This Key Management Interoperability Protocol Usage Guide
Version 2.0 is intended to complement the Key Management Interoperability
Protocol Specification [KMIP-Spec]
by providing guidance on how to implement the Key Management Interoperability
Protocol (KMIP) most effectively to ensure interoperability and to address key
management usage scenarios. It includes the following guidance:
·
Clarification of assumptions and requirements that drive or
influence the design of KMIP and the implementation of KMIP-compliant key
management.
·
Specific recommendations for implementation of KMIP
functionality.
·
Clarification of mandatory and optional capabilities for
conformant implementations.
·
Descriptions of how to use KMIP functionality to address specific
key management usage scenarios or to solve key management related issues.
A selected set of conformance profiles and authentication
suites are defined in the KMIP Profiles specification [KMIP-PROFILES].
Further assistance for implementing KMIP is provided by the
KMIP Test Cases document [KMIP-TC]
that describes a set of recommended test cases and provides the TTLV
(Tag/Type/Length/Value) format for the message exchanges defined by those test
cases.
1.1 References
(non-normative)
[KMIP-SPEC] Key Management Interoperability Protocol Specification
Version 2.0.
Edited by Tony Cox and Charles White. Latest version: https://docs.oasis-open.org/kmip/kmip-spec/v2.0/kmip-spec-v2.0.html.
[KMIP-PROF] Key Management Interoperability Protocol
Profiles Version 2.0.
Edited by Tim Hudson and Robert Lockhart. Latest version: https://docs.oasis-open.org/kmip/kmip-profiles/v2.0/kmip-profiles-v2.0.html.
[KMIP-TC] Key Management Interoperability
Protocol Test Cases Version 2.0. Edited by Tim Hudson and Mark Joseph. Latest version: https://docs.oasis-open.org/kmip/kmip-testcases/v2.0/kmip-testcases-v2.0.html.
[DoD5220.22M] National Industrial Security
Program Operating Manual, February 2006 (Incorporating Change 1 March 28,
2013), http://www.dtic.mil/whs/directives/corres/pdf/522022m.pdf
[ECC-Brainpool] ECC Brainpool Standard Curves
and Curve Generation v. 1.0.19.10.2005, http://www.ecc-brainpool.org/download/Domain-parameters.pdf.
[FIPS 180-4] Secure
Hash Standard (SHS), FIPS PUB 180-4, March 2012, http://csrc.nist.gov/publications/fips/fips180-4/fips-180-4.pdf
[FIPS 186-4] Digital Signature Standard
(DSS).
FIPS PUB 186-4. July 2013. http://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.186-4.pdf
[FIPS
197] Advanced Encryption Standard (AES). FIPS PUB 197. November 26,
2001. http://csrc.nist.gov/publications/fips/fips197/fips-197.pdf
[FIPS 198-1] The Keyed-Hash Message
Authentication Code (HMAC). FIPS PUB 198-1. July 2008. http://csrc.nist.gov/publications/fips/fips198-1/FIPS-198-1_final.pdf
[PKCS#1] RSA Laboratories. PKCS #1 v2.1: RSA
Cryptography Standard. June 14, 2002. http://www.preserveitall.org/emc-plus/rsa-labs/standards-initiatives/pkcs-rsa-cryptography-standard.htm.
[PKCS#10] RSA
Laboratories. PKCS #10 v1.7: Certification Request Syntax Standard. May 26,
2000. http://www.preserveitall.org/emc-plus/rsa-labs/standards-initiatives/pkcs10-certification-request-syntax-standard.htm.
[PKCS#11] OASIS PKCS#11 Cryptographic Token Interface Base Specification Version
3.0
[RFC1321] R. Rivest, The MD5 Message-Digest Algorithm, IETF RFC 1321, Apr
1992, http://www.ietf.org/rfc/rfc1321.txt
[RFC1421] J. Linn, Privacy Enhancement for Internet Electronic Mail:
Part I: Message
Encryption and Authentication Procedures, IETF RFC 1421, Feb 1993, http://www.ietf.org/rfc/rfc1421.txt
[RFC3647] S. Chokhani, W. Ford, R. Sabett, C.
Merrill, and S. Wu. RFC3647: Internet X.509 Public Key Infrastructure Certificate
Policy and Certification Practices Framework. November 2003. http://www.ietf.org/rfc/rfc3647.txt
[RFC4210] C. Adams, S. Farrell, T. Kause and T.
Mononen, Internet X.509 Public Key Infrastructure Certificate Management
Protocol (CMP), IETF RFC 2510, Sep 2005, http://www.ietf.org/rfc/rfc4210.txt
[RFC4211] J. Schaad, Internet X.509 Public Key Infrastructure Certificate
Request Message Format (CRMF), IETF RFC 4211, Sep 2005, http://www.ietf.org/rfc/rfc4211.txt
[RFC4949] R. Shirey. RFC4949: Internet Security
Glossary, Version 2. August 2007. http://www.ietf.org/rfc/rfc4949.txt
[RFC4880] J.
Callas, L. Donnerhacke, H. Finney, D. Shaw and R. Thayer. RFC4880: OpenPGP
Message Format. November 2007. http://www.ietf.org/rfc/rfc4880.txt
[RFC5272] J. Schaad and M. Meyers, Certificate
Management over CMS (CMC), IETF RFC 5272, Jun 2008, http://www.ietf.org/rfc/rfc5272.txt
[RFC5280] D. Cooper, S. Santesson, S. Farrell, S.
Boeyen, R. Housley, and W. Polk. RFC5280: Internet X.509 Public Key
Infrastructure Certificate and Certificate Revocation List (CRL) Profile.
May 2008. http://www.ietf.org/rfc/rfc5280.txt
[RFC5639] M. Lochter, J. Merkle, Elliptic Curve Cryptography (ECC) Brainpool
Standard Curves and Curve Generation, IETF RFC 5639, March 2010, http://www.ietf.org/rfc/rfc5639.txt.
[RFC7292] K. Moriarty, M. Nystrom, S.
Parkinson, A. Rusch, M. Scott, PKCS#12: Personal Information Exchange Syntax
v1.1, IETF RFC 7292, July 2014, https://tools.ietf.org/html/rfc7292
[RFC7748] Langley, A., Hamburg, M.,
and S. Turner, "Elliptic Curves for Security", RFC 7748, DOI
10.17487/RFC7748, January 2016, https://www.rfc-editor.org/info/rfc7748
[SEC2] SEC 2: Recommended Elliptic Curve Domain Parameters,
,http://www.secg.org/SEC2-Ver-1.0.pdf.
[SP800-38A] M. Dworkin. Recommendation
for Block Cipher Modes of Operation – Methods and Techniques. NIST Special
Publication 800-38A, Dec 2001. http://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-38a.pdf
[SP800-38D] M. Dworkin. Recommendation
for Block Cipher Modes of Operation: Galois/Counter Mode (GCM) and GMAC.
NIST Special Publication 800-38D. Nov 2007. http://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-38d.pdf.
[SP800-38E] M. Dworkin, Recommendation
for Block Cipher Modes of Operation: The XTS-AES Mode for Confidentiality on
Block-Oriented Storage Devices, NIST Special Publication 800-38E, January
2010, http://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-38e.pdf.
[SP800-56A] E.
Barker, L. Chen, A. Roginsky, and M. Smid, Recommendations for Pair-Wise Key
Establishment Schemes Using Discrete Logarithm Cryptography, NIST Special
Publication 800-56A Revision 2, May 2013, http://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-56Ar2.pdf
[SP800-57-1] E.
Barker, W. Barker, W. Burr, W. Polk and M. Smid, Recommendations for Key
Management – Part 1: General (Revision 3), NIST Special Publication 800-57
Part 1 Revision 3, July 2012, http://csrc.nist.gov/publications/nistpubs/800-57/sp800-57_part1_rev3_general.pdf
[SP800-67] W.
Barker and E. Barker, Recommendations for the Triple Data Encryption
Algorithm (TDEA) Block Cipher, NIST Special Publication 800-67 Revision 1,
January 2012, http://csrc.nist.gov/publications/nistpubs/800-67-Rev1/SP-800-67-Rev1.pdf
[SP800-88] R. Kissel, A. Regenscheid, M. Scholl, K. Stine, Guidelines for Media
Sanitization, NIST Special Publication 800-88 Revision 1, December 2014, http://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-88r1.pdf
[X.509] International Telecommunications Union
(ITU)-T, X.509: Information technology – Open systems interconnection – The
Directory: Public-key and attribute certificate frameworks, November 2008,
https://www.itu.int/rec/T-REC-X.509/recommendation.asp?lang=en&parent=T-REC-X.509-200811-S
[X9.31] ANSI, X9.31: Digital Signatures Using Reversible
Public Key Cryptography for the Financial Services Industry (rDSA).
September 1998.
[X9.42] ANSI, X9.42: Public Key Cryptography for the
Financial Services Industry: Agreement of Symmetric Keys Using Discrete
Logarithm Cryptography. 2003.
[X9.62] ANSI, X9.62: Public Key Cryptography for the
Financial Services Industry, The Elliptic Curve Digital Signature Algorithm
(ECDSA), 2005.
The section describes assumptions that underlie the KMIP
protocol and the implementation of clients and servers that utilize the
protocol.
Clients may be provided key material by the server, but they
only use that keying material for the purposes explicitly listed in the
delivery payload. Clients that ignore these instructions and use the keys in
ways not explicitly allowed by the server are non-compliant. There is no
requirement for the key management system, however, to enforce this behavior.
KMIP
relies on the chosen authentication suite as specified in [KMIP-PROF]
to authenticate the client and on the underlying transport protocol to provide
confidentiality, integrity, message authentication and protection against
replay attack. KMIP offers a wrapping mechanism for the Key Value that does not
rely on the transport mechanism used for the messages; the wrapping mechanism
is intended for importing or exporting managed cryptographic objects.
The
protocol operates on the assumption that the server is state-less, which means
that there is no concept of “sessions” inherent in the protocol. This does not
mean that the server itself maintains no state, only that the protocol does not
require it.
The
protocol provides for “private” or vendor-specific extensions, which allow for
differentiation among vendor implementations. However, any objects, attributes
and operations included in an implementation are always implemented as
specified in [KMIP-SPEC], regardless of whether
they are optional or mandatory.
A server is expected to be conformant to KMIP and supports
the conformance clauses as specified in [KMIP-SPEC].
However, a server may refuse a server-supported operation or
client-settable attribute if disallowed by the server policy (whether expressed
within or outside KMIP). Such a decision by the server may reflect the trust
relationship with a particular client, performance impact of the requested
operation, or any of a number of other considerations.
The protocol allows two modes of operation: synchronous and
asynchronous. Synchronous operations are those in which a client sends a
request and waits for a response from the server. Asynchronous operations (are
those in which the client sends a request, the server responds with a “pending”
status, and the client polls the server for the completed response and
completion status. Server implementations must support synchronous operations
but may choose not to support asynchronous operations.
2.7 Support for “Intelligent Clients” and
“Key Using Devices”
The protocol supports intelligent clients, such as end-user
workstations, which are capable of requesting all of the functions of KMIP. It
also allows subsets of the protocol and possible alternate message
representations in order to support less-capable devices, which only need a
subset of the features of KMIP.
2.8 Batched Requests and
Responses
The protocol contains a mechanism for sending batched
requests and receiving the corresponding batched responses, to allow for higher
throughput on operations that deal with a large number of entities, e. g.,
requesting dozens or hundreds of keys from a server at one time, and performing
operations in a group. A Batch Error Continuation option is provided to indicate
whether to undo all previous successful operations once a request in the batch
fails; to continue processing requests after an earlier request in the batch
fails; or to stop processing the remaining requests in the batch (but not undo
previously successful operations). Information may be pipelined between one or
more batch items using either the ID Placeholder mechanism (which involves
omitting the Unique Identifier) or via an indirect reference using an
enumerated type or an integer to indicate explicitly the origin of the value.
The
reliable message delivery function is relegated to the transport protocol and
is not part of the key management protocol itself.
For
requests that could result in large responses, a mechanism in the protocol
allows a client to specify in a request the maximum allowed size of a response,
or in the case of the Locate operation, the maximum number of items which
should be returned. The server indicates in a response to such a request that
the response would have been too large and, therefore, is not returned.
[KMIP-SPEC] describes the
key life-cycle model, based on the [SP800-57-1] key
state definitions, supported by the KMIP protocol. Particular implications of
the key life-cycle model in terms of defining time-related attributes of
objects are discussed in Section 3.15.
This section provides guidance on using the functionality
described in the Key Management Interoperability Protocol (KMIP) Specification.
3.1 Authentication
As discussed in [KMIP-SPEC], a
conforming KMIP implementation establishes and maintains channel
confidentiality and integrity and provides assurance of server authenticity for
KMIP messaging. Client authentication is performed according to the chosen KMIP
authentication suite as specified in [KMIP-PROF].
Other mechanisms for client and server authentication are possible and optional
for KMIP implementations.
KMIP
implementations that support the KMIP-defined Credential Types or use other
vendor-specific mechanisms for authentication may use the optional
Authentication structure specified inside the Request Header to include
additional identification information. Depending on the server’s configuration,
the server may interpret the identity of the requestor from the Credential
structure, contained in the Authentication structure if it is not provided
during the channel-level authentication. For example, in addition to performing
mutual authentication during a TLS handshake, the client passes the Credential
structure (e.g., a username and password) in the request. If the requestor’s
username is not specified inside the client certificate and is instead
specified in the Credential structure, the server interprets the identity of the
requestor from the Credential structure. This supports use cases where
channel-level authentication authenticates a machine or service that is used by
multiple users of the KMIP server. If the client provides the username of the
requestor in both the client certificate and the Credential structure, the
server verifies that the usernames are the same. If they differ, the
authentication fails, and the server returns an error. If no Credential
structure is included in the request, the username of the requestor is expected
to be provided inside the certificate. If no username is provided in the client
certificate and no Credential structure is included in the request message, the
server is expected to refuse authentication and return an error.
If authentication is unsuccessful, and it is possible to
return an “authentication not successful” error, this error should be returned
in preference to any other result status. This prevents status code probing by
a client that is not able to authenticate.
Server decisions regarding which operations to reject if
there is insufficiently strong authentication of the client are not specified
in the protocol. However, see Section 3.2 for operations for which
authentication and authorization are particularly important.
The Credential object defined in the [KMIP-SPEC]
is a structure used to convey information about the client, but the contents of
this object are not managed by the key management server. The type of
information convey within this object varies based on the type of credential
(e.g. Username and Password, Device Credential and Attestation).
[KMIP-SPEC] defines the
Username and Password structure for the Credential Type Username and Password.
The structure consists of two fields: Username and Password. Password is a
recommended, but optional, field, which may be excluded only if the client is
authenticated using one of the authentication suites defined in [KMIP-PROF] For example, if the client performs client
certificate authentication during the TLS handshake, and the Authentication
structure is provided in the Message Request, the Password field is an optional
field in the Username and Password structure of the Credential structure.
The Credential structure is used to provide additional
identification information. As described above, for certain use cases,
channel-level authentication may only authenticate a machine or service that is
used by multiple clients of the KMIP server. The Credential structure may be
used in this scenario to identify individual clients by specifying the username
in the Username and Password structure.
The Device Credential may be used to uniquely identify
back-end devices by specifying Device as the Credential Type in the Credential
structure.
