Dynamic Symmetric Key Provisioning Protocol
(DSKPP)RSA, The Security Division of EMC174 Middlesex TurnpikeBedfordMA01730USAandrea.doherty@rsa.comVerisign, Inc.487 E. Middlefield RoadMountain ViewCA94043USAmpei@verisign.comDiversinet Corp.2225 Sheppard Avenue East, Suite 1801TorontoOntarioM2J 5C2Canadasmachani@diversinet.comRSA, The Security Division of EMCArenavagen 29StockholmStockholm Ln121 29SEmagnus.nystrom@rsa.com
Security Area
KEYPROV Working GroupDSKPP is a client-server protocol for initialization (and
configuration) of symmetric keys to locally and remotely accessible
cryptographic modules. The protocol can be run with or without
private-key capabilities in the cryptographic modules, and with or
without an established public-key infrastructure.Two variations of the protocol support multiple usage scenarios. With
the four-pass variant, keys are mutually generated by the provisioning
server and cryptographic module; provisioned keys are not transferred
over-the-wire or over-the-air. The two-pass variant enables secure and
efficient download and installation of pre-generated symmetric keys to a
cryptographic module.This document builds on information contained in , adding specific enhancements in response to
implementation experience and liaison requests.Symmetric key based cryptographic systems (e.g., those providing
authentication mechanisms such as one-time passwords and
challenge-response) offer performance and operational advantages over
public key schemes. Such use requires a mechanism for provisioning of
symmetric keys providing equivalent functionality to mechanisms such as
CMP and CMMC in a Public Key Infrastructure.Traditionally, cryptographic modules have been provisioned with keys
during device manufacturing, and the keys have been imported to the
cryptographic server using, e.g., a CD-ROM disc shipped with the
devices. Some vendors also have proprietary provisioning protocols,
which often have not been publicly documented (CT-KIP is one exception
).This document describes the Dynamic Symmetric Key Provisioning
Protocol (DSKPP), a client-server protocol for provisioning symmetric
keys between a cryptographic module (corresponding to DSKPP client) and
a key provisioning server (corresponding to DSKPP server).DSKPP provides an open and interoperable mechanism for initializing
and configuring symmetric keys to cryptographic modules that are
accessible over the Internet. The description is based on the
information contained in , and contains
specific enhancements, such as User Authentication and support for the
format for transmission of keying
material.DSKPP has two principal protocol variants. The four-pass protocol
variant permits a symmetric key to be established that includes
randomness contributed by both the client and the server. The two-pass
protocol requires only one round trip instead of two and permits a
server specified key to be established.The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in .There is a provision made in the syntax for an explicit version
number. Only version "1.0" is currently specified.This document uses Uniform Resource Identifiers to identify resources, algorithms, and
semantics.The XML namespace URI for Version
1.0 of DSKPP protocol is:xmlns:dskpp="urn:ietf:params:xml:ns:keyprov:dskpp:1.0"References to qualified elements in the DSKPP schema defined
herein use the prefix "dskpp".This document relies on qualified elements already defined in the
Portable Symmetric Key Container
namespace, which is represented by the prefix "pskc" and declared
as:xmlns:pskc="urn:ietf:params:xml:ns:keyprov:pskc:1.0"Finally, the DSKPP syntax presented in this document relies on
algorithm identifiers defined in the XML Signature namespace:xmlns:ds="http://www.w3.org/2000/09/xmldsig#"References to algorithm identifiers in the XML Signature
namespace are represented by the prefix "ds".The definitions provided below are defined as used in this
document. The same terms may be defined differently in other
documents.User Authentication Code
comprised of a string of numeric characters known to the device
and the server and containing a client identifier and a password.
This ClientID/password combination is used only once, and then
discarded.User Authentication Data
that is derived from the Authentication Code (AC)An identifier that the DSKPP Server uses
to locate the real user name or account identifier on the server.
It can be a short random identifier that is unrelated to any real
usernames.A component of an application,
which enables symmetric key cryptographic functionalityA physical piece of hardware, or a software
framework, that hosts symmetric key cryptographic modulesA unique identifier for the
device that houses the cryptographic module, e.g., a mobile
phoneManages communication between the
symmetric key cryptographic module and the DSKPP serverThe symmetric key provisioning server
that participates in the DSKPP protocol runThe unique identifier of
a DSKPP serverAn organization that issues symmetric
keys to end-usersAn object that encapsulates a
symmetric key and its configuration dataA unique identifier for the
symmetric keyThe key transport
method used during two-pass DSKPPThe list of key
protection methods supported by a cryptographic moduleA lifecycle management
system that provides a key issuer with the ability to provision
keys to cryptographic modules hosted on end-users' devicesA key establishment procedure whereby
the DSKPP server selects and encrypts the keying material and then
sends the material to the DSKPP client The private key that resides on
the cryptographic module. This key is paired with the DSKPP
client's public key, which the DSKPP server uses to encrypt keying
material during key transport The type of symmetric key cryptographic
methods for which the key will be used (e.g., OATH HOTP or RSA
SecurID authentication, AES encryption, etc.)A method of encrypting keys for key
transport A symmetric key encrypting key
used for key wrapping The data necessary (e.g., keys and
key configuration data) necessary to establish and maintain
cryptographic keying relationships A unique master key pre-issued to
a hardware device, e.g., a smart card, during the manufacturing
process. If present, this key may be used by a cryptographic
module to derive secret keysA payload that
contains the DSKPP version, DSKPP variant (four- or two-pass), key
package formats, key types, and cryptographic algorithms that the
cryptographic module is capable of supportingA payload that contains the
DSKPP version, DSKPP variant (four- or two-pass), key package
format, key type, and cryptographic algorithms relevant to the
current protocol runString concatenationOptional element xExclusive-OR operation on strings A and B
(where A and B are of equal length)A typographical convention used
in the body of the textA keyed pseudo-random
functionEncryption of m with the key kKey used to encrypt R_C (either K_SERVER or
K_SHARED), or in MAC or DSKPP_PRF computationsSecret key that is derived from the
Authentication Code and used for user authentication
purposesSecret key derived during a DSKPP exchange for
use with key confirmationA second secret key used for server
authenticationA provisioning master key from which two keys
are derived: K_TOKEN and K_MACPublic key of the DSKPP server; used for
encrypting R_C in the four-pass protocol variantSecret key that is pre-shared between the
DSKPP client and the DSKPP server; used for encrypting R_C in the
four-pass protocol variantSecret key that is established in a
cryptographic module using DSKPPPseudorandom value chosen by the DSKPP client and
used for MAC computationsPseudorandom value chosen by the DSKPP client
and used as input to the generation of K_TOKENPseudorandom value chosen by the DSKPP server
and used as input to the generation of K_TOKENDSKPP server address, as a URLAuthentication CodeAuthentication DataDynamic Symmetric Key Provisioning
ProtocolHypertext Transfer ProtocolKey PackageKey Protection MethodKey Protection Method ListMessage Authentication CodePersonal ComputerProtocol Data UnitPublic-Key Cryptography StandardsPseudo-Random FunctionPortable Symmetric Key ContainerSecurity Attribute List (see )Security Context (see )Transport Layer SecurityUniform Resource LocatorUniversal Serial BuseXtensible Markup LanguageThe following sub-sections provide a high-level view of protocol
internals and how they interact with external provisioning applications.
