Request for Comments: 7210
Category: Standards Track
Huawei Technologies Co. Ltd.
Database of Long-Lived Symmetric Cryptographic Keys
This document specifies the information contained in a conceptual database of long-lived cryptographic keys used by many different routing protocols for message security. The database is designed to support both manual and automated key management. In addition to describing the schema for the database, this document describes the operations that can be performed on the database as well as the requirements for the routing protocols that wish to use the database. In many typical scenarios, the protocols do not directly use the long-lived key, but rather a key derivation function is used to derive a short-lived key from a long-lived key.
Status of This Memo
This is an Internet Standards Track document.
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in 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/rfc7210.
Copyright © 2014 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.
This document specifies the information that needs to be included in a database of long-lived cryptographic keys in order to key the cryptographic authentication of routing protocols. This conceptual database is designed to separate protocol-specific aspects from both manual and automated key management. The intent is to allow many different implementation approaches to the specified cryptographic key database, while simplifying specification and heterogeneous deployments. This conceptual database avoids the need to build knowledge of any security protocol into key management protocols. It minimizes protocol-specific knowledge in operational/management interfaces, and it constrains where that knowledge can appear. Textual conventions are provided for the representation of keys and other identifiers. These conventions should be used when presenting keys and identifiers to operational/management interfaces or reading keys/identifiers from these interfaces. This satisfies the operational requirement that all implementations represent the keys and key identifiers in the same way so that cross-vendor configuration instructions can be provided.
Routing protocols can employ the services of more-generic security protocols such as TCP-AO [RFC5925]. Implementations of routing protocols may need to supply keys to databases specific to these security protocols as the associated entries in this document's conceptual database are manipulated.
In many instances, the long-lived keys are not used directly in security protocols, but rather a key derivation function is used to derive short-lived keys from the long-lived key in the database. In other instances, security protocols will directly use the long-lived key from the database. The database design supports both use cases.
1.1. Requirements Notation
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 [RFC2119].
2. Conceptual Database Structure
The database is characterized as a table, where each row represents a single long-lived symmetric cryptographic key. Normally, each key should only have one row. Only in the (hopefully) very rare cases where a key is used for more than one purpose, or where the same key is used with multiple key derivation functions (KDFs) will multiple rows contain the same key value. The columns in the table represent the key value and attributes of the key.
To accommodate manual key management, the format of the fields has been purposefully chosen to allow updates with a plain-text editor and to provide equivalent display on multiple systems.
The columns that the table consists of are listed as follows:
The AdminKeyName field contains a human-readable string meant to identify the key for the user. Implementations can use this field to uniquely identify rows in the key table. The same string can be used on the local system and peer systems, but this is not required. Routing protocols do not make use of this string; they use the LocalKeyName and the PeerKeyName. However, if these strings are to be used as protocol elements in other protocols or otherwise transferred between systems, they will need to follow the requirements of Section 5.1.
The LocalKeyName field contains a string identifying the key. It can be used to retrieve the key in the local database when received in a message. As discussed in Section 4, the protocol defines the form of this field. For example, many routing protocols restrict the format of their key names to integers that can be represented in 16 or 32 bits. Typically, this field does not contain data in human character sets requiring internationalization. If there ever are any routing Protocols with key names requiring internationalization, those specifications need to address issues of canonicalization and normalization so that key names can be compared using binary comparison.
PeerKeyName is the name of the key used by the local system in an outgoing message. For unicast communication, the PeerKeyName of a key on a system matches the LocalKeyName of the identical key that is maintained on one or multiple peer systems. Similar to LocalKeyName, a protocol defines the form of this identifier and will often restrict it to be an integer. For group keys, the protocol will typically require this field be an empty string as the sending and the receiving key names need to be the same.
Typically for unicast keys, this field lists the peer systems that have this key in their database. For group keys, this field names the groups for which the key is appropriate. For example, this might name a routing area for a multicast routing protocol. Formally, this field provides a protocol-specific set of restrictions on the scope in which the key is appropriate. The format of the identifiers in the Peers field is specified by the protocol.