The Device Credential may be used in a proxy environment
where the proxy authenticates with the client certificate and supports KMIP
while the back-end devices may not support KMIP or TLS. An example is
illustrated below:
Figure 1: Aggregator
Client Example
The end device identifies itself with a device unique set of
identifier values that include the device hardware serial number, the network
identifier, the machine identifier, or the media identifier. For many of the
self-encrypting devices there is a unique serial number assigned to the device
during manufacturing. The ability to use network, machine, or media identifier
explicitly should map to different device types and achieve better
interoperability since different types of identifier values are explicitly
enumerated. The device identifier is included for more generic usage. An
optional password or shared secret may be used to further authenticate the
device.
Server implementations may choose to enforce rules for
uniqueness for different types of identifier values, combinations of TLS
certificate used in combination with the Device Credential, and optionally
enforce the use of a Device Credential password.
Four identifiers are optionally provided but are unique in
aggregate:
1. Serial
Number, for example the hardware serial number of the device
2. Network
Identifier, for example the MAC address for Ethernet connected devices
3. Machine
Identifier, for example the client aggregator identifier, such as a tape
library aggregating tape drives
4. Media
Identifier, for example the volume identifier used for a tape cartridge
The device identifier by choice of server policy may or may
not be used in conjunction with the above identifiers to insure uniqueness.
These additional identifiers are generally useful for
auditing and monitoring encryption and could according to server policy be
logged or used in server implementation specific validation.
A specific example for self-encrypting tape drive and tape
library would be:
1. the tape
drive has a serial number that is unique for that manufacturer and the vendor
has procedures for maintaining and tracking serial number usage
2. a password
optionally is created and stored either on the drive or the library to help
authenticate the drive
3. the tape
drives may be connected via fiber channel to the library and therefore have a
World Wide Name assigned
4. a machine
identifier can be used to identify the tape library that is aggregating the
device in question
5. the media
identifier helps identify the individual media such as a tape cartridge for
proof of encryption reporting
Another example using self-encrypting disk drives inside of
a server would be:
1. the disk
drive has a unique serial number
2. a password
may be supplied by configuration of the drive or the server where the drive is
located
3. the
network identifier may come from the internal attachment identifier for the
disk drive in the server
4. the
machine identifier may come from a server’s motherboard or service processor
identifier,
5. and the
media identifier comes from the volume name used by the server’s operating
system to identify the volume on the disk drive
Server implementations could control what devices may read
and write keys and use the device credential fields to influence access control
enforcement.
Another example applied to server virtualization and
encryption built into virtualization would be:
1. the
virtual machine instance has a unique identifier that is used for the serial
number
2. the
hypervisor supplies a shared secret that is used as the password to
authenticate the virtual machine
3. the
network identifier could be used to identify the MAC address of the physical
server where the virtual machine is running
4. the
machine identifier could be used to identify the hypervisor
5. the media
identifier could be used to identify the storage volume used by the virtual
machine
These are examples of usage and are not meant to define all
device credential usage patterns nor restrict server specific implementations.
The device credentials may be explicitly added by the
administrator or may be captured in line with the request and implicitly registered
depending upon server policy.
When a server is not able to resolve the identifier values
in the device credential to a unique client identification, it may choose to
reject the request with an error code of operation failed and reason code of item
not found.
3.2 Authorization for Revoke, Recover, Destroy
and Archive Operations
The authentication suite, as specified in [KMIP-PROF], describes how the client identity is
established for KMIP-compliant implementations. This authentication is
performed for all KMIP operations.
Certain operations that may be requested by a client via
KMIP, particularly Revoke, Recover, Destroy and Archive, may have a significant
impact on the availability of a key, on server performance and/or on key
security. When a server receives a request for one of these operations, it
should ensure that the client has authenticated its identity (see the
Authentication Suites section in [KMIP-PROF]. The
server should also ensure that the client requesting the operation is an object
owner, security officer or other identity authorized to issue the request. It
may also require additional authentication to ensure that the object owner or a
security officer has issued that request. Even with such authentication and
authorization, requests for these operations should be considered only a “hint”
to the key management system, which may or may not choose to act upon this
request depending on server policy.
The key block returns a single object, with associated
attributes and other data. For those cases in which multiple related objects
are needed by a client, such as the private key and the related certificate,
the client should issue multiple Get requests to obtain these related objects.
A Cryptographic Object or Opaque Object may be compromised
for a variety of reasons. In KMIP, a client indicates to the server that a
Cryptographic Object is to be considered compromised by performing a Revoke
Operation with a Revocation Reason of Key Compromise or CA Compromise.
The KMIP client should provide a Compromise Occurrence Date (if the Revocation
Reason is Key Compromise or CA Compromise) and if it is unable to
estimate when the compromise occurred then it should provide a Compromise
Occurrence Date equal to the Initial Date.
The KMIP specification [KMIP-SPEC] places
no requirements on a KMIP server to perform any action on any Managed Object
that references (i.e., via Link attributes) a Cryptographic Object or Opaque
Object that a client has performed a Revoke operation with a Revocation Reason
of Key Compromise or CA Compromise. However, KMIP users should
be aware that there may be security relevant implications in continuing to use
a Managed Cryptographic Object in the following circumstances:
- For a compromised Private Key, the linked Public Key and/or
Certificate;
- For a compromised Public Key, the linked Private Key and/or
Certificate;
- For a compromised Derived Key, the linked derived key and/or Secret
Data Object
In these circumstances, it is the responsibility of the
client to either check the state of the referenced Managed Object or to also
perform a Revoke operation on the referenced Managed Object.
A number of operations are affected by a mechanism referred
to as the ID Placeholder. This is a temporary variable consisting of a single
Unique Identifier that is stored inside the server for the duration of
executing a batch of operations.
The Unique Identifier value when omitted indicates use of
the ID Placeholder. A general mechanism of specification of the Unique
Identifier either via an enumeration that indicates which Unique Identifier
value from the response batch item should be used or via an integer indicating
which batch item in the request should be referenced allows for additional
flexibility in pipelining requests.
The ID Placeholder is obtained from the Unique Identifier
returned by certain operations; the applicable operations are identified in Table 1, along with a list of operations that accept the ID Placeholder as input.
|
Operation
|
ID Placeholder at the
beginning of the operation
|
ID Placeholder upon
completion of the operation (in case of operation failure, a batch using the
ID Placeholder stops)
|
|
Create
|
-
|
ID of new Object
|
|
Create Key Pair
|
-
|
ID of new Private Key (ID of new
Public Key may be obtained via a Locate)
|
|
Create Split Key
|
-
|
ID of the split whose Key Part Identifier is 1
|
|
Join Split Key
|
|
ID of returned object
|
|
Register
|
-
|
ID of newly registered Object
|
|
Derive Key
|
- (multiple Unique Identifiers
may be specified in the request)
|
ID of new Symmetric Key
|
|
Locate
|
-
|
ID of located Object
|
|
Get
|
ID of Object
|
no change
|
|
Validate
|
-
|
-
|
|
Get Attributes
List/Modify/Add/Delete
|
ID of Object
|
no change
|
|
Activate
|
ID of Object
|
no change
|
|
Revoke
|
ID of Object
|
no change
|
|
Destroy
|
ID of Object
|
no change
|
|
Archive/Recover
|
ID of Object
|
no change
|
|
Certify
|
ID of Public Key
|
ID of new Certificate
|
|
Re-certify
|
ID of Certificate
|
ID of new Certificate
|
|
Re-key
|
ID of Symmetric Key to be
rekeyed
|
ID of new Symmetric Key
|
|
Re-key Key Pair
|
ID of Private Key to be
rekeyed
|
ID of new Private Key (ID of new
Public Key may be obtained via a Locate)
|
|
Obtain Lease
|
ID of Object
|
no change
|
|
Get Usage Allocation
|
ID of Key
|
no
change
|
|
Check
|
ID of Object
|
no
change
|
Table 1: ID
Placeholder Prior to and Resulting from a KMIP Operation
Meta-Data Only (MDO) keys are those Managed Key Objects for
which no Key Value is present. MDO objects can be one of the following:
Symmetric Keys, Private Keys, Split Keys, or Secret Data.
This may be a result of the KMIP client only wanting to
register information (Meta-Data) about the key with a Key Management System,
without having the key itself leave the client’s physical boundary. One such
example could be for keys created and stored within a Hardware Security Module
(HSM), with a policy that does not allow for the keys to leave its hardware.
In such cases, the KMIP client will not include a Key Value within the Key
Block during a Register operation, although it may optionally include a Key
Value Location attribute indicating the location of the Key Value instead. For
such keys, as part of the Register operation, the server will create a Key
Value Present attribute and set it to false to indicate the key value is not
stored on the server.
The KMIP protocol does not support the addition of a Key
Value to an existing MDO key object on the server. If for some reason the
client wanted to do this, it would have to carry out another Register operation
and create a new managed object with the Key Value.
Finally, because there is no Key Value associated with an
MDO key on the server, KMIP operations for Re-key, Re-key Key Pair and Derive
Key cannot be carried out on an MDO key object. An attempt to do so will return
an appropriate error as specified in the Error Handling section of [KMIP-SPEC].
There are mathematical relationships between certain
asymmetric cryptographic algorithms such as the Digital Signature Algorithm
(DSA) and Diffie-Hellman (DH) and their elliptic curve equivalents ECDSA and
ECDH that allow the same asymmetric key pair to be used in both algorithms. In
addition, there are overlaps in the key format used to represent the asymmetric
key pair for each algorithm type.
Even though a single key pair may be used in multiple
algorithms, the KMIP Specification has chosen to specify separate key formats
for representing the asymmetric key pair for use in each algorithm. This
approach keeps KMIP in line with the reference standards (e.g., NIST [FIPS 186-4], ANSI [X9.42], etc.)
from which the key formats are obtained and the best practice documents (e.g.,
NIST [SP800-57-1], NIST [SP800-56A]
etc.) which recommend that a key pair only be used for one purpose.
The KMIP protocol allows clients to register Secret Data
objects. Secret Data objects may include passwords or data that are used to
derive keys.
KMIP defines Secret Data as cryptographic objects. Even if
the object is not used for cryptographic purposes, clients may still set
certain attributes, such as the Cryptographic Usage Mask, for this object
unless otherwise stated. Similarly, servers set certain attributes for this
object, including the Digest, State, and certain Date attributes, even if the
attributes may seem relevant only for other types of cryptographic objects.
When registering a Secret Data object, the following
attributes are set by the server:
·
Unique Identifier
·
Object Type
·
Digest
·
State
·
Initial Date
·
Last Change Date
When registering a Secret Data object for non-cryptographic
purposes, the following attributes are set by either the client or the server:
·
Cryptographic Usage Mask
KMIP allow a client to register a Split Key that was created
or otherwise obtained by the client. To request that the server generate a
split key, the client sends a Create Split Key request that includes the Split
Key parameters (Split Key Parts, Split Key Threshold, Split Key Method) and
desired key attributes (e.g. Object Type, Cryptographic Length). If the client
supplies the Unique Identifier of an existing base key in a Create Split Key
request, the server will use the supplied key in the key splitting operation
instead of generating a new one. The server will respond with a list of Unique
Identifiers for the newly created Split Keys.
The client may want to add link
attributes to more easily locate the complete set of related Split Keys as
follows. The client adds a Previous Link from the Split Key with Key Part
Identifier K to the Split Key with Key Part Identifier K-1 and a Next Link to
the Split Key with Key Part Identifier K+1. Denoting the value of Split Key
Parts by N, the client adds a Previous Link from the Split Key with Key Part
Identifier 1 to the Split Key with Key Part Identifier N and a Next Link from
the Split Key with Key Part Identifier N to the Split Key with Key Part
Identifier 1. If the client supplies the Unique Identifier of an existing base
key in a Create Split Key request, the client may want to add a Parent Link
attribute from each newly generated Split Key to the base key that was supplied
in the Create Split Key request.
To request that the server recombine a set of split keys,
the client sends a Join Split Key request that includes the type of object to
be returned (e.g. Symmetric Key, Private Key, or Secret Data) and a list of
Unique Identifiers of the Split Keys to be combined. The number of Unique
Identifiers in the request needs to be at least the value of Split Key
Threshold in the Split Keys to ensure the server will be able to combine the
keys according to the Split Key Method. The server will respond with the Unique
Identifier of the key obtained by combining the provided Split Keys.
The Cryptographic Usage Mask attribute is intended to
support asymmetric concepts using symmetric keys. This is common practice in
established crypto systems: the MAC is an example of an operation where a
single symmetric key is used at both ends, but policy dictates that one end may
only generate cryptographic tokens using this key (the MAC) and the other end
may only verify tokens. The security of the system fails if the verifying end
can use the key to perform generation operations.
In these cases, it is not sufficient to describe the usage
policy on the keys in terms of cryptographic primitives like “encrypt” vs. “decrypt”
or “sign” vs. “verify”. There are two reasons why this is the case.
·
In some of these operations, such as MAC generation and
verification, the same cryptographic primitive is used in both of the
complementary operations. MAC generation involves computing and returning the
MAC, while MAC verification involves computing that same MAC and comparing it
to a supplied value to determine if they are the same. Thus, both generation
and verification use the “encrypt” operation, and the two usages are not able to
be distinguished by considering only “encrypt” vs. “decrypt”.
·
Some operations which require separate key types use the same
fundamental cryptographic primitives. For example, encryption of data,
encryption of a key, and computation of a MAC all use the fundamental operation
“encrypt”, but in many applications, securely differentiated keys are used for
these three operations. Simply looking for an attribute that permits “encrypt”
is not sufficient.
Allowing the use of these keys outside of their specialized purposes
may compromise security. Instead, specialized application-level permissions are
necessary to control the use of these keys. KMIP provides several pairs of such
permissions in the Cryptographic Usage Mask, such as:
|
MAC GENERATE
MAC VERIFY
|
For cryptographic MAC operations. Although it is possible
to compose certain MACs using a series of encrypt calls, the security of the
MAC relies on the operation being atomic and specific.
|
|
GENERATE CRYPTOGRAM
VALIDATE CRYPTOGRAM
|
For composite cryptogram operations such as financial CVC
or ARQC. To specify exactly which cryptogram the key is used for it is also
necessary to specify a role for the key
|
|
TRANSLATE ENCRYPT
TRANSLATE DECRYPT
TRANSLATE WRAP
TRANSLATE UNWRAP
|
To accommodate secure routing of traffic and data. In
many areas that rely on symmetric techniques (notably, but not exclusively
financial networks), information is sent from place to place encrypted using
shared symmetric keys. When encryption keys are changed, it is desirable for
the change to be an atomic operation, otherwise distinct unwrap-wrap or
decrypt-encrypt steps risk leaking the plaintext data during the translation
process.
TRANSLATE ENCRYPT/DECRYPT is used for data
encipherment.
TRANSLATE WRAP/UNWRAP
is used for key wrapping.
|
Table 2: Cryptographic
Usage Masks Pairs
The protocol uses the Key Block structure to transport a key
to the client or server. This Key Block consists of the Key Value Type, the Key
Value, and the Key Wrapping Data. The Key Value Type identifies the format of
the Key Material, e.g., Raw format or Transparent Key structure. The Key Value
consists of the Key Material and optional attributes. The Key Wrapping Data
provides information about the wrapping key and the wrapping mechanism and is
returned only if the client requests the Key Value to be wrapped by specifying
the Key Wrapping Specification inside the Get Request Payload. The Key Wrapping
Data may also be included inside the Key Block if the client registers a
wrapped key.