Usage scenarios are provided in .A DSKPP provisioning transaction has three entities:The DSKPP provisioning server.The cryptographic module to
which the symmetric keys are to be provisioned, e.g., an
authentication token.The DSKPP client which manages communication
between the cryptographic module and the key provisioning
server.While it is highly desirable for the entire communication between
the DSKPP client and server to be protected by means of a transport
providing confidentiality and integrity protection such as HTTP over
Transport Layer Security (TLS), such protection is not sufficient to
protect the exchange of the symmetric key data between the server and
the cryptographic module and the DSKPP protocol is designed to permit
implementations that satisfy this requirement.The server only communicates to the client. As far as the server is
concerned, the client and cryptographic module may be considered to be
a single entity.From a client-side security perspective, however, the client and
the cryptographic module are separate logical entities and may in some
implementations be separate physical entities as well.It is assumed that a device will host an application layered above
the cryptographic module, and this application will manage
communication between the DSKPP client and cryptographic module. The
manner in which the communicating application will transfer DSKPP
protocol elements to and from the cryptographic module is transparent
to the DSKPP server. One method for this transfer is described in
.In a DSKPP message flow, the user has obtained a new hardware or
software device embedded with a cryptographic module. The goal of
DSKPP is to provision the same symmetric key and related information
to the cryptographic module and the key management server, and
associate the key with the correct user name (or other account
identifier) on the server. To do this, the DSKPP Server MUST
authenticate the user to be sure he is authorized for the new
key.User authentication occurs within the protocol itself afterthe DSKPP client initiates the first message.
In this case, the DSKPP client MUST have access to the DSKPP Server
URL.Alternatively, a DSKPP web service or other form of web
application can authenticate a user beforethe first message is exchanged. In this case, the DSKPP
server MUST trigger the DSKPP client to initiate the first message
in the protocol transaction.In the following example, the DSKPP client first initiates DSKPP,
and then the user is authenticated using a Client ID and
Authentication Code.Before DSKPP begins:The Authentication Code is generated by the DSKPP Server, and
delivered to the user via an out-of-band trustworthy channel
(e.g., a paper slip delivered by IT department staff).The user typically enters the Client ID and Authentication
Code manually, possibly on a device with only numeric keypad.
Thus, they are often short numeric values (for example, 8
decimal digits). However, the DSKPP Server is free to generate
them in any way it wishes.The DSKPP client needs the URL of the DSKPP server (which is
not user-specific or secret, and may be pre-configured somehow),
and a set of trust anchors for verifying the server
certificate.There must be an account for the user that has an identifier
and long-term user name (or other account identifier) to which
the token will be associated. The DSKPP server will use the
Client ID to find the corresponding Authentication Code for user
authenticationIn Step 1, the client establishes a TLS connection, and
authenticates the server (that is, validates the certificate, and
compares the host name in the URL with the certificate).Next, the DSKPP Client and DSKPP Server exchange DSKPP messages
(which are sent over HTTPS). In these messages:The client and server negotiate which cryptographic
algorithms they want to use; which algorithms are supported for
protecting DSKPP messages, and other DSKPP protocol details.The client sends the Client ID to the server, and proves that
it knows the corresponding Authentication Code.The client and server agree on a secret key (token key or
K_TOKEN); depending on the negotiated protocol variant, this is
either a fresh key derived during the DSKPP protocol run (called
"four-pass variant", since it involves four DSKPP messages), or
it is generated by (or pre-exists on) the server and transported
to the client (called "two-pass variant" in the rest of this
document, since it involves two DSKPP messages).The server sends a "key package" to the client. The package
only includes the key itself in the case of the "two-pass
variant"; with either variant, the key package contains
attributes that influence how the provisioned key will be later
used by the cryptographic module and cryptographic server. The
exact contents depend on the cryptographic algorithm (e.g., for
a one-time password algorithm that supports variable-length OTP
values, the length of the OTP value would be one attribute in
the key package).After the protocol run has been successfully completed, the
cryptographic modules stores the contents of the key package.
Likewise, the DSKPP provisioning server stores the contents of the
key package with the cryptographic server, and associates these with
the correct user name. The user can now use the their device to
perform symmetric-key based operations.The exact division of work between the cryptographic module and
the DSKPP client -- and key Provisioning server and DSKPP server --
are not specified in this document. The figure above shows one
possible case, but this is intended for illustrative purposes
only.In the first message flow (previous section), the Client ID and
Authentication Code were delivered to the user by some out-of-band
means (such as paper).In the second message flow, the user first authenticates to a web
server (for example, IT department's "self-service" Intranet page),
using an ordinary web browser and some existing credentials.The user then requests (by clicking a link or submitting a form)
provisioning of a new key to the cryptographic module. The web
server will reply with a <KeyProvTrigger> message that
contains the Client ID, Authentication Code, and URL of the DSKPP
server. This information is also needed by the DSKPP server; how the
web server and DSKPP server interact is beyond the scope of this
document.The <KeyProvTrigger> message is sent in a HTTP response,
and it is marked with MIME type
"application/vnd.ietf.keyprov.dskpp+xml". It is assumed the web
browser has been configured to recognize this MIME type; the browser
will start the DSKPP client, and provides it with the
<KeyProvTrigger> message.The DSKPP client then contacts the DSKPP server, and uses the
Client ID and Authentication Code (from the <KeyProvTrigger>
messsage) the same way as in the first message flow.As noted in the previous section, once the protocol has started,
the client and server MAY engage in either a two-pass or four-pass
message exchange. The four-pass and two-pass protocols are
appropriate in different deployment scenarios. The biggest
differentiator between the two is that the two-pass protocol
supports transport of an existing key to a cryptographic module,
while the four-pass involves key generation on-the-fly via key
agreement. In either case, both protocol variants support algorithm
agility through negotiation of encryption mechanisms and key types
at the beginning of each protocol run.The four-pass protocol is needed under one or more of the
following conditions:Policy requires that both parties engaged in the protocol
jointly contribute entropy to the key. Enforcing this policy
mitigates the risk of exposing a key during the provisioning
process as the key is generated through mutual agreement
without being transferred over-the-air or over-the-wire. It
also mitigates risk of exposure after the key is provisioned,
as the key will be not be vulnerable to a single point of
attack in the system.A cryptographic module does not have private-key
capabilities.The cryptographic module is hosted by a device that was
neither pre-issued with a manufacturer's key or other form of
pre-shared key (as might be the case with a smart card or SIM
card) nor has a keypad that can be used for entering a
passphrase (such as present on a mobile phone).The two-pass protocol is needed under one or more of the
following conditions:Pre-existing (i.e., legacy) keys must be provisioned via
transport to the cryptographic module.The cryptographic module is hosted on a device that was
pre-issued with a manufacturer's key (such as may exist on a
smart card), or other form of pre-shared key (such as may
exist on a SIM-card), and is capable of performing private-key
operations.The cryptographic module is hosted by a device that has a
built-in keypad with which a user may enter a passphrase,
useful for deriving a key wrapping key for distribution of
keying material.Upon transmission or receipt of a message for which the Status
attribute's value is not "Success" or "Continue", the default
behavior, unless explicitly stated otherwise below, is that both the
DSKPP server and the DSKPP client MUST immediately terminate the DSKPP
protocol run. DSKPP servers and DSKPP clients MUST delete any secret
values generated as a result of failed runs of the DSKPP protocol.
Session identifiers MAY be retained from successful or failed protocol
runs for replay detection purposes, but such retained identifiers MUST
NOT be reused for subsequent runs of the protocol.When possible, the DSKPP client SHOULD present an appropriate error
message to the user.These status codes are valid in all DSKPP Response messages unless
explicitly stated otherwise:"Continue" indicates that the DSKPP server is ready for a
subsequent request from the DSKPP client. It cannot be sent in the
server's final message."Success" indicates successful completion of the DSKPP session.
It can only be sent in the server's final message."Abort" indicates that the DSKPP server rejected the DSKPP
client's request for unspecified reasons."AccessDenied" indicates that the DSKPP client is not
authorized to contact this DSKPP server."MalformedRequest" indicates that the DSKPP server failed to
parse the DSKPP client's request."UnknownRequest" indicates that the DSKPP client made a request
that is unknown to the DSKPP server."UnknownCriticalExtension" indicates that a critical DSKPP
extension (see below) used by the DSKPP client was not supported
or recognized by the DSKPP server."UnsupportedVersion" indicates that the DSKPP client used a
DSKPP protocol version not supported by the DSKPP server. This
error is only valid in the DSKPP server's first response
message."NoSupportedKeyTypes" indicates that the DSKPP client only
suggested key types that are not supported by the DSKPP server.