The Interfaces field identifies the set of physical and/or virtual interfaces for which it is appropriate to use this key. When the long-lived value in the Key field is intended for use on any interface, this field is set to "all". The interfaces field consists of a set of strings; the form of these strings is specified by the implementation and is independent of the protocol in question. Protocols may require support for the Interfaces field or may indicate that support for constraining keys based on interface is not required. As an example, TCP-AO implementations are unlikely to make the decision of what interface to use prior to key selection. In that case, the implementations are expected to use the same keying material across all of the interfaces and then require the "all" setting.
The Protocol field identifies a single routing protocol where this key may be used to provide cryptographic protection. This specification establishes a registry for this field; the registry also specifies the format of the following field, ProtocolSpecificInfo, for each registered protocol.
This field contains the protocol-specified information that may be useful for a protocol to apply the key correctly. Note that such information MUST NOT be required for a protocol to locate an appropriate key. When a protocol does not need the information in ProtocolSpecificInfo, it will require this field be empty. Key table rows MAY specify a Direction of "both". As a result, the encoding of this field needs to support encoding protocol-specific information for sending and receiving in the same row.
The KDF field indicates the key derivation function that is used to generate short-lived keys from the long-lived value in the Key field. When the long-lived value in the Key field is intended for direct use, the KDF field is set to "none". A key derivation function is a one-way function that provides cryptographic separation of key material. The KDF MAY use inputs from the row in the key table and the message being sent or received but MUST NOT depend on other configuration state. This document establishes an IANA registry for the values in the KDF field to simplify references in future specifications. The protocol indicates what (if any) KDFs are valid.
The AlgID field indicates which cryptographic algorithm is to be used with the security protocol for the specified peer or peers. Such an algorithm can be an encryption algorithm and mode (e.g., AES-128-CBC), an authentication algorithm (e.g., HMAC-SHA1-96 or AES-128-CMAC), or any other symmetric cryptographic algorithm needed by a security protocol. If the KDF field contains "none", then the long-lived key is used directly with this algorithm; otherwise, the derived short- lived key is used with this algorithm. When the long-lived key is used to generate a set of short-lived keys for use with the security protocol, the AlgID field identifies a ciphersuite rather than a single cryptographic algorithm. This document establishes an IANA registry for the values in the AlgID field to simplify references in future specifications. Protocols indicate which algorithms are appropriate.
The Key field contains a long-lived symmetric cryptographic key in the format of a lowercase hexadecimal string. The size of the Key depends on the KDF and the AlgID. For instance, KDF=none and AlgID=AES128 require a 128-bit key, which is represented by 32 hexadecimal digits.
The Direction field indicates whether this key may be used for inbound traffic, outbound traffic, both, or whether the key has been disabled and may not currently be used at all. The supported values are "in", "out", "both", and "disabled", respectively. The Protocol field will determine which of these values are valid.
The SendLifetimeStart field specifies the earliest date and time in Coordinated Universal Time (UTC) at which this key should be considered for use when sending traffic. The format is YYYYMMDDHHSSZ, where four digits specify the year, two digits specify the month, two digits specify the day, two digits specify the hour, two digits specify the minute, and two digits specify the second. The "Z" is included as a clear indication that the time is in UTC.
The SendLifeTimeEnd field specifies the latest date and time at which this key should be considered for use when sending traffic. The format is the same as the SendLifetimeStart field.
The AcceptLifeTimeStart field specifies the earliest date and time in Coordinated Universal Time (UTC) at which this key should be considered for use when processing received traffic. The format is YYYYMMDDHHSSZ, where four digits specify the year, two digits specify the month, two digits specify the day, two digits specify the hour, two digits specify the minute, and two digits specify the second. The "Z" is included as a clear indication that the time is in UTC.