The protocol allows any attribute to be included inside the
Key Value and allows these attributes to be cryptographically bound to the Key
Material (i.e., by signing, MACing, encrypting, or both encrypting and
signing/MACing the Key Value). Some of the attributes that may be included
include the following:
·
Unique Identifier – uniquely identifies the key
·
Cryptographic Algorithm (e.g., AES, 3DES, RSA) – this attribute
is either specified inside the Key Block structure or the Key Value structure
·
Cryptographic Length (e.g., 128, 256, 2048) – this attribute is
either specified inside the Key Block structure or the Key Value structure
·
Cryptographic Usage Mask– identifies the cryptographic usage of
the key (e.g., Encrypt, Wrap Key, Export)
·
Cryptographic Parameters – provides additional parameters for
determining how the key may be used
· Block Cipher
Mode (e.g., CBC, NISTKeyWrap, GCM) – this parameter identifies the mode of
operation, including block cipher-based MACs or wrapping mechanisms
· Padding Method
(e.g., OAEP, X9.31, PSS) – identifies the padding method and if applicable the
signature or encryption scheme
· Hashing
Algorithm (e.g., SHA-256) – identifies the hash algorithm to be used with the
signature/encryption mechanism or Mask Generation Function; note that the
different HMACs are defined individually as algorithms and do not require the
Hashing Algorithm parameter to be set
· Key Role Type
– Identifies the functional key role (e.g., DEK, KEK)
·
State (e.g., Active)
·
Dates (e.g., Activation Date, Process Start Date, Protect Stop
Date)
·
Custom Attribute – allows vendors and clients to define
vendor-specific attributes; may also be used to prevent replay attacks by
setting a nonce
Two parties receiving the same key as a Key Value Byte
String make use of the key in exactly the same way in order to interoperate. To
ensure that, it is necessary to define a correspondence between the abstract
syntax of Key and the notation in the standard algorithm description that
defines how the key is used. The next sections establish that correspondence
for the algorithms AES [FIPS 197] and Triple-DES [SP800-67].
AES Key Encoding [FIPS 197] section
5.2, titled Key Expansion, uses the input key as an array of bytes indexed
starting at 0. The first byte of the Key becomes the key byte in AES that is
labeled index 0 in [FIPS 197] and the other key
bytes follow in index order.
Proper parsing and key load of the contents of the Key for
AES is determined by using the following Key byte string to generate and match
the key expansion test vectors in [FIPS 197] Appendix A
for the 128-bit (16 byte) AES Cipher Key: 2B 7E 15 16 28 AE D2 A6 AB F7 15 88
09 CF 4F 3C.
A Triple-DES key consists of three keys for the
cryptographic engine (Key1, Key2, and Key3) that are each 64 bits (even though
only 56 are used); the three keys are also referred to as a key bundle (KEY) [SP800-67]. A key bundle may employ either two or three
mutually independent keys. When only two are employed (called two-key
Triple-DES), then Key1 = Key3.
Each key in a Triple-DES key bundle is expanded into a key
schedule according to a procedure defined in [SP800-67]
Appendix A. That procedure numbers the bits in the key from 1 to 64, with
number 1 being the left most, or most significant bit. The first byte of the
Key is bits 1 through 8 of Key1, with bit 1 being the most significant bit. The
second byte of the Key is bits 9 through 16 of Key1, and so forth, so that the
last byte of the KEY is bits 57 through 64 of Key3 (or Key2 for two-key
Triple-DES).
Proper parsing and key load of the contents of Key for
Triple-DES is determined by using the following Key byte string to generate and
match the key expansion test vectors in [SP800-67]
Appendix B for the key bundle:
Key1 = 0123456789ABCDEF
Key2 = 23456789ABCDEF01
Key3 = 456789ABCDEF0123
AES-XTS
is an implementation of the AES encryption algorithm with the XTS (XEX (XOR
Encrypt XOR) Tweakable Block Cipher with Ciphertext Stealing) cipher mode (see [SP800-38E]). What is unique about the XTS mode is that
it requires the use of two AES keys. AES-XTS 512 requires two 256-bit AES keys
while AES-XTS 256 requires two 128-bit AES keys.
KMIP implementations should treat the required XTS keys as
two distinct keys (e.g. two 256-bit AES keys). KMIP operations (e.g. Create,
Register, Get) associated with these two keys are batched within the profile to
logically keep the handling of these associated keys together. A simple custom
attribute (X-ID) is used to specify a naming convention that associates the two
AES keys. The Link attribute (with Link Types of Next Link and Previous Link)
is used to associate the two AES keys with one another on the KMIP server.
The KMIP Specification [KMIP-SPEC]
specifies a number of ECC recommended curves ([FIPS 186-4]
[SEC2] [X9.62] [ECC-Brainpool] [RFC5639] [RFC7748]). These recommended curves are
defined in multiple source documents and in some cases, the same algorithm is
known by multiple names since the algorithm is defined in multiple documents.
The following table provides a mapping of the ECC recommended curves specified
in the KMIP specification [KMIP-SPEC]. The table
identifies the KMIP enumeration, the Object Identifier (OID) and multiples
names (synonyms) for the ECC recommended curves.
|
Recommended
Curve Name
|
KMIP Enumeration
Value
|
OID
|
Recommended
Curve Synonym(s)
|
|
P-192
|
00000001
|
1.2.840.10045.3.1.1
|
SECP192R1
ANSIX9P192V1
|
|
K-163
|
00000002
|
1.3.132.0.1
|
SECT163K1
|
|
B-163
|
00000003
|
1.3.132.0.15
|
SECT163R2
|
|
P-224
|
00000004
|
1.3.132.0.33
|
SECP224R1
|
|
K-233
|
00000005
|
1.3.132.0.26
|
SECT233K1
|
|
B-233
|
00000006
|
1.3.132.0.27
|
SECT233R1
|
|
P-256
|
00000007
|
1.2.840.10045.3.1.7
|
SECP256R1
ANSIX9P256V1
|
|
K-283
|
00000008
|
1.3.132.0.16
|
SECT283K1
|
|
B-283
|
00000009
|
1.3.132.0.17
|
SECT283R1
|
|
P-384
|
0000000A
|
1.3.132.0.34
|
SECP384R1
|
|
K-409
|
0000000B
|
1.3.132.0.36
|
SECT409K1
|
|
B-409
|
0000000C
|
1.3.132.0.37
|
SECT409R1
|
|
P-521
|
0000000D
|
1.3.132.0.35
|
SECP521R1
|
|
K-571
|
0000000E
|
1.3.132.0.38
|
SECT571K1
|
|
B-571
|
0000000F
|
1.3.132.0.39
|
SECT571R1
|
|
SECP112R1
|
00000010
|
1.3.132.0.6
|
|
|
SECP112R2
|
00000011
|
1.3.132.0.7
|
|
|
SECP128R1
|
00000012
|
1.3.132.0.28
|
|
|
SECP128R2
|
00000013
|
1.3.132.0.29
|
|
|
SECP160K1
|
00000014
|
1.3.132.0.9
|
|
|
SECP160R1
|
00000015
|
1.3.132.0.8
|
|
|
SECP160R2
|
00000016
|
1.3.132.0.30
|
|
|
SECP192K1
|
00000017
|
1.3.132.0.31
|
|
|
SECP192R1
|
00000001
|
1.2.840.10045.3.1.1
|
P-192
ANSIX9P192V1
|
|
SECP224K1
|
00000018
|
1.3.132.0.32
|
|
|
SECP224R1
|
00000004
|
1.3.132.0.33
|
P-224
|
|
SECP256K1
|
00000019
|
1.3.132.0.10
|
|
|
SECP256R1
|
00000007
|
1.2.840.10045.3.1.7
|
P-256
ANSIX9P256V1
|
|
SECP384R1
|
0000000A
|
1.3.132.0.34
|
P-384
|
|
SECP521R1
|
0000000D
|
1.3.132.0.35
|
P-521
|
|
SECT113R1
|
0000001A
|
1.3.132.0.4
|
|
|
SECT113R2
|
0000001B
|
1.3.132.0.5
|
|
|
SECT131R1
|
0000001C
|
1.3.132.0.22
|
|
|
SECT131R2
|
0000001D
|
1.3.132.0.23
|
|
|
SECT163K1
|
00000002
|
1.3.132.0.1
|
K-163
|
|
SECT163R1
|
0000001E
|
1.3.132.0.2
|
|
|
SECT163R2
|
00000003
|
1.3.132.0.15
|
B-163
|
|
SECT193R1
|
0000001F
|
1.3.132.0.24
|
|
|
SECT193R2
|
00000020
|
1.3.132.0.25
|
|
|
SECT233K1
|
00000005
|
1.3.132.0.26
|
K-233
|
|
SECT233R1
|
00000006
|
1.3.132.0.27
|
B-233
|
|
SECT239K1
|
00000021
|
1.3.132.0.3
|
|
|
SECT283K1
|
00000008
|
1.3.132.0.16
|
K-283
|
|
SECT283R1
|
00000009
|
1.3.132.0.17
|
B-283
|
|
SECT409K1
|
0000000B
|
1.3.132.0.36
|
K-409
|
|
SECT409R1
|
0000000C
|
1.3.132.0.37
|
B-409
|
|
SECT571K1
|
0000000E
|
1.3.132.0.38
|
K-571
|
|
SECT571R1
|
0000000F
|
1.3.132.0.39
|
B-571
|
|
ANSIX9P192V1
|
00000001
|
1.2.840.10045.3.1.1
|
P-192
SECP192R1
|
|
ANSIX9P192V2
|
00000022
|
1.2.840.10045.3.1.2
|
|
|
ANSIX9P192V3
|
00000023
|
1.2.840.10045.3.1.3
|
|
|
ANSIX9P239V1
|
00000024
|
1.2.840.10045.3.1.4
|
|
|
ANSIX9P239V2
|
00000025
|
1.2.840.10045.3.1.5
|
|
|
ANSIX9P239V3
|
00000026
|
1.2.840.10045.3.1.6
|
|
|
ANSIX9P256V1
|
00000007
|
1.2.840.10045.3.1.7
|
P-256
SECP256R1
|
|
ANSIX9C2PNB163V1
|
00000027
|
1.2.840.10045.3.0.1
|
|
|
ANSIX9C2PNB163V2
|
00000028
|
1.2.840.10045.3.0.2
|
|
|
ANSIX9C2PNB163V3
|
00000029
|
1.2.840.10045.3.0.3
|
|
|
ANSIX9C2PNB176V1
|
0000002A
|
1.2.840.10045.3.0.4
|
|
|
ANSIX9C2TNB191V1
|
0000002B
|
1.2.840.10045.3.0.5
|
|
|
ANSIX9C2TNB191V2
|
0000002C
|
1.2.840.10045.3.0.6
|
|
|
ANSIX9C2TNB191V3
|
0000002D
|
1.2.840.10045.3.0.7
|
|
|
ANSIX9C2PNB208W1
|
0000002E
|
1.2.840.10045.3.0.10
|
|
|
ANSIX9C2TNB239V1
|
0000002F
|
1.2.840.10045.3.0.11
|
|
|
ANSIX9C2TNB239V2
|
00000030
|
1.2.840.10045.3.0.12
|
|
|
ANSIX9C2TNB239V3
|
00000031
|
1.2.840.10045.3.0.13
|
|
|
ANSIX9C2PNB272W1
|
00000032
|
1.2.840.10045.3.0.16
|
|
|
ANSIX9C2PNB304W1
|
00000033
|
1.2.840.10045.3.0.17
|
|
|
ANSIX9C2TNB359V1
|
00000034
|
1.2.840.10045.3.0.18
|
|
|
ANSIX9C2PNB368W1
|
00000035
|
1.2.840.10045.3.0.19
|
|
|
ANSIX9C2TNB431R1
|
00000036
|
1.2.840.10045.3.0.20
|
|
|
BRAINPOOLP160R1
|
00000037
|
1.3.36.3.3.2.8.1.1.1
|
|
|
BRAINPOOLP160T1
|
00000038
|
1.3.36.3.3.2.8.1.1.2
|
|
|
BRAINPOOLP192R1
|
00000039
|
1.3.36.3.3.2.8.1.1.3
|
|
|
BRAINPOOLP192T1
|
0000003A
|
1.3.36.3.3.2.8.1.1.4
|
|
|
BRAINPOOLP224R1
|
0000003B
|
1.3.36.3.3.2.8.1.1.5
|
|
|
BRAINPOOLP224T1
|
0000003C
|
1.3.36.3.3.2.8.1.1.6
|
|
|
BRAINPOOLP256R1
|
0000003D
|
1.3.36.3.3.2.8.1.1.7
|
|
|
BRAINPOOLP256T1
|
0000003E
|
1.3.36.3.3.2.8.1.1.8
|
|
|
BRAINPOOLP320R1
|
0000003F
|
1.3.36.3.3.2.8.1.1.9
|
|
|
BRAINPOOLP320T1
|
00000040
|
1.3.36.3.3.2.8.1.1.10
|
|
|
BRAINPOOLP384R1
|
00000041
|
1.3.36.3.3.2.8.1.1.11
|
|
|
BRAINPOOLP384T1
|
00000042
|
1.3.36.3.3.2.8.1.1.12
|
|
|
BRAINPOOLP512R1
|
00000043
|
1.3.36.3.3.2.8.1.1.13
|
|
|
BRAINPOOLP512T1
|
00000044
|
1.3.36.3.3.2.8.1.1.14
|
|
|
CURVE25519
|
00000045
|
1.3.101.110
|
X25519
ed25519
|
|
CURVE448
|
00000046
|
1.3.101.111
|
X448
|
Table 3: ECC
Recommended Curve Mapping
All managed objects within KMIP have associated attributes.
Initial attribute values are provided during object creation or registration
(via Create, CreateKeyPair, Certify, DeriveKey, ReKey, ReKeyKeyPair, ReCertify,
and Register).
Attributes have associated rules. They may be mandatory or
optional. They may be single instance, or multi-instance. They may be set or
updated only by the client or by the server. Within the Specification, each
attribute section contains a table defining these rules.
Specific operations allow for working with attributes.
·
Attribute values may be created using either Add Attribute or Set
Attribute
·
Attribute values may be modified using either Modify Attribute or
Set Attribute.
·
Set Attribute is more appropriate for single-instance attributes
as it does not require knowledge of whether the attribute previously had a
value. Set Attribute cannot be used for multi-instance attributes.
·
Both multi-instance and single-instance attributes may be removed
using Delete Attributes.
·
Instances of multi-instance attributes are identified by
specifying their current value when using either Modify Attribute of Delete
Attribute.
·
Adjust Attribute should be used where applicable, both because it
is more convenient and is transaction-safe.
[KMIP-SPEC] provides a number of
time-related attributes, including the following:
- Initial Date: The date and time when the managed cryptographic object
was first created by or registered at the server.
- Activation Date: The date and time when the managed cryptographic object
should begin to be used for applying cryptographic protection to data.
- Process Start Date: The date and time when a managed symmetric key object
should begin to be used for processing cryptographically protected data.
The managed symmetric key object should not be used prior to this date.