This error is only valid in the DSKPP server's first response
message."NoSupportedEncryptionAlgorithms" indicates that the DSKPP
client only suggested encryption algorithms that are not supported
by the DSKPP server. This error is only valid in the DSKPP
server's first response message."NoSupportedMacAlgorithms" indicates that the DSKPP client only
suggested MAC algorithms that are not supported by the DSKPP
server. This error is only valid in the DSKPP server's first
response message."NoProtocolVariants" indicates that the DSKPP client only
suggested a protocol variant (either 2-pass or 4-pass) that is not
supported by the DSKPP server. This error is only valid in the
DSKPP server's first response message."NoSupportedKeyPackages" indicates that the DSKPP client only
suggested key package formats that are not supported by the DSKPP
server. This error is only valid in the DSKPP server's first
response message."AuthenticationDataMissing" indicates that the DSKPP client
didn't provide authentication data that the DSKPP server
required."AuthenticationDataInvalid" indicates that the DSKPP client
supplied user authentication data that the DSKPP server failed to
validate."InitializationFailed" indicates that the DSKPP server could
not generate a valid key given the provided data. When this status
code is received, the DSKPP client SHOULD try to restart DSKPP, as
it is possible that a new run will succeed."ProvisioningPeriodExpired" indicates that the provisioning
period set by the DSKPP server has expired. When the status code
is received, the DSKPP client SHOULD report the reason for key
initialization failure to the user and the user MUST register with
the DSKPP server to initialize a new key.The following calculations are used in both DSKPP protocol
variants.User authentication data (AD) is derived from a Client ID and
Authentication Code that the user enters before the first DSKPP
message is sent.Note: The user will typically enter the Client ID and
Authentication Code manually, possibly on a device with only numeric
keypad. Thus, they are often short numeric values (for example, 8
decimal digits). However, the DSKPP Server is free to generate them
in any way it wishes.AC is encoded in Type-Length-Value (TLV) format. The format
consists of a minimum of two TLVs and a variable number of
additional TLVs, depending on implementation.The TLV fields are defined as follows:The integer value identifying the
type of information contained in the value field.The length, in hexadecimal, of
the value field to follow.A variable-length
hexadecimal value containing the instance-specific information
for this TLV.A 1 byte type field identifies the specific TLV, and a 1 byte
length, in hexadecimal, indicates the length of the value field
contained in the TLV. A TLV MUST start on a 4 byte boundary. Pad
bytes MUST be placed at the end of the previous TLV in order to
align the next TLV. These pad bytes are not counted in the length
field of the TLV.The following table summarizes the TLVs defined in this
document. Optional TLVs are allowed for vendor-specific extensions
with the constraint that the high bit MUST be set to indicate a
vendor-specific type. Other TLVs are left for later revisions of
this protocol.The Client ID is a mandatory TLV that represents the
requester's identifier of maximum length 128. The value is
represented as an ASCII string that identifies the key request.
The clientID MUST be HEX encoded. For example, suppose clientID is
set to "AC00000A", the hexadecimal equivalent is
0x4143303030303041, resulting in a TLV of {0x1, 0x8,
0x4143303030303041}.The Password is a mandatory TLV the contains a one-time use
shared secret known by the user and the Provisioning Server. The
password value is unique and SHOULD be a random string to make AC
more difficult to guess. The string MUST be UTF-8 encoded in
accordance with . For example,
suppose password is set to "3582", then the TLV would be {0x2,
0x4, UTF-8("3582")}.The Checksum is an OPTIONAL TLV, which is generated by the
issuing server and sent to the user as part of the AC. If the TLV
is provided, the checksum value MUST be computed using the CRC16
algorithm . When the user enters the
AC, the typed password is verified with the checksum to ensure it
is correctly entered by the user. For example, suppose the
Password is set to "3582", then the CRC16 calculation would
generate a checksum of 0x5F8D, resulting in TLV {0x3, 0x2,
0x5F8D}.The Authentication Data consists of a Client ID (extracted from
the AC) and a value, which is derived from AC as follows (refer to
for a description of DSKPP-PRF in
general and for a
description of DSKPP-PRF-AES):MAC = DSKPP-PRF(K_AC, AC->clientID||URL_S||R_C||[R_S],
16)In four-pass DSKPP, the cryptographic module uses R_C, R_S, and
URL_S to calculate the MAC, where URL_S is the URL the DSKPP
client uses when contacting the DSKPP server. In two-pass DSKPP,
the cryptographic module does not have access to R_S, therefore
only R_C is used in combination with URL_S to produce the MAC. In
either case, K_AC MUST be derived from AC>password as follows
:K_AC = PBKDF2(AC->password, R_C || K, iter_count, 16)One of the following values for K MUST be used:In four-pass: The public key of the DSKPP server (K_SERVER), or (in
the pre-shared key variant) the pre-shared key between the
client and the server (K_SHARED)In two-pass:The public key of the DSKPP client, or the public key
of the device when a device certificate is availableThe pre-shared key between the client and the server
(K_SHARED)A passphrase-derived keyThe iteration count, iter_count,
MUST be set to at least 100,000 except for case (b) and (c),
above, in which case it MUST be set to 1.Regardless of the protocol variant employed, there is a
requirement for a cryptographic primitive that provides a
deterministic transformation of a secret key k and a varying length
octet string s to a bitstring of specified length dsLen.This primitive must meet the same requirements as for a keyed
hash function: It MUST take an arbitrary length input, and generate
an output that is one-way and collision-free (for a definition of
these terms, see, e.g., ). Further, its
output MUST be unpredictable even if other outputs for the same key
are known.From the point of view of this specification, DSKPP-PRF is a
"black-box" function that, given the inputs, generates a
pseudorandom value and MAY be realized by any appropriate and
competent cryptographic technique. contains two example
realizations of DSKPP-PRF.DSKPP-PRF(k, s, dsLen)Input:secret key in octet string formatoctet string of varying length consisting of
variable data distinguishing the particular string being
deriveddesired length of the outputOutput:pseudorandom string, dsLen-octets longFor the purposes of this document, the secret key k MUST be at
least 16 octets long.When sending its last message in a protocol run, the DSKPP server
generates a MAC that is used by the client for key confirmation.