The AcceptLifeTimeEnd field specifies the latest date and time at which this key should be considered for use when processing the received traffic. The format of this field is identical to the format of AcceptLifeTimeStart.
3. Key Selection and Rollover
A protocol may directly consult the key table to find the key to use on an outgoing message. The protocol provides a protocol (P) and a peer identifier (H) into the key selection function. Optionally, an interface identifier (I) may also need to be provided. Any key that satisfies the following conditions may be selected:
(1) the Peers field includes H;
(2) the Protocol field matches P;
(3) If an interface is specified by the protocol, the Interfaces
field in the key table row includes I or "all";
(4) the Direction field is either "out" or "both"; and
(5) SendLifetimeStart <= current time <= SendLifeTimeEnd.
During key selection, there may be multiple entries that simultaneously exist and are associated with different cryptographic algorithms or ciphersuites. Systems should support selection of keys based on algorithm preference to facilitate algorithm transition.
In addition, multiple entries with overlapping valid periods are expected to be available for orderly key rollover. In these cases, the expectation is that systems will transition to the newest key available. To meet this requirement, this specification recommends supplementing the key selection algorithm with the following differentiation: select the long-lived key specifying the most recent time in the SendLifetimeStart field.
In order to look up a key for validating an incoming message, the protocol provides its protocol (P), the peer identifier (H), the key identifier (L), and optionally the interface (I). If one key matches the following conditions, it is selected:
(1) the Peer field includes H;
(2) the Protocol field matches P;
(3) if the Interface field is provided, it includes I or is
(4) the Direction field is either "in" or "both";
(5) the LocalKeyName is L; and
(6) AcceptLifeTimeStart <= current time <= AcceptLifeTimeEnd.
Note that the key usage is loosely bound by the times specified in the AcceptLifeTimeStart and AcceptLifeTimeEnd fields. New security associations should not be established except within the period of use specified by these fields, while allowing some grace time for clock skew. However, if a security association has already been established based on a particular long-lived key, exceeding the lifetime does not have any direct impact. The implementations of security protocols that involve long-lived security associations should be designed to periodically interrogate the database and rollover to new keys without tearing down the security associations.
Rather than consulting the conceptual database, a security protocol such as TCP-AO may update its own tables as keys are added and removed. In this case, the protocol needs to maintain its own key information. Some routing protocols use IP Security (IPsec) to provide integrity. If a specification describes how to use the conceptual database described in this document to configure keys for these routing protocols, similar concerns apply. The specification mapping those routing protocols onto this conceptual database needs to describe how the Security Policy Database is manipulated as rows are added and removed from the conceptual database.
4. Application of the Database in a Security Protocol
In order to use the key table database in a protocol specification, a protocol needs to specify certain information. This section enumerates items that a protocol must specify.
(1) The ways of mapping the information in a key table row to the
information needed to produce an outgoing message; specified as an explanation of either how to fill in authentication- related fields in a message based on key table information, or (for protocols such as TCP-AO) how to construct Master Key Tuples (MKTs) or other protocol-specific structures from a key table row
(2) The ways of locating the peer identifier (a member of the
Peers set) and the LocalKeyName inside an incoming message
(3) The methods of verifying a message given a key table row;
this may be stated directly or in terms of protocol-specific structures such as MKTs
(4) The form and validation rules for LocalKeyName and
PeerKeyName; if either of these is an integer, the conventions in Section 5.1 are used as a vendor-independent format
(5) The form and validation rules for members of the Peers set
(6) The algorithms and KDFs supported
(7) The form of the ProtocolSpecificInfo field
(8) The rules for canonicalizing LocalKeyName, PeerKeyName,
entries in the Peers set, or ProtocolSpecificInfo; this may include normalizations such as lowercasing hexadecimal strings
(9) The Indication whether the support for Interfaces is required
by this protocol
The form of the interfaces field is not protocol specific but instead is shared among all protocols on an implementation. If a protocol needs to distinguish instances running over the same interface, this is included in the specification of peers. Generally, it is desirable to define the specification of peers so that an operator can use the Interfaces field to refer to all instances of a protocol on a link without having to specify both generic interfaces information and protocol-specific peer information.