- Protect Stop Date: The date and time when a managed symmetric key object
should no longer be used for applying cryptographic protection to data
- Deactivation Date: The date and time when the managed cryptographic object
should no longer be used for applying
cryptographic protection (e.g., encryption, signing, wrapping, MACing,
deriving). Under extraordinary circumstances and when special
permission is granted the managed symmetric key object can be used for
decryption, signature verification, unwrapping, or MAC verification,
- Destroy Date: The date and time when the managed cryptographic object
was destroyed
- Compromise Occurrence Date: The date and time when the managed cryptographic
object was first believed to be compromised.
- Compromise Date: The date and time when the managed cryptographic object
entered the compromised state.
- Archive Date: The date and time when the managed object was placed
in Off-Line storage.
These attributes apply to all cryptographic objects (symmetric
keys, asymmetric keys, etc.) with exceptions as noted in [KMIP-SPEC].
However, certain of these attributes (such as the Initial Date) are not
specified by the client and are implicitly set by the server.
In using these attributes, the following guidelines should
be observed:
- As discussed for each of these attributes in [KMIP-SPEC],
a number of these times are set once, and it is not possible for the
client or server to modify them. However, several of the time attributes
(particularly the Activation Date, Protect Start Date, Process Stop Date
and Deactivation Date) may be set by the server and/or requested by the
client. Coordination of time-related attributes between client and server,
therefore, is primarily the responsibility of the server, as it manages
the cryptographic object and its state. However, special conditions
related to time-related attributes, governing when the server accepts
client modifications to time-related attributes, may be communicated
out-of-band between the client and server outside the scope of KMIP.
- In general, state transitions occur as a result of operational
requests, such as Create, Create Key Pair, Register, Activate, Revoke, and
Destroy. However, clients may need to specify times in the future for such
things as Activation Date, Deactivation Date, Process Start Date, and
Protect Stop Date. KMIP also allows clients to specify times in the past
for such attributes as Activation Date and Deactivation Date..
- It is valid to have a projected Deactivation Date when there is no
Activation Date. This means, however, that the key is not yet active, even
though its projected Deactivation Date has been specified. A valid
Deactivation Date is greater than or equal to the Activation Date (if the
Activation Date has been set).
- The Protect Stop Date may be equal to but may not be later than the
Deactivation Date. Similarly, the Process Start Date may be equal to, but
may not precede, the Activation Date. KMIP implementations should consider
specifying both these attributes, particularly for symmetric keys, as a
key may be needed for processing protected data (e.g., decryption) long
after it is no longer appropriate to use it for applying cryptographic
protection to data (e.g., encryption).
- KMIP does not allow an Active object to be destroyed with the
Destroy operation. The server returns an error, if the client invokes the
Destroy operation on an Active object. To destroy an Active object,
clients first call the Revoke operation or explicitly set the Deactivation
Date of the object. Once the object is in Deactivated state, clients may destroy
the object by calling the Destroy operation. These operations may be
performed in a batch. If other time-related attributes (e.g., Protect Stop
Date) are set to a future date, the server should set these to the
Deactivation Date.
- After a cryptographic object is destroyed, a key management server
may retain certain information about the object, such as the Unique
Identifier.
KMIP does not support server mutation of client-supplied
attributes. If a server does not accept an attribute value that is being
specified inside the request by the client, the server returns an error and
specifies “Invalid Field” as Result Reason.
If a client sets a time-related attribute to the current
date and time (as perceived by the client), but as a result of a clock skew,
the specified date of the attribute is earlier than the time perceived by the
server, the server’s policy is used to determine whether to accept the “backdated
attribute”. KMIP does not require the server to fail a request if a backdated
attribute is set by the client.
If a server does not support backdated attributes, and
cryptographic objects are expected to change state at the specified current
date and time (as perceived by the client), clients are recommended to issue
the operation that would implicitly set the date for the client. For example,
instead of explicitly setting the Activation Date, clients could issue the
Activate operation. This would require the server to set the Activation Date to
the current date and time as perceived by the server.
If it is not possible to set a date attribute via an
operation, and the server does not support backdated attributes, clients need
to take into account that potential clock skew issues may cause the server to
return an error even if a date attribute is set to the client’s current date
and time.
For additional information, refer to the sections describing
the State attribute and the Time Stamp field in [KMIP-SPEC].
The Application Specific Information attribute is used to
store data which is specific to the application(s) using the object. Some
examples of Application Namespace and Application Data pairs are given below.
·
SMIME, 'someuser@company.com'
·
TLS, 'some.domain.name'
·
Volume Identification, '123343434'
·
File Name, 'secret.doc'
·
Client Generated Key ID, ‘450994003'
The following Application Namespaces are recommended:
·
SMIME
·
TLS
·
IPSEC
·
HTTPS
·
PGP
·
Volume Identification
·
File Name
·
LTO4, LTO5, and LTO6
·
LIBRARY-LTO, LIBRARY-LTO4, LIBRARY-LTO5 and LIBRARY-LTO6
KMIP provides optional support for server-generated
Application Data. Clients may request the server to generate the Application
Data for the client by omitting Application Data while setting or modifying the
Application Specific Information attribute. A server only generates the
Application Data if the Application Data is completely omitted from the
request, and the client-specified Application Namespace is recognized and
supported by the server. An example for requesting the server to generate the
Application Data is shown below:
AddAttribute(Unique ID,
AppSpecInfo{AppNameSpace=’LIBRARY-LTO4’});
If the server does not recognize the namespace, the “Application
Namespace Not Supported” error is returned to the client.
If the Application Data is provided, and the Application
Namespace is recognized by the server, the server uses the provided Application
Data, and does not generate the Application Data for the client. In the example
below, the server stores the Application Specific Information attribute with
the Application Data value set to null.
AddAttribute(Unique ID,
AppSpecInfo{AppNameSpace=’LIBRARY-LTO4’, AppData=null});
The Description and Comment attributes are used to convey
information about the purpose and use of an object. Description is intended to
be a concise imperative statement about an object (e.g. “Root Key for internal
servers”). Comment is a complementary field which allows more verbose
communication regarding what activity or why an activity applies to an object
(e.g. “Make sure to update new internal servers with the Root key as they are
brought on line”). Both attributes are optional, and their contents should be
used for informational purposes and not for policy enforcement.
The value (e.g., 2048 bits) referred to in the KMIP Cryptographic
Length attribute for an asymmetric (public or private) key may be misleading,
since this length only refers to certain portions of the mathematical values
that comprise the key. The actual length of all the mathematical values
comprising the public or the private key is longer than the referenced value.
This point may be illustrated by looking at the components of a RSA public and
private key.
The RSA public key is comprised of a modulus (n) and an
(public) exponent (e). When one indicates that the RSA public key is 2048 bits
in length that is a reference to the bit length of the modulus (n) only. So
the full length of the RSA public key is actually longer than 2048 bits, since
it also includes the length of the exponent (e) and the overhead of the
encoding (e.g., ASN.1) of the key material.
The RSA private key is comprised of a modulus (n), the
public exponent (e), the private exponent (d), prime 1 (p), prime 2 (q),
exponent 1 (d mod (p-1)), exponent 2 (d mod (p-1)), and coefficient ((inverse
of q) mod p). Once again, the 2048 bit key length is referring only to the
length of the modulus (n), so the overall length of the private key would be
longer given the number of additional components which comprise the key and the
overhead of encoding (e.g., ASN.1) of the key material.
KMIP implementations need to ensure they do not make
assumptions about the actual length of asymmetric (public and private) key
material based on the value specified in the Cryptographic Length
attribute.
There are certain cryptographic parameters that the server
must know in order to create objects. These parameters are the
CryptographicAlgorithm (e.g., AES), the CryptographicLength (e.g., 128) and the
CryptographicUsageMask. These parameters can be provided to the server by the
clientServers may supply defaults for some or all of these parameters, and if a
client omits them, the server will fill in the values and satisfy the client
request rather than failing the request. The server defaults are discoverable
via Query Defaults Information.
The Digest attribute contains the output of hashing a
managed object, such as a key or a certificate. The server always generates the
SHA-256 hash value when the object is created or generated. KMIP allows
multiple instances of the digest attribute to be associated with the same
managed object. For example, it is common practice for publicly trusted CAs to
publish two digests (often referred to as the fingerprint or the thumbprint) of
their certificate: one calculated using the SHA-1 algorithm and another using
the MD5 algorithm. In this case, each digest would be calculated by the server
using a different hash algorithm.
KMIP supports attributes which can be used to indicate the
sensitivity of a cryptographic object and whether a cryptographic object may
leave the server. These attributes are modeled after similar attributes within
[PKCS#11].
The Sensitive attribute is used to specify whether a
cryptographic object can only leave the KMIP server (via the Get operation) in
wrapped (encrypted) form.
The Always Sensitive attribute is used to track
whether a cryptographic object has ever been considered sensitive during its
existence within the KMIP server.
The Extractable attribute is used to control whether
a cryptographic object is ever allowed to leave the KMIP server.
The Never Extractable attribute is used to track
whether extraction of a cryptographic object has been restricted during its
existence within the KMIP server.
A client may wish to use the digest attribute of an object
to compare objects. If a client does not specify a key format type on a
constructor, the server may choose any of a number of key format types to use
to encode that object. If the server chooses a format that the client cannot
handle, then the client will be unable to use the digest of that format in any
meaningful way. The key format attribute was added so that the client can tell
the server exactly which format to use else fail the construction. The server is
free to add digests of other formats.
The “raw” key format is intended to be used for “a key that
contains only cryptographic key material, encoded as a string of bytes”. The
“raw” key format supports situations such as “non-KMIP-aware end-clients are
aware how wrapped cryptographic objects (possibly Raw keys) from the KMIP
server should be used without having to rely on the attributes provided by the
Get Attributes operation” and in that regard is similar to the Opaque key format
type. “Raw” key format is intended to be applied to symmetric keys and
not asymmetric keys.
PKCS#12 was originally introduced in the early days of PKI
and is used as a keystore format and as a key and certificate exchange mechanism.
The addition of PKCS#12 to KMIP as a key format type is intended as a means of
admitting the basic forms of PKCS#12 into a managed environment, as well as
extracting basic forms of PKCS#12 from the KMIP world to be consumed/used in
existing environments and workflows.
The intent of KMIP’s PKCS#12 key format is that a
PKCS#12-conformant blob can be Registered by the client, and a
PKCS#12-conformant blob can be returned from a KMIP server on a Get. It is
presumed that a single private key is what is being Registered (along with its
related certificate or certificate chain), so the Unique Identifier that KMIP
servers return on a Register PKCS#12 operation is that of the private key. It
is also presumed that the client doing a Get with a key format type of PKCS#12
will be passing in the private key’s Unique Identifier, and that the KMIP
server is responsible for returning a PKCS#12 blob with the referenced private
key and any related certificates.
If the PKCS#12 blob being Registered has multiple private keys,
then the client can indicate which ONE of them is being Registered on a
particular request by specifying the PKCS#12 Friendly Name attribute. If the
client does not supply the Friendly Name (a.k.a. “alias”), then the server is
free to just Register the first that it finds in the PKCS#12 blob. If the
client wants more than one private key within a PKCS#12 blob to be Registered
from a single PKCS#12 blob, then it will have to request repeatedly with that
PKCS#12 blob using a different Friendly Name for each request.
Those familiar with PKCS#12 objects, particularly those
containing sensitive material such as private keys, know that such objects are
usually password-protected. KMIP honors such constraints by requiring
SecretData objects to be referenced and used for the normal password-based
encryption (PBE) schemes employed to protect PKCS#12 objects. Hence Register
requests for objects with PKCS#12 key format must supply a Link attribute with
a LinkType of “PKCS#12 Password Link”, so that the server can find the
SecretData object containing the password for the PKCS#12 blob being sent to
the server and decrypt and verify its contents. Such a SecretData object must
be in the Active state and have a Cryptographic Usage Mask that indicates
“DeriveKey” for it to be used for password-based key derivation. At the end of
the Register operation, one can expect the KMIP server to have decrypted and
verified the contents of the PKCS#12 blob, registered the (specified or
defaulted) private key, added the PKCS#12 Password Link to the SecretData in
the new PrivateKey, found and Registered the certificate(s) that make up the
private key’s certificate chain, and linked these Certificates from the
PrivateKey via the PKCS#12 CertificateLink (the link to the first Certificate
is from the PrivateKey; any subsequent Certificates would be linked by a
PKCS#12 Certificate Link in the previous Certificate).
When a client issues a Get PKCS#12 key format request on a
PrivateKey Unique Identifier, the KMIP server traverses all the links set up on
the Register to reassemble the PKCS#12 blob and protect it with the SecretData
specified on the PrivateKey PKCS#12 Password Link. If the client wants a
different password on the PKCS#12 blob resulting from a Get, it should modify
the PKCS#12 Password Link on the PrivateKey to a different SecretData before
performing the Get. Similarly, if the client wants a different Friendly Name
for the key entry in a PKCS#12 blob, it should modify the PKCS#12 Friendly Name
attribute in the PrivateKey object before performing the Get.
A client can set up all this machinery by hand to be able to
Get PKCS#12 key format without a corresponding Register PKCS#12 key format.
The key management system may specify rules for valid group
names which may be created by the client. Clients are informed of such rules by
a mechanism that is not specified by [KMIP-SPEC]. In
the protocol, the group names themselves are text strings of no specified
format. Specific key management system implementations may choose to support
hierarchical naming schemes or other syntax restrictions on the names. Groups
may be used to associate objects for a variety of purposes. A set of keys used
for a common purpose, but for different time intervals, may be linked by a
common Object Group. Servers may create predefined groups and add objects to
them independently of client requests.
KMIP allows clients to specify whether it wants a “fresh” or
“default” object from a common Object Group. Fresh is an indication of whether
a member of a group has been retrieved by a client with the Get operation. The
value of fresh may be set as an attribute when creating or registering an
object. Subsequently, the Fresh attribute is modifiable only by the server.
For example, a set of symmetric keys belong to the Object Group
“SymmetricKeyGroup1” and the Fresh attribute is set to true for members of the
group at the time of creating or registering the member. To add a new symmetric
key to the group, the Object Group attribute is set to “SymmetricKeyGroup1” and
the Fresh attribute is set to true when creating or registering the symmetric
key object.
The definition of a “default” object in a group is based on
server policy. One example of server policy is to use round robin selection to
serve a key from a group. In this case when a client requests the default key
from a group, the server uses round robin selection to serve the key.
An object may be removed from a group by deleting the Object
Group attribute, as long as server policy permits it. A client would need to
delete each individual member of a group to remove all members of a group.
The Object Group Member flag is specified in the Locate
request to indicate the type of group member to return. Object Group Member is
an enumeration that can take the value Group Member Fresh or Group Member
Default. Following are examples of how the Object Group Member flag is used:
When a Locate request is made by specifying the Object Group
attribute (e.g., "symmetricKeyGroup1) and setting the Object Group Member
flag to "Group Member Fresh", matching objects from the specified
group (e.g., "symmetricKeyGroup1”) have the Fresh attribute set to true.
If there are no fresh objects remaining in the group, the server may generate a
new object on the fly based on server policy.
When a Locate request is made by specifying the Object Group
attribute (e.g., "symmetricKeyGroup2) and setting the Object Group Member
flag to "Group Member Default", a default object is returned from the
group. In this example, the server policy defines default to be the next key in
the group "symmetricKeyGroup2”; the group has three group members whose
Unique Identifiers are uuid1, uuid2, uuid3. If the client performs four
consecutive batched Locate and Get operations with Object Group set to
"symmetricKeyGroup2" and Object Group Member set to “Group Member
Default” in the Locate request, the server returns uuid1, uuid2, uuid3, and
uuid1 (restarting from the beginning with uuid1 for the fourth request) in the
four Get responses.