Computation of the MAC MUST include a hash of all DSKPP messages
sent by the client and server during the transaction. To compute a
message hash for the MAC given a sequence of DSKPP messages msg_1,
..., msg_n, the following operations MUST be carried out:The sequence of messages contains all DSKPP Request and
Response messages up to but not including this message.Re-transmitted messages are removed from the sequence of
messages.Note: The resulting sequence
of messages MUST be an alternating sequence of DSKPP Request and
DSKPP Response messagesThe contents of each message is concatenated together.The resultant string is hashed using SHA-256 in accordance
with .This section describes the methods and message flow that comprise the
four-pass protocol variant. Four-pass DSKPP depends on a client-server
key agreement mechanism.With 4-pass DSKPP, the symmetric key that is the target of
provisioning, is generated on-the-fly without being transferred
between the DSKPP client and DSKPP server. The data flow and
computation are described below.A sample data flow showing key generation during the 4-pass
protocol is shown in .The inclusion of the two random nonces (R_S and R_C) in the key
generation provides assurance to both sides (the cryptographic
module and the DSKPP server) that they have contributed to the key's
randomness and that the key is unique. The inclusion of the
encryption key (K) ensures that no man-in-the-middle may be present,
or else the cryptographic module will end up with a key different
from the one stored by the legitimate DSKPP server.Notes: Conceptually, although R_C is one pseudorandom string, it may
be viewed as consisting of two components, R_C1 and R_C2, where
R_C1 is generated during the protocol run, and R_C2 can be
pre-generated and loaded on the cryptographic module before the
device is issued to the user. In that case, the latter string,
R_C2, SHOULD be unique for each cryptographic module.A man-in-the-middle (in the form of corrupt client software
or a mistakenly contacted server) may present his own public key
to the cryptographic module. This will enable the attacker to
learn the client's version of K_TOKEN. However, the attacker is
not able to persuade the legitimate server to derive the same
value for K_TOKEN, since K_TOKEN is a function of the public key
involved, and the attacker's public key must be different than
the correct server's (or else the attacker would not be able to
decrypt the information received from the client). Therefore,
once the attacker is no longer "in the middle," the client and
server will detect that they are "out of sync" when they try to
use their keys. In the case of encrypting R_C with K_SERVER, it
is therefore important to verify that K_SERVER really is the
legitimate server's key. One way to do this is to independently
validate a newly generated K_TOKEN against some validation
service at the server (e.g. using a connection independent from
the one used for the key generation).In DSKPP, the client and server both generate K_TOKEN and K_MAC
by deriving them from a provisioning key (K_PROV) using the
DSKPP-PRF function (refer to ) as
follows:K_PROV = DSKPP-PRF(k,s,dsLen), wherek = R_C (i.e., the secret random value chosen by the DSKPP
client)s = "Key generation" || K || R_S (where K is the key used to
encrypt R_C and R_S is the random value chosen by the DSKPP
server)dsLen = (desired length of K_PROV whose first half
constitutes K_MAC and second half constitutes K_TOKEN)Then K_TOKEN and K_MAC are derived from K_PROV, whereK_PROV = K_MAC || K_TOKENWhen computing K_PROV, the derived keys, K_MAC and K_TOKEN, MAY
be subject to an algorithm-dependent transform before being adopted
as a key of the selected type. One example of this is the need for
parity in DES keys.The four-pass protocol flow consists of two message exchanges:Pass 1 = <KeyProvClientHello>, Pass 2 =
<KeyProvServerHello>Pass 3 = <KeyProvClientNonce>, Pass 4 =
<KeyProvServerFinished>The first pair of messages negotiate cryptographic algorithms and
exchange nonces. The second pair of messages establishes a symmetric
key using mutually authenticated key agreement.The purpose and content of each message are described below. XML
format and examples are in and
.When this message is sent:The "trigger" message is optional. The DSKPP server sends
this message after the following out-of-band steps are
performed:A user directed their browser to a key provisioning web
application and signs in (i.e., authenticates)The user requests a keyThe web application processes the request and returns an
authentication code to the user, e.g., in the form of an
email messageThe web application retrieves the authentication code
from the user (possibly by asking the user to enter it using
a web form, or alternatively by the user selecting a URL in
which the authentication code is embedded)The web application derives authentication data (AD) from
the authentication code as described in The web application passes AD, and possibly a DeviceID
(identifies a particular device to which the key MUST be
provisioned) and/or KeyID (identifies a key that will be
replaced) to the DSKPP serverPurpose of this message:To start a DSKPP session: The DSKPP server uses this message
to trigger a client-side application to send the first DSKPP
message.To provide a way for the key provisioning system to get the
DSKPP server URL to the DSKPP client.So
the key provisioning system can point the DSKPP client to a
particular cryptographic module that was pre-configured in the
DSKPP provisioning server.In the case
of key renewal, to identify the key to be replaced.What is contained in this message:AD MUST be provided to allow the DSKPP server to authenticate
the user before completing the protocol run.A DeviceID MAY be included to allow a key
provisioning application to bind the provisioned key to a
specific device.A KeyID MAY be included
to allow the key provisioning application to identify a key to
be replaced, e.g., in the case of key renewal.The Server URL MAY be included to allow the key
provisioning application to inform the DSKPP client of which
server to contactWhen this message is sent:When a DSKPP client first connects to a DSKPP server, it is
required to send the <KeyProvClientHello> as its first
message. The client can also send a <KeyProvClientHello>
in response to a <KeyProvTrigger>.What is contained in this message:The Security Attribute List (SAL) included with
<KeyProvClientHello> contains the combinations of DSKPP
versions, variants, key package formats, key types, and
cryptographic algorithms that the DSKPP client supports in order
of the client's preference (favorite choice first). If <KeyProvClientHello> was preceded by a
<KeyProvTrigger>, then this message MUST also include the
Authentication (AD), DeviceID, and/or KeyID that was provided
with the trigger. If
<KeyProvClientHello> was not preceded by a
<KeyProvTrigger>, then this message MAY contain a device
ID that was pre-shared with the DSKPP server, and a key ID
associated with a key previously provisioned by the DSKPP
provisioning server.Application note:If this message is preceded by trigger message
<KeyProvTrigger>, then the application will already have
AD available (see ).
However, if this message was not preceded by
<KeyProvTrigger>, then the application MUST retrieve the
user authentication code, possibly by prompting the user to
manually enter their authentication code, e.g., on a device with
only a numeric keypad.The application MUST also derive Authentication Data (AD)
from the authentication code, as described in , and save it for use in its next
message, <KeyProvClientNonce>.How the DSKPP server uses this message:The DSKPP server will look for an acceptable combination of
DSKPP version, variant (in this case, four-pass), key package
format, key type, and cryptographic algorithms. If the DSKPP
Client's SAL does not match the capabilities of the DSKPP
Server, or does not comply with key provisioning policy, then
the DSKPP Server will set the Status attribute to something
other than "Continue". Otherwise, Status will be set to
"Continue".If included in <KeyProvClientHello>, the DSKPP server
will validate the Authentication Data (AD), DeviceID, and KeyID.
The DSKPP server MUST NOT accept the DeviceID unless the server
sent the DeviceID in a preceding trigger message. Note that it
is also legitimate for a DSKPP client to initiate the DSKPP
protocol run without having received a <KeyProvTrigger>
message from a server, but in this case any provided DeviceID
MUST NOT be accepted by the DSKPP server unless the server has
access to a unique key for the identified device and that key
will be used in the protocol.When this message is sent:The DSKPP server will send this message in response to a
<KeyProvClientHello> message after it looks for an
acceptable combination of DSKPP version, variant (in this case,
four-pass), key package format, key type, and set of
cryptographic algorithms. If it could not find an acceptable
combination, then it will still send the message, but with a
failure status.Purpose of this message:With this message, the context for the protocol run is set.
Furthermore, the DSKPP server uses this message to transmit a
random nonce, which is required for each side to agree upon the
same symmetric key (K_TOKEN).What is contained in this message:A status attribute equivalent to the server's return code to
<KeyProvClientHello>. If the server found an acceptable
set of attributes from the client's SAL, then it sets status to
Continue and returns an SC, which specifies the DSKPP version
and variant (in this case, four-pass), key type, cryptographic
algorithms, and key package format that the DSKPP Client MUST
use for the remainder of the protocol run.A random nonce (R_S) for use in generating a symmetric key
through key agreement; the length of R_S may depend on the
selected key type. A key (K) for the DSKPP Client to use for encrypting the
client nonce included with <KeyProvClientNonce>. K
represents the server's public key (K_SERVER) or a pre-shared
secret key (K_SHARED).A MAC MUST be
present if a key is being renewed so that the DSKPP client can
confirm that the replacement key came from a trusted server.