5. Textual Conventions
5.1. Key Names
When a key for a given protocol is identified by an integer key identifier, the associated key name will be represented as lowercase hexadecimal digits with the most significant octet first. This integer is padded with leading zero digits until the width of the key identifier field in the protocol is reached. If a key name contains non-integer human-readable text, its format and encoding may be an issue, particularly if it is used in protocol between two different types of systems. If characters from the ASCII range [RFC20] are sufficient for a key name, then they SHOULD be used. If characters outside of that range are desirable or required, then they MUST be in an encoding of Unicode [UNICODE].
In the case of an AdminKeyName that uses characters outside of the ASCII range, the AdminKeyName MUST be encoded using UTF-8 [RFC3629] and SHOULD be normalized using Unicode Normalization Form KC [UAX15] to maximize the chance that the strings will compare correctly.
However, simply using Unicode Normalization Form KC is not sufficient to account for all issues of string comparison; refer to [PRECIS-FRAMEWORK] for additional information.
A key is represented as a lowercase hexadecimal string with the most significant octet of the key first. As discussed in Section 2, the length of this string depends on the associated algorithm and KDF.
6. Operational Considerations
If the valid periods for long-lived keys do not overlap or the system clocks are inconsistent, it is possible to construct scenarios where systems cannot agree upon a long-lived key. When installing a series of keys to be used one after another, operators should configure the SendLifetimeStart field of the key to be several hours after the AcceptLifeTimeStart field of the key to guarantee there is some overlap. This overlap is intended to address the clock-skew issue and allow for basic operational considerations. Operators may choose to specify a longer overlap (e.g., several days) to allow for exceptional circumstances.
7. Security Considerations
Management of encryption and authentication keys has been a significant operational problem, both in terms of key synchronization and key selection. For instance, the current guidance [RFC3562] warns against sharing TCP MD5 keying material between systems and recommends changing keys according to a schedule. The same general operational issues are relevant for the management of other cryptographic keys.
It has been recognized in [RFC4107] that automated key management is not viable in multiple scenarios. The conceptual database specified in this document is designed to accommodate both manual key management and automated key management. A future specification to automatically populate rows in the database is envisioned.
Designers should recognize the warning provided in [RFC4107]:
Automated key management and manual key management provide very different features. In particular, the protocol associated with an automated key management technique will confirm the liveness of the peer, protect against replay, authenticate the source of the short-term session key, associate protocol state information with the short-term session key, and ensure that a fresh short-term session key is generated. Further, an automated key management protocol can improve interoperability by including negotiation mechanisms for cryptographic algorithms. These valuable features are impossible or extremely cumbersome to accomplish with manual key management.
8. IANA Considerations
This specification defines three registries.
8.1. KeyTable Protocols
Per this document, IANA has established a registry called "KeyTable Protocols".
All assignments to the KeyTable Protocols registry are made on a Specification Required basis per Section 4.1 of [RFC5226].
Each registration entry must contain the three fields:
- Protocol Name (unique within the registry); - Protocol-Specific Info; and - Reference.
The specification needs to describe parameters required for using the conceptual database as outlined in Section 4. This typically means that the specification focuses more on the application of security protocols with the key tables rather than being a new security protocol specification for general purposes. Of course, new protocols may combine information on how to use the key table database with the protocol specification.
The registry has three columns. The first column is a string of Unicode characters encoded in UTF-8 representing the name protocol. The second column is a string of Unicode characters encoded in UTF-8 providing a brief description of Protocol-Specific Info. The third column is a reference to a specification defining how the protocol is used with the key table.
There are no initial registrations.