The server may have multiple ways in which keys can be
stored and protected. The client may wish to avail themselves of particular
variations. The client can specify an ordered list of preferences for said
storage, one of which the server must meet in its entirety or fail the request.
One example of this might be the requirement for cryptographic material to
remain within a geographical location or boundary (data sovereignty)
The enumerations for the Revocation Reason attribute
specified in KMIP [KMIP-SPEC]) are aligned with the
Reason Code specified in [X.509] and referenced in [RFC5280] with the following exceptions. The certificateHold
and removeFromCRL reason codes have been excluded from [KMIP-SPEC] since KMIP does not support certificate
suspension (putting a certificate hold) or unsuspension (removing a certificate
from hold). The aaCompromise reason code has been excluded from [KMIP-SPEC] since it only applies to attribute
certificates, which are out-of-scope for [KMIP-SPEC].
The privilegeWithdrawn reason code is included in [KMIP-SPEC]
since it may be used for either attribute or public key certificates. In the
context of its use within KMIP it is assumed to only apply to public key
certificates.
For clients that require unique identifiers in a special
form, out-of-band registration/configuration may be used to communicate this
requirement to the server.
Usage should be allocated and handled carefully at the
client, since power outages or other types of client failures (crashes) may
render allocated usage lost. For example, in the case of a key being used for the
encryption of tapes, such a loss of the usage allocation information following
a client failure during encryption may result in the necessity for the entire
tape backup session to be re-encrypted using a different key, if the server is
not able to allocate more usage. It is possible to address this through such
approaches as caching usage allocation information on stable storage at the
client, and/or having conservative allocation policies at the server (e.g., by
keeping the maximum possible usage allocation per client request moderate). In
general, usage allocations should be as small as possible; it is preferable to
use multiple smaller allocation requests rather than a single larger request to
minimize the likelihood of unused allocation.
If an asynchronous operation is cancelled by the client, no
information is returned by the server in the result code regarding any
operations that may have been partially completed. Identification and
remediation of partially completed operations is the responsibility of the
server.
It is the responsibility of the server to determine when to
discard the status of asynchronous operations. The determination of how long a
server should retain the status of an asynchronous operation is
implementation-dependent and not defined by KMIP.
Once a client has received the status on an asynchronous
operation other than “pending”, any subsequent request for status of that
operation may return either the same status as in a previous polling request or
an “unavailable” response.
"Mandatory" Asynchronous operations in the request
header establishes that the server will return Asynchronous Correlation Value
for all the batch items. The expectation is that one of the batch items may
have a longer response time than a client is willing to maintain a connection.
This is useful in an IoT environment that has limited resources.
KMIP supports creation and registration of managed objects
and retrieval of managed objects in both plaintext and optionally wrapped with
another managed object. KMIP also includes support for a subset of the
operations necessary for certificate management (certifying certificate
requests and validating certificate hierarchies). KMIP defines a range of
Hash-based and MAC-based key derivation options.
There are certain situations in which having capability for
a KMIP client to request cryptographic operations from a KMIP server is
beneficial in terms of simplifying the client implementation, strengthening the
integration between the key management and cryptographic operations, or improving
the overall security of a solution.
KMIP includes support for cryptographic services in the form
of client-to-server operations for cryptographic services using managed objects
for encryption, decryption, signature generation, signature verification, MAC
generation, MAC verification, random number generation, and general hashing.
This support for cryptographic services is similar to the
approach taken in KMIP for certificates. The protocol supports a base set of
operations on certificates that enable a key manager to act as a proxy for a
Certification Authority (CA) or in fact operate as a CA in the contexts where
that is appropriate. A KMIP server supporting cryptographic services may be
acting as a proxy for another cryptographic device or in fact operating as a
cryptographic device in the contexts where that is appropriate.
KMIP clients and KMIP servers using cryptographic services
operations should be mindful of selecting a level of protection for the
communication channel (the TLS connection) that provides sufficient protection
of the plaintext data included in cryptographic operations and commensurate
with the security strength of the operation. There is no requirement for the KMIP
server to enforce selection of a level of protection.
Similarly, server policy regarding accepting random data from
a client should reflect the level of confidence that that server has in a
particular client or all clients. Issues in the quality or integrity of random
data provided in RNG Seed can affect key creation, nonce and
IV generation, client-server TLS session key creation, and the random data delivered
to clients with the RNG Retrieve Operation. KMIP, as a protocol, does
not itself enforce restrictions on the quality or nature of the random data provided
by a client in the RNG Seed operation.
A KMIP server that supports the RNG Retrieve and RNG Seed
operations may have a single RNG for the server, an RNG which is shared in an
unspecified manner by KMIP clients or a separate RNG for each KMIP client.
There is no requirement for the KMIP server to implement any specific RNG
model.
KMIP allows for multi-part (streaming) operations where the
data required for the operations may be provided in multiple separate requests.
Each of the existing encryption, decryption, signature generation, signature
verification, MAC generation, MAC verification, and general hashing operations
now support additional parameters to allow for each of the stages of multi-part
usage – initialization, update, and finalization.
The server returns to the client a Correlation Value in the
first response. This correlation value is used by the client in each subsequent
request for performing multi-part (streaming) operations. Each cryptographic
operation has additional optional parameters in the request that enable the
client to indicate which stage of multi-part usage (initialization, update, and
finalization) is being requested.
A client can use the Query Capability Information to
determine if a KMIP server supports the multi-part (streaming) operations by
checking if the Streaming Capability is set to true in the response to a Query
operation with the Query Function value of Query Capability Information.
When a server is processing a correlation value either for
asynchronous or multi-part (streaming) operations and matching an incoming
request against server state on behalf of a client, the following may be
relevant to consider as part of determining whether or not to accept the
specified correlation value:
-
Matching TLS mutual authentication client
credentials
-
Matching KMIP authentication header information
-
Matching link-level (network end-point) details
(i.e. source address information)
-
Matching TLS session identifiers
-
Requiring the first and last operations to be
within a specified time period
-
Requiring each operation to be within a
specified time period
-
Re-check the access rights on all referenced
managed objects
The KMIP specification does not require that any of these
items are considered as the determination of which items are relevant is both
server implementation and security context dependent.
When the Archive operation is performed, it is recommended
that a unique identifier and a minimal set of attributes be retained within the
server for operational efficiency. In such a case, the retained attributes may
include Unique Identifier and State.
The key management system may contain multiple embedded CAs
or may have access to multiple external CAs. How the server routes a certificate
request to a CA is vendor-specific and outside the scope of KMIP. If the server
requires and supports the capability for clients to specify the CA to be used
for signing a Certificate Request, then this information may be provided by
including the X.509 Certificate Issuer attribute in the Certify or Re-certify
request.
[KMIP-SPEC] supports multiple
options for submitting a certificate request to the key management server
within a Certify or Re-Certify operation. It is a vendor decision as to whether
the key management server offers certification authority (CA) functionality or
proxies the certificate request onto a separate CA for processing. The type of
certificate request formats supported is also a vendor decision, and this may,
in part, be based upon the request formats supported by any CA to which the
server proxies the certificate requests.
All certificate request formats for requesting X.509
certificates specified in [KMIP-SPEC] (i.e., PKCS#10,
PEM and CRMF) provide a means for allowing the CA to verify that the client
that created the certificate request possesses the private key corresponding to
the public key in the certificate request. This is referred to as
Proof-of-Possession (POP). However, it should be noted that in the case of the
CRMF format, some CAs may not support the CRMF POP option, but instead rely
upon the underlying certificate management protocols (i.e., CMP and CMC) to
provide POP. In the case where the CA does not support POP via the CRMF format
(including CA functionality within the key management server), an alternative
certificate request format (i.e., PKCS#10, PEM) would need to be used if POP
needs to be verified.
The Re-key, Re-key Key Pair, and Re-certify operations allow
the specification of an offset interval.
The Re-key and the Re-key Key Pair operations allow the
client to specify an offset interval for activation of the key. This offset
specifies the duration of time between the time the request is made and the
time when the activation of the key occurs. If an offset is specified, all
other times for the new key are determined from the new Activation Date, based
on the intervals used by the previous key, i.e., from the Activation Date to the
Process Start Date, Protect Stop Date, etc.
The Re-certify operation allows the client to specify an
offset interval that indicates the difference between the Initial Date of the
new certificate and the Activation Date of the new certificate. As with the Re-key
operation, all other times for the certificate are determined using the intervals
used for the previous certificate.
Note that in the Re-key and Re-key Key Pair operations if
activation date, process start date, protect stop date and deactivation date
are obtained from the existing key, and the initial date is obtained from the current
time, then the deactivation/activation date/process start date/protect stop
date is smaller or less than initial date. KMIP allows forward-dating of these
values to prevent a contradiction.
The process of generating a new certificate to replace an
existing certificate may be referred to by multiple terms, based upon what data
within the certificate is changed when the new certificate is created. In all
situations, the new certificate includes a new serial number and new validity
dates.
[KMIP-SPEC] uses the following
terminology which is aligned with the definitions found in IETF [RFC3647] and [RFC4949]:
·
Certificate Renewal: The issuance of a new certificate to
the subject without changing the subject public key or other information (except
the serial number and certificate validity dates) in the certificate.
·
Certificate Update: The issuance of a new certificate,
due to changes in the information in the certificate other than the subject
public key.
·
Certificate Rekey: The generation of a new key pair for
the subject and the issuance of a new certificate that certifies the new public
key.
The KMIP Specification supports certificate renewals using
the Re-Certify operation and certificate updates using the Certify operation.
Certificate rekey is supported through the submission of a Re-key Key Pair
operation, which generates a replacement (new) key pair, followed by a Certify
operation, which issues a new certificate containing the replacement (new)
public key.
Any Certificate Revocation List (CRL) checking which may be
required for certificate-related operations such as Register and Re-key should
be performed by the client prior to requesting the operation from a server.
The Delegated Login operation supports a client delegating
a subset of client rights. The operation allows specification of a time limit
and/or a request usage count, along with a list of operations and a list of
objects that the returned login ticket is permitted to use.
The Discover Versions operation allows clients and servers
to identify a KMIP protocol version that both client and server understand. The
operation addresses both the “dumb” and “smart” client scenarios. Dumb clients
may simply pick the first protocol version that is returned by the server,
assuming that the client provides the server with a list of supported protocol
version. Smart clients may request the server to return a complete list of
supported protocol versions by sending an empty request payload and picking a
protocol version that is supported by both client and server.
Clients specify the protocol version in the request header
and optionally provide a list of protocol versions in the request payload. If
the protocol version in the request header is not specified in the request
payload and the server does not support any protocol version specified in the
request payload, the server returns an empty list in the response payload. In
this scenario, clients are aware that the request did not result in an error
and could communicate with the server using the protocol version specified in
the request header.
During formal KMIP interoperability testing it is helpful
for a client to be able to indicate which test case is being executed to enable
a server to denote this within its logs and use the information for automated
comparison against expected results.
The Offset Items field in the Locate operation allows a KMIP
client to request different parts of a larger Locate result set from the KMIP
server. KMIP servers are also permitted to return the number of items in the
Located Items field which match a Locate.
With the knowledge of the number of Located items and the
combined use of the Offset Items and Maximum Items fields, a KMIP client is
able to “page” or “browse” through a large Locate result set to find the items
of interest.
The KMIP specification does not require that the results be
returned in a specific order. If operations are performed between Locate
requests that alter the result set that is returned by a Locate operation, then
items may be missed, or items may be returned more than once across separate
Locate requests that simply vary the offset items value. The same circumstances
apply when performing separate Locate requests without an offset value.
The Login and Logout operations support caching of authentication
checks to enable more efficient interaction with authentication services. The
Login operation allows specification of a time limit or request usage count for
the returned login ticket.
Query is defined as both a Client-to-Server and Server to
Client operation. Conceptually, the Client-to-Server Query and Server to
Client Query operations are equivalent, but the parameters used in each of
these queries may differ since some concepts are only applicable to a server or
to a client.
Function values allow a KMIP client to obtain a wide range
of information about a KMIP server and allow a KMIP client to
determine which extensions a KMIP server supports (Query Extension List and
Query Extension Map); which random number generators a KMIP server supports
(Query RNGs); which external validations (e.g., FIPS140, Common Criteria) apply
to the KMIP server (Query Validations); which KMIP Profiles from each version
of the KMIP specification a KMIP server supports (Query Profiles); which optional
areas of the KMIP specification and alternate behaviors a KMIP server supports
(Query Capabilities); and which methods of automatic client credential
registration (Query Client Registration Methods) are supported by the KMIP
server.
Test cases for using these various Query
functions may be found in the KMIP Test Case document [KMIP-TC].
Query does not explicitly support client requests to
determine what operations require authentication. To determine whether an
operation requires authentication, a client should request that operation.
During a Create Key Pair or Re-key Key Pair operation, a
Link Attribute is automatically created by the server for each object (i.e., a
link is created from the private key to the public key and vice versa). Certain
attributes are the same for both objects and are set by the server while
creating the key pair. The KMIP protocol does not support an equivalent
operation for registering a key pair. Clients can register the objects
independently and manually set the Link attributes to make the server aware
that these keys are associated with each other. When the Link attribute is set
for both objects, the server should verify that the registered objects indeed
correspond to each other and apply similar restrictions as if the key pair was
created on the server.
Clients should perform the following steps when registering
a key pair:
1. Register the
public key and set all associated attributes:
a.
Cryptographic
Algorithm
b.
Cryptographic
Length
c.
Cryptographic
Usage Mask
1. Register the
private key and set all associated attributes
a.
Cryptographic
Algorithm is the same for both public and private key
b.
Cryptographic
Length is the same for both public and private key
c.
Cryptographic
Parameters may be set; if set, the value is the same for both the public and
private key
d.
Cryptographic
Usage Mask is set, but does not contain the same value for both the public and
private key
e.
Link is
set for the Private Key with Link Type Public Key Link and the Linked
Object Identifier of the corresponding Public Key
f.
Link is
set for the Public Key with Link Type Private Key Link and the Linked
Object Identifier of the corresponding Private Key
3.46
Using Notify and Put Operations
The Notify and Put operations are operations in the KMIP
protocol that are initiated by the server, rather than the client. As
client-initiated requests are able to perform these functions (e.g., by polling
to request notification), these operations are optional for conforming KMIP
implementations. However, they provide a mechanism for optimized communication
between KMIP servers and clients.
In using Notify and Put, the following constraints and
guidelines should be observed:
·
The client enrolls with the server, so that the server knows how
to locate the client to which a Notify or Put is being sent and which events
for the Notify are supported. However, such registration is outside the scope
of the KMIP protocol. Registration also includes a specification of whether a
given client supports Put and Notify, and what attributes may be included in a
Put for a particular client.
·
Communication between the client and the server is authenticated.
Authentication for a particular client/server implementation is at a minimum
accomplished using one of the mandatory authentication mechanisms (see [KMIP-PROF]). Further strengthening of the client/server
communications integrity by means of signed message content and/or wrapped keys
is recommended.