This MAC MUST be computed using DSKPP-PRF (see ), where the input parameter k MUST be
set to the existing MAC key K_MAC' (i.e., the value of the MAC
key that existed before this protocol run; the implementation
MAY specify K_MAC' to be the value of the K_TOKEN that is being
replaced, or a version of K_MAC from the previous protocol run),
and input parameter dsLen MUST be set to the length of R_S.How the DSKPP client uses this message:When the Status attribute is not set to "Continue", this
indicates failure and the DSKPP client MUST abort the
protocol.If successful execution of the protocol will result in the
replacement of an existing key with a newly generated one, the
DSKPP client MUST verify the MAC provided in
<KeyProvServerHello>. The DSKPP client MUST terminate the
DSKPP session if the MAC does not verify, and MUST delete any
nonces, keys, and/or secrets associated with the failed
run.If Status is set to "Continue" the cryptographic module
generates a random nonce (R_C) using the cryptographic algorithm
specified in SC. The length of the nonce R_C will depend on the
selected key type.Encrypt R_C using K and the encryption algorithm included in
SC.The method the DSKPP client MUST use to encrypt R_C:If K is equivalent to K_SERVER (i.e., the public key of the
DSKPP server), then an RSA encryption scheme from PKCS #1 MAY be used. If K is equivalent to
K_SERVER, then the cryptographic module SHOULD verify the
server's certificate before using it to encrypt R_C in
accordance with .If K is equivalent to K_SHARED, the DSKPP
client MAY use the DSKPP-PRF function to avoid dependence on
other algorithms. In this case, the client uses K_SHARED as
input parameter k (K_SHARED SHOULD be used solely for this
purpose) as follows:dsLen = len(R_C),
where "len" is the length of R_CDS =
DSKPP-PRF(K_SHARED, "Encryption" || R_S, dsLen)This will produce a pseudorandom string DS of
length equal to R_C. Encryption of R_C MAY then be achieved by
XOR-ing DS with R_C:E(DS, R_C) = DS ^
R_CThe DSKPP server will then perform
the reverse operation to extract R_C from E(DS, R_C).When this message is sent:The DSKPP client will send this message immediately following
a <KeyProvServerHello> message whose status was set to
"Continue".Purpose of this message:With this message the DSKPP client transmits user
authentication data (AD) and a random nonce encrypted with the
DSKPP server's key (K). The client's random nonce is required
for each side to agree upon the same symmetric key
(K_TOKEN).What is contained in this message:Authentication Data (AD) that was derived from an
authentication code entered by the user before
<KeyProvClientHello> was sent (refer to ).The DSKPP client's random nonce (R_C), which was encrypted as
described in .How the DSKPP server uses this message:The DSKPP server MUST use AD to authenticate the user. If
authentication fails, then the DSKPP server MUST set the return
code to a failure status. If user
authentication passes, the DSKPP server decrypts R_C using its
key (K). The decryption method is based on whether K that was
transmitted to the client in <KeyProvServerHello> was
equal to the server's public key (K_SERVER) or a pre-shared key
(K_SHARED) (refer to for a description of how
the DSKPP client encrypts R_C).After extracting R_C, the DSKPP server computes K_TOKEN using
a combination of the two random nonces R_S and R_C and its
encryption key, K, as described in . The DSKPP server then
generates a key package that contains key usage attributes such
as expiry date and length. The key package MUST NOT include
K_TOKEN since in the four-pass variant K_TOKEN is never
transmitted between the DSKPP server and client. The server
stores K_TOKEN and the key package with the user's account on
the cryptographic server.Finally, the server generates a key
confirmation MAC that the client will use to avoid a false
"Commit" message that would cause the cryptographic module to
end up in state in which the server does not recognize the
stored key.The MAC used for key confirmation MUST be calculated as follows:
msg_hash = SHA-256(msg_1, ..., msg_n)dsLen = len(msg_hash)MAC = DSKPP-PRF (K_MAC, "MAC 2 computation" || msg_hash,
dsLen)whereThe DSKPP Pseudo-Random Function defined
in is used to compute the
MAC. The particular realization of DSKPP-PRF (e.g., those
defined in
depends on the MAC algorithm contained in the
<KeyProvServerHello> message. The MAC MUST be computed
using the existing MAC key (K_MAC), and a string that is
formed by concatenating the (ASCII) string "MAC 2
computation" and a msg_hashThe key derived from K_PROV, as
described in .The message hash (defined in ) of messages msg_1,
..., msg_n.When this message is sent:The DSKPP server will send this message after authenticating
the user and, if authentication passed, generating K_TOKEN and a
key package, and associating them with the user's account on the
cryptographic server.Purpose of this message:With this message the DSKPP server confirms generation of the
key (K_TOKEN), and transmits the associated identifier and
application-specific attributes, but not the key itself, in a
key package to the client for protocol completion.What is contained in this message:A status attribute equivalent to the server's return code to
<KeyProvClientNonce>. If user authentication passed, and
the server successfully computed K_TOKEN, generated a key
package, and associated them with the user's account on the
cryptographic server, then it sets Status to Continue.If status is Continue, then this message acts as a "commit"
message, instructing the cryptographic module to store the
generated key (K_TOKEN) and associate the given key identifier
with this key. As such, a key package (KP) MUST be included in
this message, which holds an identifier for the generated key
(but not the key itself) and additional configuration, e.g., the
identity of the DSKPP server, key usage attributes, etc. The
default symmetric key package format MUST be based on the
Portable Symmetric Key Container (PSKC) defined in . Alternative formats MAY include , PKCS#12 , or PKCS#5 XML format. With KP, the server includes a key confirmation
MAC that the client uses to avoid a false "Commit".How the DSKPP client uses this message:When the Status attribute is not set to "Continue", this
indicates failure and the DSKPP client MUST abort the
protocol.After receiving a <KeyProvServerFinished> message with
Status = "Success", the DSKPP client MUST verify the key
confirmation MAC that was transmitted with this message. The
DSKPP client MUST terminate the DSKPP session if the MAC does
not verify, and MUST, in this case, also delete any nonces,
keys, and/or secrets associated with the failed run of the
protocol. If <KeyProvServerFinished> has Status = "Success" and
the MAC was verified, then the DSKPP client MUST calculate
K_TOKEN from the combination of the two random nonces R_S and
R_C and the server's encryption key, K, as described in . The DSKPP client
associates the key package contained in
<KeyProvServerFinished> with the generated key, K_TOKEN,
and stores this data permanently on the cryptographic module.
After this operation, it MUST NOT be possible to overwrite
the key unless knowledge of an authorizing key is proven through
a MAC on a later <KeyProvServerHello> (and
<KeyProvServerFinished>) message.This section describes the methods and message flow that comprise the
two-pass protocol variant. Two-pass DSKPP is essentially a transport of
keying material from the DSKPP server to the DSKPP client. The DSKPP
server transmits keying material in a key package formatted in
accordance with , , PKCS#12 , or
PKCS#5 XML .The keying material includes a provisioning master key, K_PROV, from
which the DSKPP client derives two keys: the symmetric key to be
established in the cryptographic module, K_TOKEN, and a key, K_MAC, used
for server authentication and key confirmation. The keying material also
includes key usage attributes, such as expiry date and length.The DSKPP server encrypts K_PROV to ensure that it is not exposed to
any other entity than the DSKPP server and the cryptographic module
itself. The DSKPP server uses any of three key protection methods to
encrypt K_PROV: Key Transport, Key Wrap, and Passphrase-Based Key Wrap
Key Protection Methods.This section introduces three key protection methods for the
two-pass variant. Additional methods MAY be defined by external
entities or through the IETF process.Purpose of this method:This method is intended for PKI-capable devices. The DSKPP
server encrypts keying material and transports it to the DSKPP
client. The server encrypts the keying material using the public
key of the DSKPP client, whose private key part resides in the
cryptographic module. The DSKPP client decrypts the keying
material and uses it to derive the symmetric key, K_TOKEN.This method MUST be identified with the following URN: urn:ietf:params:xml:schema:keyprov:dskpp#transportThe DSKPP server and client MUST support the following
mechanism:http://www.w3.org/2001/04/xmlenc#rsa-1_5 encryption mechanism
defined in .Purpose of this method:This method is ideal for pre-keyed devices, e.g., SIM cards.
