, N. and N. Borenstein, "Multipurpose Internet Mail Extensions (MIME) Part One: Format of Internet Message Bodies", RFC 2045, November 1996. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC4843] Nikander, P., Laganier, J., and F. Dupont, "An IPv6 Prefix for Overlay Routable Cryptographic Hash Identifiers (ORCHID)", RFC 4843, April 2007. [RFC5201] Moskowitz, R., Nikander, P., Jokela, P., and T. Henderson, "Host Identity Protocol", RFC 5201, April 2008. [RFC5205] Nikander, P. and J. Laganier, "Host Identity Protocol (HIP) Domain Name System (DNS) Extensions", RFC 5205, April 2008. [RFC6194] Polk, T., Chen, L., Turner, S., and P. Hoffman, "Security Considerations for the SHA-0 and SHA-1 Message-Digest Algorithms", RFC 6194, March 2011. [RFC6253] Heer, T. and S. Varjonen, "Host Identity Protocol Certificates", RFC 6253, May 2011. 10.2. Informative References [HIT2IP] Ponomarev, O. and A. Gurtov, "Embedding Host Identity Tags Data in DNS", Work in Progress, July 2009. [RFC5204] Laganier, J. and L. Eggert, "Host Identity Protocol (HIP) Rendezvous Extension", RFC 5204, April 2008. [RFC5206] Nikander, P., Henderson, T., Vogt, C., and J. Arkko, "End- Host Mobility and Multihoming with the Host Identity Protocol", RFC 5206, April 2008. Ahrenholz Experimental [Page 19] RFC 6537 HIP DHT Interface February 2012 Author's Address Jeff Ahrenholz The Boeing Company P.O. Box 3707 Seattle, WA USA EMail: jeffrey.m.ahrenholz@boeing.com Ahrenholz Experimental [Page 20]