·
In order to minimize possible divergence of key or state
information between client and server as a result of server-initiated
communication, any client receiving Notify or Put messages returns
acknowledgements of these messages to the server. This acknowledgement may be
at communication layers below the KMIP layer, such as by using transport-level
acknowledgement provided in TCP/IP.
For client devices that are incapable of responding to
messages from the server, communication with the server happens via a proxy
entity that communicates with the server, using KMIP, on behalf of the client.
It is possible to secure communication between a proxy entity and the client
using other, potentially proprietary mechanisms.
Flow Control or
"Endpoint Role" is useful for "Put" and "Notify"
operations for clients that are available intermittently or are not directly
accessible to the server. It provides a uniform means of delivering data from
the server to the client that would be available by requests from a polling
client.
The Client Correlation Value can be used by a client to
provide additional, non-critical information to a server. The value need not
be unique. The server should log the value and be able to identify log records
that contain a given Client Correlation Value. An example use of the Client
Correlation value is to track the KMIP test case number used when performing
interoperability testing by the KMIP Technical Committee.
The Server Correlation Value is generated by a server and used to uniquely identify each request. The
value should be globally unique and should be logged by the server for each
request. The server provides the value to the client within the response to
each KMIP operation and the client should store the value in its log. Ideally
both the client and server should locate log records that contain a
given Server Correlation Value so that the workflow associated with each
request may be traced across client and server.
KMIP supports batch operations in order to reduce the number
of calls between the client and server. For example, Locate and Get are likely
to be commonly accomplished within a single batch request.
KMIP does
not ensure that batch operations are atomic on the server side. If servers
implement such atomicity, the client is able to use the optional “undo” mode to
request roll‐back for batch operations implemented as atomic
transactions. However, support for “undo” mode is optional in the protocol, and
there is no guarantee that a server that supports “undo” mode has effectively
implemented atomic batches.
A client can
determine if a server supports the optional “undo” or the optional “continue”
mode using a Query operation with the Query Function value of Query Capability Information
and by checking the value of the Batch Undo Capability or Batch Continue
Capability fields in the response.
3.50 Maximum Message Size
When a server is processing requests in a batch, it should
compare the cumulative response size of the message to be returned after each request
with the specified Maximum Response Size. If the message is too large, it should
prepare a maximum message size error response message at that point, rather
than continuing with operations in the batch. This increases the client’s
ability to understand what operations have and have not been completed.
When processing individual requests within the batch, the
server that has encountered a Maximum Response Size error should not return
attribute values or other information as part of the error response.
The Locate operation also supports the concept of a maximum
item count to include in the returned list of unique identifiers.
Any number of vendor-specific extensions may be included in
the Message Extension optional structure. This allows KMIP implementations to
create multiple extensions to the protocol.
KMIP specifies the Result Status, the Result Reason and the
Result Message as normative message contents. For the Result Status and Result Reason,
the enumerations provided in [KMIP-SPEC] are the
normative values. The values for the Result Message text are
implementation-specific. In consideration of internationalization, it is
recommended that any vendor implementation of KMIP provide appropriate language
support for the Return Message. How a client specifies the language for Result
Messages is outside the scope of the KMIP.
Additional Result Reasons have been added
to enable a client to self-diagnose error situations with additional
information about the nature of the error. Previously many common error
conditions resulted in a single Result Reason. More specificity improves the
adoption of additional capabilities within the specification.
In some scenarios the server may want assurance of the
integrity of the client’s system before honoring a client’s request.
Additionally, the server may want a guarantee of the freshness of the
attestation computation in the integrity measurement.
Generally, the process takes four passes:
- The client sends a request to the server which
requires attestation.
- The server returns a random nonce to the client
that will be used in the attestation computation to guarantee the
freshness of the measurement.
- The client sends a request to the server which
includes the measurement of the client’s system, and the measurement
contains the nonce from the server.
- The
server verifies the measurement and sends the appropriate response to the
client.
Passing attestation data with a
client request can be achieved in KMIP as follows:
- The client sends a request to
the server with the Attestation Capable Indicator set to True in the
request header.
- If the request requires attestation, the server will return an
“Attestation Required” error with a Nonce object in the response header. {If the client request fails for any reason other
than “Attestation Required”, the server will not include a nonce in the
error message.}
- The client uses the nonce received from the server in the
attestation computation that will be used in the measurement.
- The client forms an Attestation Credential Object which contains
either the measurement from the client or an assertion from a third party
if the server is not capable or willing to verify the attestation data
from the client.
- The client then issues a request which contains the Attestation
Credential Object in the request header.
- The server validates the measurement or assertion data in the
Credential Object, checks that the nonce in the Credential Object matches
one sent recently by the server, then sends the appropriate response to
complete the request issued by the client. {If the measurement or
assertion data in the Credential Object does not validate or if the nonce
does not match one sent recently by the server, the server will return an
“Attestation Failed” error instead of completing the request issued by the
client.}
The server needs to be capable
of processing and verifying multiple Credential Objects in the same request
header since Attestation Credentials do not provide the same type of
authentication as the Username and Password or Device Credential.
How frequently (e.g. every
request, every 100 requests, etc.) the server generates a new random nonce
depends on server policy. The lifetime of the nonce once the server has sent
it to the client (i.e., the timeframe in which the client must return the nonce
before needs to request a fresh nonce from the server) also depends on server
policy.
If the client sends a request
that requires attestation, but the client has not set the Attestation
Capable Indicator to True, then the server will send a “Permission Denied”
error and will not include a Nonce object in the response header.
KMIP allows for vendor extensions in a number of areas:
1. Enumerations have
specific ranges which are noted as extensions
2. Item Tag values of the
form 0x54xxxx are reserved for vendor extensions
3. Attributes may be
defined by the client with a “x-“ prefix or by the server with a “y-“ prefix
Extensions may be used by vendors to communicate information
between a KMIP client and a KMIP server that is not currently defined within
the KMIP specification.
A common use of extensions is to allow for the structured
definition of attributes using KMIP TTLV encoding rather than encoding vendor
specific information in opaque byte strings.
This section describes how to apply the functionality
described in the Key Management Interoperability Protocol Specification to
address specific key management usage scenarios or to solve key management
related issues.
It is possible to formulate Locate queries to address any of
the following conditions:
·
Exact match of a transition to a given state. Locate the key(s)
with a transition to a certain state at a specified time (t).
·
Range match of a transition to a given state. Locate the key(s)
with a transition to a certain state at any time at or between two specified
times (t and t’).
·
Exact match of a state at a specified time. Locate the key(s)
that are in a certain state at a specified time (t).
·
Match of a state during an entire time range. Locate the key(s)
that are in a certain state during an entire time specified with times (t and
t’). Note that the Activation Date could occur at or before t and that the
Deactivation Date could occur at or after t’+1.
·
Match of a state at some point during a time range. Locate the
key(s) that are in a certain state at some time at or between two specified
times (t and t’). In this case, the transition to that state could be before
the start of the specified time range.
This is accomplished by allowing any date/time attribute to
be present either once (for an exact match) or at most twice (for a range
match).
For instance, if the state we are interested in is Active,
the Locate queries would be the following (corresponding to the bulleted list
above):
·
Exact match of a transition to a given state: Locate
(ActivationDate(t)). Locate keys with an Activation Date of t.
·
Range match of a transition to a given state: Locate
(ActivationDate(t), ActivationDate(t')). Locate keys with an Activation Date at
or between t and t’.
·
Exact match of a state at a specified time: Locate
(ActivationDate (0), ActivationDate (t), DeactivationDate (t+1),
DeactivationDate (MAX_INT), CompromiseDate (t+1), CompromiseDate (MAX_INT)).
Locate keys in the Active state at time t, by looking for keys with a
transition to Active before or until t, and a transition to Deactivated or
Compromised after t (because we don't want the keys that have a transition to
Deactivated or Compromised before t). The server assumes that keys without a
DeactivationDate or CompromiseDate is equivalent to MAX_INT (i.e., infinite).
·
Match of a state during an entire time range: Locate
(ActivationDate (0), ActivationDate (t), DeactivationDate (t'+1),
DeactivationDate (MAX_INT), CompromiseDate (t'+1), CompromiseDate (MAX_INT)).
Locate keys in the Active state during the entire time from t to t’.
·
Match of a state at some point during a time range: Locate
(ActivationDate (0), ActivationDate (t'-1), DeactivationDate (t+1), DeactivationDate
(MAX_INT), CompromiseDate (t+1), CompromiseDate (MAX_INT)). Locate keys in the
Active state at some time from t to t’, by looking for keys with a transition
to Active between 0 and t’-1 and exit out of Active on or after t+1.
The queries would be similar for Initial Date, Deactivation
Date, Compromise Date and Destroy Date.
In the case of the Destroyed-Compromise state, there are two
dates recorded: the Destroy Date and the Compromise Date. For this state, the
Locate operation would be expressed as follows:
·
Exact match of a transition to a given state: Locate
(CompromiseDate (t), State (Destroyed-Compromised)) and Locate (DestroyDate
(t), State (Destroyed-Compromised)). KMIP does not support the OR in the Locate
request, so two requests should be issued. Locate keys that were Destroyed and
transitioned to the Destroyed-Compromised state at time t, and locate keys that
were Compromised and transitioned to the Destroyed-Compromised state at time t.
·
Range match of a transition to a given state: Locate
(CompromiseDate (t), CompromiseDate (t'), State (Destroyed-Compromised)) and
Locate (DestroyDate (t), DestroyDate (t'), State (Destroyed-Compromised)).
Locate keys that are Destroyed-Compromised and were Compromised or Destroyed at
or between t and t’.
·
Exact match of a state at a specified time: Locate
(CompromiseDate (0), CompromiseDate (t), DestroyDate (0), DestroyDate (t));
nothing else is needed, since there is no exit transition. Locate keys with a
Compromise Date at or before t, and with a Destroy Date at or before t. These
keys are, therefore, in the Destroyed-Compromised state at time t.
·
Match of a state during an entire time range: Locate
(CompromiseDate (0), CompromiseDate (t), DestroyDate (0), DestroyDate (t)).
Same as above. As there is no exit transition from the Destroyed-Compromised
state, the end of the range (t’) is irrelevant.
·
Match of a state at some point during a time range: Locate
(CompromiseDate (0), CompromiseDate (t'-1), DestroyDate (0), DestroyDate
(t'-1)). Locate keys with a Compromise Date at or before t’-1, and with a
Destroy Date at or before t’-1. As there is no exit transition from the
Destroyed-Compromised state, the start of the range (t) is irrelevant.
KMIP provides the option to register and get keys in wrapped
format. Clients request the server to return a wrapped key by including the Key
Wrapping Specification in the Get Request Payload. Similarly, clients register
a wrapped key by including the Key Wrapping Data in the Register Request
Payload.
The Wrapping Method within the Key Wrapping Data identifies
the type of mechanism used to wrap the key, but does not identify the algorithm
or block cipher mode. It is possible to determine these from the attributes set
for the specified Encryption Key or MAC/Signing Key. If a key has multiple
Cryptographic Parameters set, clients may include the applicable parameters in
Key Wrapping Specification. If omitted, the server chooses the Cryptographic
Parameter attribute with the lowest index.
The Key Value includes both the Key Material and,
optionally, attributes of the key; these may be provided by the client in the
Register Request Payload; the server only includes attributes when requested in
the Key Wrapping Specification of the Get Request Payload. The Key Value may be
encrypted, signed/MACed, or both encrypted and signed/MACed (and vice versa).
In addition, clients have the option to request or import a wrapped Key Block
according to standards, such as ANSI TR-31, or vendor-specific key wrapping
methods.
It is important to note that if the Key Wrapping
Specification is included in the Get Request Payload, the Key Value may not
necessarily be encrypted. If the Wrapping Method is MAC/sign, the returned Key
Value is in plaintext, and the Key Wrapping Data includes the MAC or Signature
of the Key Value.
The Key Wrap Type parameter in the Get operation may be used
to determine how a key that was registered as a wrapped key is returned to a
client. Specifying a value of Not Wrapped ensures that the server returns the
unwrapped key value. A value of As Registered can be used to retrieve the key
value as it was provided in the Register operation. In the latter case, the
wrapping key need not be known to the server. If no Key Wrap Type is provided,
then the server may choose to return the key either wrapped or unwrapped. A
Get operation may use both a Key Wrap Type and a Wrapping Key Specification, in
which case the Key Wrap Type is processed as if there was no Wrapping Key
Specification, and the result is then wrapped as specified.
Prior to wrapping or unwrapping a key, the server should verify
that the wrapping key is allowed to be used for the specified purpose. For
example, if the Unique ID of a symmetric key is specified in the Key Wrapping
Specification inside the Get request, the symmetric key should have the “Wrap Key”
bit set in its Cryptographic Usage Mask. Similarly, if the client registers a
signed key, the server should verify that the Signature Key, as specified by
the client inside the Key Wrapping Data, has the “Verify” bit set in the Cryptographic
Usage Mask. If the wrapping key is not permitted to be used for the requested
purpose (e.g., when the Cryptographic Usage Mask is not set), the server should
return the Operation Failed result status.
The client sends a Get request to obtain a key that is
stored on the server. When the client sends a Get request to the server, a Key
Wrapping Specification may be included. If a Key Wrapping Specification is
included in the Get request, and a client wants the requested key and its
Cryptographic Usage Mask attribute to be wrapped with AES key wrap, the client
includes the following information in the Key Wrapping Specification:
·
Wrapping Method: Encrypt
·
Encryption Key Information
· Unique Key
ID: Key ID of the AES wrapping key
· Cryptographic
Parameters: The Block Cipher Mode is NISTKeyWrap (not necessary if default
block cipher mode for wrapping key is NISTKeyWrap)
·
Attribute Name: Cryptographic Usage Mask
The server uses the Unique Key ID specified by the client to
determine the attributes set for the proposed wrapping key. For example, the
algorithm of the wrapping key is not explicitly specified inside the Key
Wrapping Specification. The server determines the algorithm to be used for
wrapping the key by identifying the Algorithm attribute set for the specified
Encryption Key.
The Cryptographic Parameters attribute should be specified
by the client if multiple instances of the Cryptographic Parameters exist, and
the lowest index does not correspond to the NIST key wrap mode of operation.
The server should verify that the AES wrapping key has NISTKeyWrap set as an
allowable Block Cipher Mode, and that the “Wrap Key” bit is set in the
Cryptographic Usage Mask.
If the correct data was provided to the server, and no
conflicts exist, the server AES key wraps the Key Value (both the Key Material
and the Cryptographic Usage Mask attribute) for the requested key with the
wrapping key specified in the Encryption Key Information. The wrapped key (byte
string) is returned in the server’s response inside the Key Value of the Key
Block.
The Key Wrapping Data of the Key Block in the Get Response
Payload includes the same data as specified in the Key Wrapping Specification of
the Get Request Payload except for the Attribute Name.
The client sends a Register request to the server and
includes the wrapped key and the Unique ID of the wrapping key inside the
Request Payload. The wrapped key is provided to the server inside the Key
Block. The Key Block includes the Key Value Type, the Key Value, and the Key
Wrapping Data. The Key Value Type identifies the format of the Key Material,
the Key Value consists of the Key Material and optional attributes that may be
included to cryptographically bind the attributes to the Key Material, and the
Key Wrapping Data identifies the wrapping mechanism and the encryption key used
to wrap the object and the wrapping mechanism.