The DSKPP server encrypts keying material using a pre-shared key
wrapping key and transports it to the DSKPP client. The DSKPP
client decrypts the keying material, and uses it to derive the
symmetric key, K_TOKEN.This method MUST be identified with the following URN: urn:ietf:params:xml:schema:keyprov:dskpp#wrapThe DSKPP server and client MUST support one of the following key
wrapping mechanisms:KW-AES128 without padding; refer to
http://www.w3.org/2001/04/xmlenc#kw-aes128 in KW-AES128 with padding; refer to
http://www.w3.org/2001/04/xmlenc#kw-aes128 in AES-CBC-128; refer to Purpose of this method:This method is a variation of the Key Wrap Method that is
applicable to constrained devices with keypads, e.g., mobile
phones. The DSKPP server encrypts keying material using a
wrapping key derived from a user-provided passphrase, and
transports the encrypted material to the DSKPP client. The DSKPP
client decrypts the keying material, and uses it to derive the
symmetric key, K_TOKEN. To preserve the property of not exposing K_TOKEN to any other
entity than the DSKPP server and the cryptographic module
itself, the method SHOULD be employed only when the device
contains facilities (e.g. a keypad) for direct entry of the
passphrase.This method MUST be identified with the following URN: urn:ietf:params:xml:schema:keyprov:dskpp#passphrase-wrapThe DSKPP server and client MUST support the following:The PBES2 password-based encryption scheme defined in
(and identified as
http://www.rsasecurity.com/rsalabs/pkcs/schemas/pkcs-5#pbes2
in )The PBKDF2 passphrase-based key derivation function also
defined in (and identified as
http://www.rsasecurity.com/rsalabs/pkcs/schemas/pkcs-5#pbkdf2
in )One of the following key wrapping mechanisms:KW-AES128 without padding; refer to
http://www.w3.org/2001/04/xmlenc#kw-aes128 in KW-AES128 without padding; refer to
http://www.w3.org/2001/04/xmlenc#kw-aes128 in AES-CBC-128; refer to The two-pass protocol flow consists of one exchange:Pass 1 = <KeyProvClientHello>, Pass 2 =
<KeyProvServerFinished>Although there is no exchange of the <ServerHello> message or
the <ClientNonce> message, the DSKPP client is still able to
specify algorithm preferences and supported key types in the
<KeyProvClientHello> message.The purpose and content of each message are described below. XML
format and examples are in and
.The trigger message is used in exactly the same way for the
two-pass variant as for the four-pass variant; refer to .When this message is sent:When a DSKPP client first connects to a DSKPP server, it is
required to send the <KeyProvClientHello> as its first
message. The client can also send <KeyProvClientHello> in
response to a <KeyProvTrigger> message.Purpose of this message:With this message, the DSKPP client specifies its algorithm
preferences and supported key types as well as which DSKPP
versions, protocol variants (in this case "two-pass"), key
package formats, and key protection methods that it supports.
Furthermore, the DSKPP client facilitates user authentication by
transmitting the authentication data (AD) that was provided by
the user before the first DSKPP message was sent.Application note:This message MUST send user authentication data (AD) to the
DSKPP server. If this message is preceded by trigger message
<KeyProvTrigger>, then the application will already have
AD available (see ).
However, if this message was not preceded by
<KeyProvTrigger>, then the application MUST retrieve the
user authentication code, possibly by prompting the user to
manually enter their authentication code, e.g., on a device with
only a numeric keypad.The application MUST also derive Authentication Data (AD)
from the authentication code, as described in , and save it for use in its next
message, <KeyProvClientNonce>.What is contained in this message:The Security Attribute List (SAL) included with
<KeyProvClientHello> contains the combinations of DSKPP
versions, variants, key package formats, key types, and
cryptographic algorithms that the DSKPP client supports in order
of the client's preference (favorite choice first). Authentication Data (AD) that was either
included with <KeyProvTrigger>, or generated as described
in the "Application Note" above.The
DSKPP client's random nonce (R_C), which is used to compute
provisioning key (K_PROV). By inserting R_C into the DSKPP
session, the DSKPP client is able to ensure the DSKPP server is
live before committing the key. If
<KeyProvClientHello> was preceded by a
<KeyProvTrigger>, then this message MUST also include the
DeviceID and/or KeyID that was provided with the trigger.
Otherwise, if a trigger message did not precede
<KeyProvClientHello>, then this message MAY include a
device ID that was pre-shared with the DSKPP server, and MAY
contain a key ID associated with a key previously provisioned by
the DSKPP provisioning server.The list of key protection methods
(KPML) that the DSKPP client supports. Each item in the list MAY
include an encryption key "payload" for the DSKPP server to use
to protect keying material that it sends back to the client. The
payload MUST be of type <ds:KeyInfoType> (). For each key protection method, the
allowable choices for <ds:KeyInfoType> are:Key TransportOnly those choices
of <ds:KeyInfoType> that identify a public key (i.e.,
<ds:KeyName>, <ds:KeyValue>,
<ds:X509Data>, or <ds:PGPData>). The
<ds:X509Certificate> option of the <ds:X509Data>
alternative is RECOMMENDED when the public key corresponding
to the private key on the cryptographic module has been
certified.Key WrapOnly those choices of
<ds:KeyInfoType> that identify a symmetric key (i.e.,
<ds:KeyName> and <ds:KeyValue>). The
<ds:KeyName> alternative is RECOMMENDED.Passphrase-Based Key WrapThe
<ds:KeyName> option MUST be used and the key name MUST
identify the passphrase that will be used by the server to
generate the key wrapping key. The identifier and passphrase
components of <ds:KeyName> MUST be set to the Client
ID and authentication code components of AD (same AD as
contained in this message).How the DSKPP server uses this message:The DSKPP server will look for an acceptable combination of
DSKPP version, variant (in this case, two-pass), key package
format, key type, and cryptographic algorithms. If the DSKPP
Client's SAL does not match the capabilities of the DSKPP
Server, or does not comply with key provisioning policy, then
the DSKPP Server will set the Status attribute to something
other than "Continue". Otherwise, Status will be set to
"Continue".The DSKPP server will validate the DeviceID and KeyID if
included in <KeyProvClientHello>. The DSKPP server MUST
NOT accept the DeviceID unless the server sent the DeviceID in a
preceding trigger message. Note that it is also legitimate for a
DSKPP client to initiate the DSKPP protocol run without having
received a <KeyProvTrigger> message from a server, but in
this case any provided DeviceID MUST NOT be accepted by the
DSKPP server unless the server has access to a unique key for
the identified device and that key will be used in the
protocol.The DSKPP server MUST use AD
to authenticate the user. If authentication fails, then the
DSKPP server MUST set the return code to a failure
status.If user authentication passes,
the DSKPP server generates a key K_PROV, which MUST consist of
two parts of equal length, where the first half constitutes
K_MAC and the second half constitutes K_TOKEN, i.e.,K_PROV = K_MAC || K_TOKENThe length of K_TOKEN (and hence also the length of K_MAC) is
determined by the type of K_TOKEN, which MUST be one of the key
types supported by the DSKPP client. Once K_PROV is computed, the DSKPP server
selects one of the key protection methods from the DSKPP
client's KPML, and uses that method and corresponding payload to
encrypt K_PROV. The result of the operation MUST be of type
<xenc:EncryptedKeyType> ().