8.2. KeyTable KDFs
Per this document, IANA has established a registry called "KeyTable KDFs". The remainder of this section describes the registry.
All assignments to the KeyTable KDFs registry are made on a First Come First Served basis per Section 4.1 of RFC 5226.
The registry has three columns. The first column is a string of Unicode characters encoded in UTF-8 representing the name of a KDF. The second column is a string of Unicode characters encoded in UTF-8 providing a brief description of the KDF. The third column is a reference to a specification defining the KDF, if available.
The initial contents of this registry and that in Section 8.3 are chosen based on the algorithms defined for TCP-AO [RFC5926].
8.3. KeyTable AlgIDs
Per this document, IANA has established a registry called "KeyTable AlgIDs". The remainder of this section describes the registry.
All assignments to the KeyTable AlgIDs registry are made on a First Come First Served basis per Section 4.1 of RFC 5226.
The registry has three columns. The first column is a string of Unicode characters encoded in UTF-8 representing the algorithm identifier (AlgID). The second column is a string of Unicode characters encoded in UTF-8 providing a brief description of the identified algorithm. The third column is a reference to a specification defining the identified algorithm.
The initial contents of this registry and that in Section 8.2 are chosen based on the algorithms defined for TCP-AO [RFC5926].
This document reflects many discussions with many different people over many years. In particular, the authors thank Jari Arkko, Ran Atkinson, Ron Bonica, Ross Callon, Lars Eggert, Pasi Eronen, Adrian Farrel, Gregory Lebovitz, Acee Lindem, Sandy Murphy, Eric Rescorla, Mike Shand, Dave Ward, and Brian Weis for their insights. The authors additionally thank Brian Weis for supplying text to address IANA concerns and for help with formatting.
Sam Hartman's work on this document is funded by Huawei.
10. Normative References
[RFC20] Cerf, V., "ASCII format for network interchange", RFC 20, October 1969. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [UAX15] The Unicode Consortium, "Unicode Standard Annex #15: Unicode Normalization Forms", Unicode 6.3.0, September 2013, <http://www.unicode.org/reports/tr15/tr15-39.html>. [UNICODE] The Unicode Consortium, "The Unicode Standard, Version 6.3.0", (Mountain View, CA: The Unicode Consortium, 2013. ISBN 978-1-936213-08-5), <http://www.unicode.org/versions/Unicode6.3.0/>.
11. Informative References
Saint-Andre, P. and M. Blanchet, "PRECIS Framework: Preparation and Comparison of Internationalized Strings in Application Protocols", Work in Progress, March 2014. [RFC2104] Krawczyk, H., Bellare, M., and R. Canetti, "HMAC: Keyed- Hashing for Message Authentication", RFC 2104, February 1997. [RFC3562] Leech, M., "Key Management Considerations for the TCP MD5 Signature Option", RFC 3562, July 2003. [RFC3629] Yergeau, F., "UTF-8, a transformation format of ISO 10646", STD 63, RFC 3629, November 2003. [RFC4107] Bellovin, S. and R. Housley, "Guidelines for Cryptographic Key Management", BCP 107, RFC 4107, June 2005. [RFC4493] Song, JH., Poovendran, R., Lee, J., and T. Iwata, "The AES-CMAC Algorithm", RFC 4493, June 2006. [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 5226, May 2008. [RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP Authentication Option", RFC 5925, June 2010. [RFC5926] Lebovitz, G. and E. Rescorla, "Cryptographic Algorithms for the TCP Authentication Option (TCP-AO)", RFC 5926, June 2010.
Vigil Security, LLC
918 Spring Knoll Drive
Herndon, VA 20170
National Institute of Standards and Technology
100 Bureau Drive, Mail Stop 8930
Gaithersburg, MD 20899-8930
Sam Hartman Painless Security, LLC USA EMail: firstname.lastname@example.org Dacheng Zhang Huawei Technologies Co. Ltd. China EMail: email@example.com