Similar to the example in 4.2.1 the key is wrapped
using the AES key wrap. The Key Value includes four attributes: Cryptographic
Algorithm, Cryptographic Length, Cryptographic Parameters, and Cryptographic
Usage Mask.
The Key Wrapping Data includes the following information:
·
Wrapping Method: Encrypt
·
Encryption Key Information
· Unique Key
ID: Key ID of the AES wrapping key
· Cryptographic
Parameters: The Block Cipher Mode is NISTKeyWrap (not necessary if default
block cipher mode for wrapping key is NISTKeyWrap)
Attributes do not need to be specified in the Key Wrapping
Data. When registering a wrapped Key Value with attributes, clients may include
these attributes inside the Key Value without specifying them inside the
Template-Attribute.
Prior to unwrapping the key, the server determines the
wrapping algorithm from the Algorithm attribute set for the specified Unique ID
in the Encryption Key Information. The server verifies that the wrapping key
may be used for the specified purpose. In particular, if the client includes
the Cryptographic Parameters in the Encryption Key Information, the server
verifies that the specified Block Cipher Mode is set for the wrapping key. The
server also verifies that the wrapping key has the “Unwrap Key” bit set in the
Cryptographic Usage Mask.
The Register Response Payload includes the Unique ID of the
newly registered key and an optional list of attributes that were implicitly
set by the server.
The client sends a Get request to obtain a key (either
symmetric or asymmetric) that is stored on the server. When the client sends a
Get request to the server, a Key Wrapping Specification may be included. If a
Key Wrapping Specification is included, and the key is to be wrapped with an
RSA public key using the OAEP encryption scheme, the client includes the
following information in the Key Wrapping Specification. Note that for this
example, attributes for the requested key are not requested.
·
Wrapping Method: Encrypt
·
Encryption Key Information
· Unique Key
ID: Key ID of the RSA public key
· Cryptographic
Parameters:
Padding
Method: OAEP
Hashing
Algorithm: SHA-256
The Cryptographic Parameters attribute is specified by the
client if multiple instances of Cryptographic Parameters exist for the wrapping
key, and the lowest index does not correspond to the associated padding method.
The server should verify that the specified Cryptographic Parameters in the Key
Wrapping Specification and the “Wrap Key” bit in the Cryptographic Usage Mask
are set for the corresponding wrapping key.
The Key Wrapping Data returned by the server in the Key
Block of the Get Response Payload includes the same data as specified in the
Key Wrapping Specification of the Get Request Payload.
The Mask Generator and Mask Generator Hashing Algorithm
enumerations allow the Mask Generation Function (MGF) and associated algorithm
to be explicitly specified. If Mask Generator Hashing Algorithm is omitted,
then the value specified for Hashing Algorithm is assumed.
The example above assumes the same algorithm (SHA-256) is
used for both the Mask Generation Function (MGF) and hashing data.
The client sends a Get request to obtain a key that is
stored on the server. When the client sends a Get request to the server, a Key
Wrapping Specification may be included. If a key and Custom Attribute (i.e.,
x-Nonce) is to be MACed with HMAC SHA-256, the following Key Wrapping
Specification is specified:
·
Wrapping Method: MAC/sign
·
MAC/Signature Key Information
· Unique Key
ID: Key ID of the MACing key (note that the algorithm associated with this key
would be HMAC-SHA256)
·
Attribute Name: x-Nonce
For HMAC, no Cryptographic Parameters need to be specified,
since the algorithm, including the hash function, may be determined from the
Algorithm attribute set for the specified MAC Key. The server should verify
that the HMAC key has the “MAC Generate” bit set in the Cryptographic Usage
Mask. Note that an HMAC key does not require the “Wrap Key” bit to be set in
the Cryptographic Usage Mask.
The server creates an HMAC value over the Key Value if the specified
MACing key may be used for the specified purpose and no conflicts exist. The
Key Value is returned in plaintext, and the Key Block includes the following
Key Wrapping Data:
·
Wrapping Method: MAC/sign
·
MAC/Signature Key Information
·
Unique Key ID: Key ID of the MACing key
·
MAC/Signature: HMAC result of the Key Value
In the example, the custom attribute x-Nonce was included to
help clients, who are relying on the proxy model, to detect replay attacks.
End-clients, who communicate with the key management server, may not support
TLS and may not be able to rely on the message protection mechanisms provided
by a security protocol. An alternative approach for these clients would be to
use the custom attribute to hold a random number, counter, nonce, date, or
time. The custom attribute needs to be created before requesting the server to
return a wrapped key and is recommended to be set if clients frequently
wrap/sign the same key with the same wrapping/signing key.
Clients may want to register and store a wrapped key on the
server without the server being able to unwrap the key (i.e., the wrapping key
is not known to the server). Instead of storing the wrapped key as an opaque
object, clients have the option to store the wrapped key inside the Key Block
as an opaque cryptographic object, i.e., the wrapped key is registered as a
managed cryptographic object, but the encoding of the key is unknown to the
server. Registering an opaque cryptographic object allows clients to set all
the applicable attributes that apply to cryptographic objects (e.g.,
Cryptographic Algorithm and Cryptographic Length),
Opaque cryptographic objects are set by specifying the
following inside the Key Block structure:
·
Key Format Type: Opaque
·
Key Material: Wrapped key as a Byte String
The Key Wrapping
Data does not need to be specified.
KMIP provides the option to specify the Encoding Option
inside the Key Wrapping Specification and Key Wrapping Data. This option allows
users to Get or Register the Key Value in a non-TTLV encoded format. This may
be desirable in a proxy environment, where the end-client is not KMIP-aware.
The Encoding Option is only available if no attributes are
specified inside the Key Value. The server returns the Encoding Option Error if
both the Encoding Option and Attribute Names are specified inside the Key
Wrapping Specification. Similarly, the server is expected to return the
Encoding Option Error when registering a wrapped object with attributes inside
the Key Value and the Encoding Option is set in the Key Wrapping Data. If no
Encoding Option is specified, KMIP assumes that the Key Value is TTLV-encoded.
Thus, by default, the complete TTLV-encoded Key Value content, as shown in the
example below, is wrapped:
Key
Material || Byte String || Length || Key Material Value
420043
|| 08 || 00000010 || 0123456789ABCDEF0123456789ABCDEF
Some end-clients may not understand or have the space for
anything more than the actual key material (i.e.,
0123456789ABCDEF0123456789ABCDEF in the above example). To wrap only the Key
Material value during a Get operation, the Encoding Option (00001 for no
encoding) should be specified inside the Key Wrapping Specification. The same
Encoding Option should be specified in the Key Wrapping Data when returning the
non-TTLV encoded wrapped object inside the Get Response Payload or when
registering a wrapped object in non-TTLV encoded format.
It is important to be aware of the risks involved when
excluding the attributes from the Key Value. Binding the attributes to the key
material in certain environments is essential to the security of the
end-client. An untrusted proxy could change the attributes (provided separately
via the Get Attributes operation) that determine how the key is being used
(e.g., Cryptographic Usage). Including the attributes inside the Key Value and
cryptographically binding it to the Key Material could prevent potential misuse
of the cryptographic object and may prevent a replay attack if, for example, a
nonce is included as a custom attribute. The exclusion of attributes and
therefore the usage of the Encoding Option are only recommended in at least one
of the following scenarios:
1. End-clients are
registered with the KMIP server and are communicating with the server directly
(i.e., the TLS connection is between the server and client).
2. The environment is
controlled and non-KMIP-aware end-clients are aware how wrapped cryptographic
objects (possibly Raw keys) from the KMIP server should be used without having
to rely on the attributes provided by the Get Attributes operation.
3. The wrapped
cryptographic object consists of attributes inside the Key Material value.
These attributes are not interpreted by the KMIP server but are understood by
the end-client. This may be the case if the Key Format Type is opaque or
vendor-specific.
4. The proxy
communicating with the KMIP server on behalf of the end-client is considered to
be trusted and is operating in a secure environment.
Registering a wrapped object without attributes is not
recommended in a proxy environment, unless scenario 4 is met.
This section describes methods and provides examples for
creating and storing key identifiers that are interoperable across multi-vendor
KMIP clients, using the KMIP Tape Library Profiles.
A common method for naming and retrieving keys is needed to
support moving tape cartridges between 2 or more KMIP-compliant tape libraries
that are all registered with the same KMIP key manager.
The method uses the KMIP Tape Library Profile. This profile
specifies use of the KMIP Application Specific Information (ASI) attribute. The
method supports both client-generated and server-generated key identifiers.
The key identifier is a KMIP string, composed of hexadecimal
numeric characters. This string of characters is unique within a chosen
namespace. Methods of generating the string are determined by policy. The
LIBRARY-LTO namespace is preferred for maximum interoperability.
A compressed (numeric) transformation of the identifier
string is stored in the tape format’s Key Associated Data. This allows for
future retrieval of the key for decryption.
Interoperability is achieved by a) standardized algorithms
to map byte values between the numeric (KAD) and text (ASI) representations of
the identifier; and b) standardized ordering of bytes within the KAD so the
identifier can be re-assembled in the correct sequence by other compliant
implementations. Examples of the algorithms are provided below.
Key Associated Data (KAD): Part of the tape format.
May be segmented into authenticated and unauthenticated fields. KAD usage is
detailed in the SCSI SSC-3 standard from the T10 organization.
Application Specific Information (ASI): A KMIP
attribute.
Hexadecimal numeric characters: Case-sensitive,
printable, single byte ASCII characters representing the numbers 0 through 9
and uppercase alpha A through F. (US-ASCII characters 30h-39h and 41h-46h).
Hexadecimal numeric characters are always paired, each pair representing a
single 8-bit numeric value. A leading zero character is provided, if necessary,
so that every byte in the tape’s KAD is represented by exactly two hexadecimal
numeric characters.
N(k): The number of bytes in the tape format combined
KAD fields (both authenticated and unauthenticated).
N(a), N(u): The number of bytes in the tape formats authenticated,
and unauthenticated KAD fields, respectively.
Refer to the KMIP Tape profile for algorithm 1.
This algorithm is associated with writing the KAD, typically
to allow future retrieval of a key. An example implementation is as follows.
1.
The client creates a key identifier or obtains one from the server.
The identifier is a KMIP string of hexadecimal numeric characters. Copy the
string to an input buffer of size 2*N(k) bytes. For LTO4, an 88 character
string is sufficient to represent any key name stored directly in the KAD
fields. For LTO5, a 184 character string is sufficient to represent any key
name stored directly in the KAD fields.
2.
Define output buffers for unauthenticated KAD, and authenticated KAD, of
size N(u) and N(a) respectively. For LTO4, this would be 32 bytes of
unauthenticated data, and 12 bytes of authenticated data. For LTO5, this would
be 32 bytes of unauthenticated data and 60 bytes of authenticated data.
3.
Define the standard POSIX (also known as C) locale. Each character in
the string is a single-byte US-ASCII character.
4.
First, populate the authenticated KAD buffer, converting a sub-string
consisting of the last (rightmost) 2*N(a) characters of the key identifier
string.
5.
When the authenticated KAD is filled, next populate the unauthenticated
KAD buffer, by converting the remaining hexadecimal character pairs (if any) of
the identifier string.
This algorithm is associated with reading the KAD, typically
in preparation for retrieving a key. An example implementation is as follows
1.
Define
an input buffer sized for N(k). For LTO4, N(k) is 44 bytes (12 bytes
authenticated, 32 unauthenticated). For LTO5, N(k) is 92 bytes (60 bytes
authenticated, 32 bytes unauthenticated).
2.
Define
an output buffer sufficient to contain a string with a maximum length of
2*N(k) bytes.
3.
Define
the standard POSIX (also known as C) locale. Each character in the string is a
single-byte US-ASCII character.
4.
First,
copy
the tape format’s unauthenticated
KAD
data (if any) to the input
buffer. Next,
bytes
from the authenticated KAD are concatenated, after the unauthenticated bytes. In many implementations the
unauthenticated KAD is empty, and in those cases the entire input buffer will
be populated with bytes from authenticated KAD.
5.
For
each byte in the input buffer, convert to US-ASCII as follows:
6.
Convert
the byte's value to exactly 2 hexadecimal numeric characters, including a
leading 0 where necessary. Append these 2 numeric characters to the output
buffer, with the high-nibble represented by the left-most hexadecimal numeric
character.
The following usage example will create a key identifier
which can be stored in ASI. The identifier will then be translated for storage
into a tape format’s KAD, using algorithm 1. Both LTO4 and LTO5 examples of
KAD contents are provided.
The reverse translation from KAD bytes to the KMIP key
identifier is not shown but would be accomplished via algorithm 2. This
re-constructed key identifier string would be used to Locate the key via ASI.
Example of creating a key identifier.
Implementation-specific material is used to generate a key identifier. The
content of this material is based on server or client policy. An example of a
text string which could be used to generate a KMIP key identifier for tape is
as follows.
SN123456_MFR:XYZ
INC_BAR12345_TM20131234
This example is a set of 40 characters which will be used to
create a KMIP key identifier for use as specified in the KMIP Tape Profile.
Every 8th character is bold.
This set of characters is suitable as a key identifier for
either LTO4 or LTO5, since it will fit within the smaller 44 character KAD
space of LTO4.
The corresponding KMIP key identifier, which is a string of
hexadecimal numeric character pairs, is shown below. This string will be
stored in ASI Application Data.
53 4E 31 32 33 34 35 36 5F 4D 46 52 3A 58 59 5A
20 49 4E 43 5F 42 41 52
31 32 33 34 35 5F 54 4D 32 30 31 33 31 32 33 34
Spaces are shown for to improve readability but are NOT part
of the ASI string. Every 8th hexadecimal numeric pair is bold.
Note the identifier has exactly 2x more characters than the
material used to generate the KMIP key identifier.
Translating the key identifier to KAD bytes (LTO4).
The corresponding KAD content, for use with an LTO4 tape cartridge is shown in
the following figure.
Figure 2: KAD Content for LTO4
Each square is 1 byte (8 bits). Each square contains the
8-bit value which represents a pair of hexadecimal numeric characters in the
KMIP key identifier string.
Every 8th byte of KAD is shaded.
The KAD was populated by converting the rightmost 24
characters (12-character pairs) of the identifier string into bytes of
authenticated KAD. The remaining characters of the identifier were written to
unauthenticated KAD.
Translating the key identifier to KAD bytes (LTO5). The
corresponding KAD for use with an LTO5 and later tape cartridge is shown in the
following figure.
Figure 3: KAD Content for LTO5
Each square is 1 byte (8 bits). Each square contains the
8-bit value which represents a pair of hexadecimal numeric characters in the
key identifier string.
Every 8th byte of KAD is shaded.
The KAD was populated by converting the rightmost 80 characters
(40-character pairs) of the identifier string into bytes of authenticated
KAD. The unauthenticated KAD is not used because all of the data fits within
authenticated KAD.
As tag values and their interpretation for the most part
should be known for a client and server to meaningfully use an extension, the
following registration procedure should be used.
1. Document the
Extensions including:
a.
Extension Tag, Extension Name, Extension Type values to be reserved
b.
A brief description of the purpose of the Extension
c.
Example use case messages (requests and responses)
d.
Example Guidance
2. Send the Document to
the KMIP TC requesting review
3. Request a KMIP TC
ballot on accepting the reservation of the Extension
It is anticipated that a template document may be produced
for this registration process.