For all three key protection methods, the Type attribute of the
<xenc:EncryptedKeyType> MUST be present and MUST identify
the type of the encrypted key. <xenc:CarriedKeyName> MAY
also be present, but MUST, when present, contain the same value
as the <KeyID> element of the
<KeyProvServerFinished> message. For each key protection
method, the following encryption method and key info values are
allowed:Key TransportOnly those
encryption methods that utilize a public key and are
supported by the DSKPP clientThis element MUST
identify the same public key as the key protection
"payload" that was received in
<KeyProvClientHello>Key WrapOnly those
encryption methods that utilize a symmetric key and are
supported by the DSKPP clientThis element MUST
identify the same symmetric key as the key protection
"payload" that was received in
<KeyProvClientHello>Passphrase-Based Key WrapOnly those
encryption methods that utilize a passphrase to derive
the key wrapping key and are supported by the DSKPP
clientThis element MUST
identify the same symmetric key as the key protection
"payload" that was received in
<KeyProvClientHello>After encrypting K_PROV, the DSKPP server generates a key
package that includes key usage attributes such as expiry date
and length. The key package MUST include the encrypted
provisioning key (K_PROV). The server stores the key package and
K_TOKEN with a user account on the cryptographic server.The server generates two MAC's, one for key confirmation and
another for server authentication) that the client will use to
avoid a false "Commit" message that would cause the
cryptographic module to end up in state in which the server does
not recognize the stored key.The method the DSKPP server MUST use to calculate the key
confirmation MAC:msg_hash = SHA-256(msg_1, ..., msg_n)dsLen = len(msg_hash)MAC = DSKPP-PRF (K_MAC, "MAC 1 computation" || msg_hash ||
ServerID, dsLen)whereThe MAC MUST be calculated using the
already established MAC algorithm and MUST be computed on
the (ASCII) string "MAC 1 computation", msg_hash, and
ServerID using the existing the MAC key K_MAC.The key, along with K_TOKEN, that is
derived from K_PROV which the DSKPP server MUST provide to
the cryptographic module.The message hash, defined in , of messages msg_1,
..., msg_n.The identifier that the DSKPP server
MUST include in the <KeyPackage> element of
<KeyProvServerFinished>.If DSKPP-PRF (defined in ) is
used as the MAC algorithm, then the input parameter s MUST
consist of the concatenation of the (ASCII) string "MAC 1
computation", msg_hash, and ServerID, and the parameter dsLen
MUST be set to the length of msg_hash.The method the DSKPP server MUST use to calculate the server
authentication MAC:The MAC MUST be computed on the (ASCII) string "MAC 2
computation", the server identifier ServerID, and R, using a
pre-existing MAC key K_MAC' (the MAC key that existed before
this protocol run). Note that the implementation may specify
K_MAC' to be the value of the K_TOKEN that is being replaced, or
a version of K_MAC from the previous protocol run.If DSKPP-PRF is used as the MAC algorithm, then the input
parameter s MUST consist of the concatenation of the (ASCII)
string "MAC 2 computation" ServerID, and R. The parameter dsLen
MUST be set to at least 16 (i.e. the length of the MAC MUST be
at least 16 octets):dsLen >= 16MAC = DSKPP-PRF (K_MAC', "MAC 2 computation" || ServerID ||
R, dsLen)The MAC algorithm MUST be the same as the algorithm used by
the DSKPP server to calculate the key confirmation MAC.When this message is sent:The DSKPP server will send this message after authenticating
the user and, if authentication passed, generating K_TOKEN and a
key package, and associating them with the user's account on the
cryptographic server.Purpose of this message:With this message the DSKPP server transports a key package
containing the encrypted provisioning key (K_PROV) and key usage
attributes.What is contained in this message:A status attribute equivalent to the server's return code to
<KeyProvClientHello>. If the server found an acceptable
set of attributes from the client's SAL, then it sets status to
Continue.The confirmation message MUST
include the Key Package (KP) that holds the DSKPP Server's ID,
key ID,key type, encrypted provisioning key (K_PROV), encryption
method, and additional configuration information. The default
symmetric key package format is based on the Portable Symmetric
Key Container (PSKC) defined in .
Alternative formats MAY include , PKCS#12 , or PKCS#5 XML . Finally,
this message MUST include a MAC that the DSKPP client will use
for key confirmation. It MUST also include a server
authentication MAC (AD). These MACs are calculated as described
in the previous section.How the DSKPP client uses this message:After receiving a <KeyProvServerFinished> message with
Status = "Success", the DSKPP client MUST verify both MACs (MAC
and AD). The DSKPP client MUST terminate the DSKPP protocol run
if either MAC does not verify, and MUST, in this case, also
delete any nonces, keys, and/or secrets associated with the
failed run of the protocol. If
<KeyProvServerFinished> has Status = "Success" and the
MACs were verified, then the DSKPP client MUST extract K_PROV
from the provided key package, and derive K_TOKEN. Finally, the
DSKPP client initializes the cryptographic module with K_TOKEN
and the corresponding key usage attributes. After this
operation, it MUST NOT be possible to overwrite the key unless
knowledge of an authorizing key is proven through a MAC on a
later <KeyProvServerFinished> message.DSKPP has been designed to be extensible. However, it is possible
that the use of extensions will harm interoperability; therefore, any
use of extensions SHOULD be carefully considered. For example, if a
particular implementation relies on the presence of a proprietary
extension, then it may not be able to interoperate with independent
implementations that have no knowledge of this extension.The ClientInfoType extension MAY contain any client-specific data
required of an application. This extension MAY be present in a
<KeyProvClientHello> or <KeyProvClientNonce> message.
DSKPP servers MUST support this extension. DSKPP servers MUST NOT
attempt to interpret the data it carries and, if received, MUST
include it unmodified in the current protocol run's next server
response. DSKPP servers need not retain the ClientInfoType data.The ServerInfoType extension MAY contain any server-specific data
required of an application, e.g., state information. This extension is
only valid in <KeyProvServerHello> messages for which the Status
attribute is set to "Continue". DSKPP clients MUST support this
extension. DSKPP clients MUST NOT attempt to interpret the data it
carries and, if received, MUST include it unmodified in the current
protocol run's next client request (i.e., the
<KeyProvClientNonce> message). DSKPP clients need not retain the
ServerInfoType data.DSKPP assumes a reliable transport.This section presents a binding of the previous messages to
HTTP/1.1 . Note that the HTTP client
will normally be different from the DSKPP client (i.e., the HTTP
client will "proxy" DSKPP messages from the DSKPP client to the DSKPP
server). Likewise, on the HTTP server side, the DSKPP server MAY
receive DSKPP message from a "front-end" HTTP server. The DSKPP server
will be identified by a specific URL, which may be pre-configured, or
provided to the client during initialization.The MIME-type for all DSKPP messages MUST beapplication/vnd.ietf.keyprov.dskpp+xmlIn order to avoid caching of responses carrying DSKPP messages by
proxies, the following holds:When using HTTP/1.1, requesters SHOULD:Include a Cache-Control header field set to "no-cache,
no-store".Include a Pragma header field set to "no-cache".When using HTTP/1.1, responders SHOULD:Include a Cache-Control header field set to "no-cache,
no-must-revalidate, private".Include a Pragma header field set to "no-cache".NOT include a Validator, such as a Last-Modified or ETag
header.To handle content negotiation, HTTP requests MAY include an HTTP
Accept header field. This header field SHOULD should be identified
using the MIME type specified in . The Accept header MAY
include additional content types defined by future versions of this
protocol.There are no other restrictions on HTTP headers, besides the
requirement to set the Content-Type header value to the MIME type
specified in .Persistent connections as defined in HTTP/1.1 are OPTIONAL. DSKPP
requests are mapped to HTTP requests with the POST method. DSKPP
responses are mapped to HTTP responses.For the 4-pass DSKPP, messages within the protocol run are bound
together. In particular, <KeyProvServerHello> is bound to the
preceding <KeyProvClientHello> by being transmitted in the
corresponding HTTP response. <KeyProvServerHello> MUST have a
SessionID attribute, and the SessionID attribute of the subsequent
<KeyProvClientNonce> message MUST be identical.