PGP, both as vendor product and as standard, provides a rich
environment for key management that addresses significant use cases related to
such areas as secure exchange of email, documents and other resources. Although
KMIP is by no means required for support of PGP environments, it can provide a
valuable mechanism for movement of PGP keys between a particular PGP
environment, such as Symantec Encryption Management Server (SEMS, née PGP
Universal), and another key management environment.
KMIP does not attempt to represent the full range of
functionality in PGP environments. However, the use cases related to movement
of PGP keys across environments, described in the KMIP Use Cases document, can
be supported by taking advantage both of the PGP-specific capabilities in KMIP,
such as the PGP Key object introduced in KMIP V1.2, and of KMIP messages,
objects, operations and attributes in general.
The PGP Key managed object
contains a PGP key (specified in [RFC4880]) as an opaque
blob. KMIP compliant servers do not need to understand the fine structure of
PGP keys. The intention here is that PGP-enabled clients be able to discover
the PGP Key managed cryptographic objects by searching for one of the various
names contained within the block. The Alternative Name attribute can be used
to specify one or more names (e.g. User IDs) that are attached to the PGP Key
object. The PGP-enabled clients are expected to digest the PGP Key object and
properly assign these Alternative Name attributes on to the cryptographic
managed object. The KMIP server does not have to do this work.
Internally, PGP keys may contain many public-private key
pairs, each tied to a specific type of encryption operations (one key for
signing, one for encryption, and one to tie the other two together in a trust
relationship is one typical arrangement.) The Link attribute supports new
values that enable the description of this set of PGP Key relationships. The
new values are parent, child, previous and next. For example, the private and
public keys associated with a PGP Key can be pointed to from the PGP Key with
the “child” link attribute. Additional Decryption Keys (ADK) can be pointed to
from the PGP Key with the “child” link attribute and can be point to each other
with the “previous” and “next” link attributes. In this way, the link
attributes can be used to define the structural relationships required to
establish the web of trust for a PGP Key.
As mentioned above, KMIP does not attempt to represent all
the information about PGP keys that would be managed within a PGP
implementation. For example, policies such as algorithms supported, by a PGP
key are not expressed within KMIP. Instead, KMIP enables the specification of
these attributes, if necessary, as information enclosed within the opaque value
defined for a given PGP key. This information would be handled by security
administration and out-of-band coordination between the PGP environments that
participate in the KMIP exchanges related to PGP keys.
KMIP complaint servers are not expected to be able to create
PGP Key objects from scratch. PGP-enabled clients will do the key creation and
pass the resulting information up to KMIP.
There are several common approaches to
registering KMIP clients with KMIP servers:
- Manual client registration within a single trust boundary
- Automatic client registration across multiple trust boundaries
- Configuring a KMIP Server for use with Automatic Client Registration
The goal of these approaches is to establish the
KMIP-interoperable secure channel or channels between KMIP servers and clients,
such as a mutually-authenticated TLS channel.
In order to support the goal of establishing an
interoperable approach to establishing this channel, this section provides more
detailed information about these approaches to client registration.
The Query Function, Client Registration Methods,
provides a client a means to determine which automated client registrations
method(s) a KMIP server supports. Although a KMIP server is not required to
support any automated registration method. Reflecting common usage for KMIP,
all three of the scenarios described below discuss the use of X.509
certificates for trust establishment; other mechanisms may be used instead but
are not described here. Similarly, all three scenarios describe the
establishment of a mutually-authenticated TLS connection as the basis trusted
exchange of KMIP messages, corresponding to the published KMIP authentication
suite profiles; other authentication mechanisms can be used with KMIP, but are
not described here.
In this approach, there is no assumption of pre-population
of authentication credentials in the client, such as by installing an X.509
certificate into a tape library or drive during the manufacturing process.
Rather, a credential is propagated out-of-band to the client administrator, who
installs it into the client environment. The credential is then used on initial
and subsequent contact between the client and server systems.
This registration model often entails the server
administrator creating a package that contains 1) X.509 certificate that the
client will use to identify itself to the server when creating a TLS
mutually-authenticated session; 2) information about the X.509 certificate that
will be presented by the server to the client during negotiation of the mutual
authentication, enabling the client to verify the server identity; and 3)
possibly additional information that can be included in the credential of the
KMIP message sent across the established channel, such as to provide finer
granularity for particular drives within a tape library. As indicated, the use
of this package of materials takes place during two phases: first during the
establishment of the TLS secure channel; second during the transmission of KMIP
messages. The server administrator must have configured the server to recognize
the X.509 certificate presented by the client, to present the correct X.509
certificate of its own to the client in return and to recognize the additional
information provided in the credential object in the KMIP message, if any.
In this model, KMIP is not used to transmit the X.509
certificate and server information used in establishing the secure channel.
There is nothing to prevent KMIP being used to send this information; but
commonly this is done using mechanisms other than KMIP, nor is there any
expectation that KMIP is a required or default mechanism for propagating the
credential and the information. The distribution mechanism, therefore, may well
vary across vendors.
The use of additional information as the credential in the
KMIP message is also neither required nor a default. Inclusion of such a
credential in the package distributed to the client administrator and in one or
more KMIP messages is also, therefore, likely to vary across vendors.
In this approach the credential used to establish a
mutually-authenticated TLS connection is not provided in the package provided
by the server administrator. Instead, the establishment of trust between the client
and server is accomplished by some other mechanism.
A client may use the Query Function, Client Registration
Methods, to identify which automated client registrations method(s) a KMIP
server supports. Methods included in the enumeration are Unspecified, Server
Pre-generated, Server On-demand, Client Generated and Client Registered. Test
cases which cover how to use the Client Registration Method Query
Function and implement the referenced automated client registration methods are
included in the KMIP Test Cases document [KMIP-TC].
One example of a Server Pre-generated method is having an
X.509 certificate installed in a client device during the manufacturing
process. This certificate is then used as a bootstrap mechanism for the
subsequent exchange of the kind of information exchanged between client
administrator and server administrator.
There will often be configuration activity for the client
device based on information, such as a Service ID, received from the server
administrator. Once the client administrator initiates auto-registration, the client
device sends the X.509 certificate to the server, for example in order to use
it to establish an initial TLS session. The server then sends the equivalent of
the registration packet in section 4.6.1 above to the client and the client
returns the certificate to be used for establishing the secure TLS channel with
the server.
In this model, administrator intervention may be required to
determine whether the initial client certificate should be accepted. The
scenario above assumes that the return of the server’s packet of registration
is immediate and automatic; alternatively, the return of the packet of
information may be done manually by the server administrator, as in section 4.6.1 above; or the return of the packet of server information may be done by the
server, but only after that action has been approved by an administrator.
As discussed in section 4.6.1,
KMIP can be used by the client in sending the X.509 certificates to the server.
However, this is not required. If it is sent to the server using a KMIP
register operation, the server must be able to distinguish that this operation
is intended not only to register the cryptographic object, but also to initiate
the registration of the client as a legitimate participant in KMIP message
exchange.
Another model is to register sub-clients of a trusted
client. In this model, the establishment of trust between the client and server
can be accomplished using either of the approaches in section 4.6.1 or 4.6.2. However, the server may also send additional information to the
client, such as a “tenant identifier”, which it will have to provide to
sub-clients for them to use they attempt to register individually. The
individual sub-clients would follow a registration model such as that described
in section 4.6.2, but would also provide the tenant identifier along with the
X.509 certificate so that the server can decide whether to accept the client,
based on such criteria as (for example) the TCP/IP address of the sub-client
relative to that of the primary client.
This approach can be used for tiered clients that need to be
grouped based on their association with a larger trusted entity, but that also
need individual identities and trust relationships established based on those
identities.
KMIP can be used for sending both the client certificate and
the tenant identifier to the server.
A One Time Pad within the Cryptographic
Algorithm enumeration allows for One Time Pad algorithms to be used within the
context of KMIP. One Time Pads can be used for cryptographic operations where
the key material is either not known to the KMIP server or the KMIP server is
not willing to provide that information to the KMIP client. Test cases and
examples which demonstrate the usage of the One Time Pad algorithm in
combination with the cryptographic operations may be found the KMIP Test
Cases document [KMIP-TC].
Cryptographic (Crypto) Shredding, which is also referred to
as Cryptographic Erasure) is the deletion of a cryptographic key in order to
render data encrypted with that key unrecoverable. Cryptographic shredding may
be supported with KMIP via the Client-to-Server Destroy operation and
via the Server-to-Client Notify operation.
If key material used to encrypt data is only persisted
(stored) at the KMIP server and not at the encryption point (KMIP client is
co-located with or a proxy to the encryption point) then cryptographic
shredding may be achieved by sending a KMIP Destroy request to the KMIP
server. How the KMIP server destroys the key and whether the server choices to
retain meta-data about the deleted key is a vendor implementation decision and
outside of the scope of KMIP. Please see section 4.9 regarding options for
determining the type of key deletion a KMIP server or a KMIP client supports.
If the key material is persisted at the encryption point (in
addition to the at the KMIP server) then the KMIP server may use the (optional)
Notify operation with the Destroy Date attribute included to
inform the KMIP client (co-customer attributed with or a proxy to the
encryption point) that the key has been deleted. Upon receipt of this
notification the KMIP Client (and encryption point) could be programmed to
destroy any copy of the key that it has stored. Note that KMIP does not
require that either a KMIP Server or KMIP client to support the Notify
operation nor does KMIP require that a KMIP client behave in a specific manner
based on a notification. This example just demonstrates how Notify
could be used to facilitate key destruction and cryptographic shredding. The
method the KMIP client uses to destroy the key and whether to maintain
meta-data for the key is a vendor implementation and is also out of scope of
KMIP.
Key
destruction is eliminating a cryptographic key so that it is no longer
accessible for use in a cryptographic function (e.g. NIST [SP800-57-1]
Destroy state). Key shredding is taking the destruction process a step further
so that the cryptographic key is destroyed using a methods where the key is no
only inaccessible from the cryptographic application in which the key was used,
but that the key is not recoverable using forensics methods outside of the
cryptographic application. This type of destruction method normally follows a data
sanitization method such as overwriting key with random data (e.g. [DoD5220.22M] NIST [SP800-88]).
KMIP does not impose any requirements on the methods a KMIP client or KMIP
server uses to destroy a key. However, it is possible to query the type of
destruction methods that a KMIP client or KMIP server support by using the Query
Capabilities function within the Query operation. The Capability
Information object which is returned in the Query response may contain the Shredding
Algorithm which specifies the supported destruction method.
There are contexts in which a client may wish to change the
key being used to protect information without having to expose the plaintext to
the client.
The Ephemeral tag within the Request Batch item allows the
client to request that the plaintext output from a cryptographic operation is
not returned to the client.
When combined with the Data Enumeration, to refer to the
output from a previous Batch Item, a client can request a Decryption followed
by an Encryption without the plaintext output from the Decryption being
returned to the client.
This section describes KMIP functionality that has been
deprecated.
Use of deprecated functionality is
discouraged since such functionality may be dropped in a future release of the [KMIP-SPEC].
Items in the normative KMIP Specification [KMIP-SPEC] document can be marked deprecated in any
document version but will be removed only in a major version. Similarly,
conformance clauses or other normative information in the KMIP Profiles [KMIP-PROFILES] document can be deprecated in any
document version but removed only in a major version. Information in the
non-normative KMIP Usage Guide [this document] and KMIP Test Cases
[KMIP-TC] documents may be removed in any document
version.
As this is the first release of
KMIP 2.0 there are presently no deprecated functions.
Appendix A.
Acknowledgments
The following individuals have participated in the creation
of this document and are gratefully acknowledged:
Participants:
Rinkesh Bansal - IBM
Jeff Bartell - Individual
Gabriel Becker - KRYPTUS
Andre Bereza - KRYPTUS
Anthony Berglas - Cryptsoft Pty
Ltd.
Mathias Bjorkqvist - IBM
Joseph Brand - Semper Fortis
Solutions
Alan Brown - Thales e-Security
Andrew Byrne - Dell
Tim Chevalier - NetApp
Kenli Chong - QuintessenceLabs Pty
Ltd.
Justin Corlett - Cryptsoft Pty Ltd.
Tony Cox - Cryptsoft Pty Ltd.
James Crossland - Northrop Grumman
Stephen Edwards - Semper Fortis
Solutions
Stan Feather - Hewlett Packard
Enterprise (HPE)
Indra Fitzgerald - Utimaco IS GmbH
Judith Furlong - Dell
Gary Gardner - Fornetix
Susan Gleeson - Oracle
Steve He - Thales e-Security
Christopher Hillier - Hewlett
Packard Enterprise (HPE)
Tim Hudson - Cryptsoft Pty Ltd.
Nitin Jain - SafeNet, Inc.
Gershon Janssen - Individual
Mark Joseph - P6R, Inc
Paul Lechner - KeyNexus Inc
John Leiseboer - QuintessenceLabs
Pty Ltd.
Jarrett Lu - Oracle
Jeff MacMillan - KeyNexus Inc
John Major - QuintessenceLabs Pty
Ltd.
Cecilia Majorel - Quintessence Labs
Gabriel Mandaji - KRYPTUS
Jon Mentzell - Fornetix
Prashant Mestri - IBM
Kevin Mooney - Fornetix
Ladan Nekuii - Thales e-Security
Jason Novecosky - KeyNexus Inc
Matt O'reilly - Fornetix
Sanjay Panchal - IBM
Mahesh Paradkar - IBM
Steve Pate - Thales e-Security
Greg Pepus - Semper Fortis
Solutions
Bruce Rich - Cryptsoft Pty Ltd.
Thad Roemer - Dyadic Security Ltd.
Thad Roemer - Unbound Tech
Greg Scott - Cryptsoft Pty Ltd.
Martin Shannon - QuintessenceLabs
Pty Ltd.
Gerald Stueve - Fornetix
Jim Susoy - P6R, Inc
Jason Thatcher - Cryptsoft Pty Ltd.
Peter Tsai - Thales e-Security
Charles White - Fornetix
Steven Wierenga - Utimaco IS GmbH
Kyle Wuolle - KeyNexus Inc
Table of Figures
Figure
1: Aggregator Client Example. 13
Figure
2: KAD Content for LTO4. 46
Figure
3: KAD Content for LTO5. 46
Table
of Tables
Table
1: ID Placeholder Prior to and Resulting from a KMIP Operation. 16
Table
2: Cryptographic Usage Masks Pairs. 18
Table
3: ECC Recommended Curve Mapping. 23
|
Revision
|
Date
|
Editor
|
Changes Made
|
|
V2.0-wd01
|
14 July 2017
|
Judy Furlong
|
Initial Draft
Copied KMIP 1.4 UG content into KMIP 2.0 starter document.
|
|
V2.0-wd01b
|
25 February 2019
|
Judy Furlong
|
Initial Draft for 2019 F2F Meeting
Now in latest KMIP CN template
|
|
V2.0-wd02
|
3 April 2019
|
Judy Furlong
|
Align with KMIP 2.0 Spec
Incorporate content developed during F2F meeting
Other editorial changes
|
|
V2.0-wd03
|
16 April 2019
|
Judy Furlong
|
Incorporate review comments
Added Acknowledgements
|
|
V2.0-wd04
|
19 April 2019
|
Judy Furlong
|
Incorporate additional review comments
|
|
[Rev number]
|
[Rev Date]
|
[Modified By]
|
[Summary of Changes]
|