<KeyProvServerFinished> is then once again bound to the rest
through HTTP (and possibly through a SessionID).A DSKPP HTTP responder that refuses to perform a message exchange
with a DSKPP HTTP requester SHOULD return a 403 (Forbidden)
response. In this case, the content of the HTTP body is not
significant. In the case of an HTTP error while processing a DSKPP
request, the HTTP server MUST return a 500 (Internal Server Error)
response. This type of error SHOULD be returned for HTTP-related
errors detected before control is passed to the DSKPP processor, or
when the DSKPP processor reports an internal error (for example, the
DSKPP XML namespace is incorrect, or the DSKPP schema cannot be
located). If a request is received that is not a DSKPP client
message, the DSKPP responder MUST return a 400 (Bad request)
response.In these cases (i.e., when the HTTP response code is 4xx or 5xx),
the content of the HTTP body is not significant.Redirection status codes (3xx) apply as usual.Whenever the HTTP POST is successfully invoked, the DSKPP HTTP
responder MUST use the 200 status code and provide a suitable DSKPP
message (possibly with DSKPP error information included) in the HTTP
body.No support for HTTP/1.1 authentication is assumed.If a user requests key initialization in a browsing session, and
if that request has an appropriate Accept header (e.g., to a
specific DSKPP server URL), the DSKPP server MAY respond by sending
a DSKPP initialization message in an HTTP response with Content-Type
set according to
and response code set to 200 (OK). The initialization message MAY
carry data in its body, such as the URL for the DSKPP client to use
when contacting the DSKPP server. If the message does carry data,
the data MUST be a valid instance of a <KeyProvTrigger>
element.Note that if the user's request was directed to some other
resource, the DSKPP server MUST NOT respond by combining the DSKPP
content type with response code 200. In that case, the DSKPP server
SHOULD respond by sending a DSKPP initialization message in an HTTP
response with Content-Type set according to and response code set to
406 (Not Acceptable).Initialization from DSKPP server:HTTP/1.1 200 OKCache-Control: no-storeContent-Type:
application/vnd.ietf.keyprov.dskpp+xmlContent-Length: <some value>DSKPP initialization data in XML form...Initial request from DSKPP client:POST http://example.com/cgi-bin/DSKPP-server HTTP/1.1Cache-Control: no-cache, no-storePragma: no-cacheHost:
www.example.comContent-Type:
application/vnd.ietf.keyprov.dskpp+xmlContent-Length: <some value>DSKPP data in XML form (supported version,
supported algorithms...)Initial response from DSKPP server:HTTP/1.1 200 OKCache-Control: no-cache, no-must-revalidate,
privatePragma: no-cacheContent-Type:
application/vnd.ietf.keyprov.dskpp+xmlContent-Length: <some value>DSKPP data in XML form (server random nonce,
server public key, ...)Some DSKPP elements rely on the parties being able to compare
received values with stored values. Unless otherwise noted, all
elements that have the XML Schema "xs:string" type, or a type derived
from it, MUST be compared using an exact binary comparison. In
particular, DSKPP implementations MUST NOT depend on case-insensitive
string comparisons, normalization or trimming of white space, or
conversion of locale-specific formats such as numbers.Implementations that compare values that are represented using
different character encodings MUST use a comparison method that
returns the same result as converting both values to the Unicode
character encoding, Normalization Form C , and then performing an exact binary
comparison.No collation or sorting order for attributes or element values is
defined. Therefore, DSKPP implementations MUST NOT depend on specific
sorting orders for values.In order to assure that all implementations of DSKPP can
interoperate, the DSKPP server:MUST implement the four-pass variation of the protocol ()MUST implement the two-pass variation of the protocol ()MUST support user authentication ()MUST support the following key derivation functions:DSKPP-PRF-AES DSKPP-PRF realization ()DSKPP-PRF-SHA256 DSKPP-PRF realization ()MUST support the following encryption mechanisms for protection
of the client nonce in the four-pass protocol:Mechanism described in MUST support one of the following encryption algorithms for
symmetric key operations, e.g., key wrap:KW-AES128 without padding; refer to
http://www.w3.org/2001/04/xmlenc#kw-aes128 in KW-AES128 without padding; refer to
http://www.w3.org/2001/04/xmlenc#kw-aes128 in AES-CBC-128; refer to MUST support the following encryption algorithms for asymmetric
key operations, e.g., key transport:RSA Encryption Scheme MUST support the following integrity/KDF MAC functions:HMAC-SHA256 AES-CMAC-128 MUST support the PSKC key package ;
all three PSKC key protection methods (Key Transport, Key Wrap, and
Passphrase-Based Key Wrap) MUST be implementedMAY support the ASN.1 key package as defined in DSKPP clients MUST support either the two-pass or the four-pass
variant of the protocol. DSKPP clients MUST fulfill all requirements
listed in item (c) - (j).Of course, DSKPP is a security protocol, and one of its major
functions is to allow only authorized parties to successfully initialize
a cryptographic module with a new symmetric key. Therefore, a particular
implementation may be configured with any of a number of restrictions
concerning algorithms and trusted authorities that will prevent
universal interoperability.DSKPP is designed to protect generated keying material from
exposure. No other entities than the DSKPP server and the
cryptographic module will have access to a generated K_TOKEN if the
cryptographic algorithms used are of sufficient strength and, on the
DSKPP client side, generation and encryption of R_C and generation of
K_TOKEN take place as specified in the cryptographic module. This
applies even if malicious software is present in the DSKPP client.
However, as discussed in the following sub-sections, DSKPP does not
protect against certain other threats resulting from man-in-the-middle
attacks and other forms of attacks. DSKPP SHOULD, therefore, be run
over a transport providing confidentiality and integrity, such as HTTP
over Transport Layer Security (TLS) with a suitable ciphersuite, when
such threats are a concern. Note that TLS ciphersuites with anonymous
key exchanges are not suitable in those situations.An active attacker MAY attempt to modify, delete, insert, replay,
or reorder messages for a variety of purposes including service
denial and compromise of generated keying material.Modifications to a <KeyProvTrigger> message will either
cause denial-of-service (modifications of any of the identifiers or
the authentication code) or will cause the DSKPP client to contact
the wrong DSKPP server. The latter is in effect a man-in-the-middle
attack and is discussed further in .An attacker may modify a <KeyProvClientHello> message. This
means that the attacker could indicate a different key or device
than the one intended by the DSKPP client, and could also suggest
other cryptographic algorithms than the ones preferred by the DSKPP
client, e.g., cryptographically weaker ones. The attacker could also
suggest earlier versions of the DSKPP protocol, in case these
versions have been shown to have vulnerabilities. These
modifications could lead to an attacker succeeding in initializing
or modifying another cryptographic module than the one intended
(i.e., the server assigning the generated key to the wrong module),
or gaining access to a generated key through the use of weak
cryptographic algorithms or protocol versions. DSKPP implementations
MAY protect against the latter by having strict policies about what
versions and algorithms they support and accept. The former threat
(assignment of a generated key to the wrong module) is not possible
when the shared-key variant of DSKPP is employed (assuming existing
shared keys are unique per cryptographic module), but is possible in
the public-key variation. Therefore, DSKPP servers MUST NOT accept
unilaterally provided device identifiers in the public-key
variation. This is also indicated in the protocol description. In
the shared-key variation, however, an attacker may be able to
provide the wrong identifier (possibly also leading to the incorrect
user being associated with the generated key) if the attacker has
real-time access to the cryptographic module with the identified
key. The result of this attack could be that the generated key is
associated with the correct cryptographic module but the module is
associated with the incorrect user. See further for a discussion of this
threat and possible countermeasures.An attacker may also modify a <KeyProvServerHello> message.
This means that the attacker could indicate different key types,
algorithms, or protocol versions than the legitimate server would,
e.g., cryptographically weaker ones. The attacker may also provide a
different nonce than the one sent by the legitimate server. Clients
MAY protect against the former through strict adherence to policies
regarding permissible algorithms and protocol versions. The latter
(wrong nonce) will not constitute a security problem, as a generated
key will not match the key generated on the legitimate server. Also,
whenever the DSKPP run would result in the replacement of an
existing key, the <Mac> element protects against modifications
of R_S.Modifications of <KeyProvClientNonce> messages are also
possible. If an attacker modifies the SessionID attribute, then, in
effect, a switch to another session will occur at the server,
assuming the new SessionID is valid at that time on the server. It
still will not allow the attacker to learn a generated K_TOKEN since
R_C has been wrapped for the legitimate server. Modifications of the
<EncryptedNonce> element, e.g., replacing it with a value for
which the attacker knows an underlying R'C, will not result in the
client changing its pre-DSKPP state, since the server will be unable
to provide a valid MAC in its final message to the client. The
server MAY, however, end up storing K'TOKEN rather than K_TOKEN. If
the cryptographic module has been associated with a particular user,
then this could constitute a security problem. For a further
discussion about this threat, and a possible countermeasure, see
below. Note that use
of TLS does not protect against this attack if the attacker has
access to the DSKPP client (e.g., through malicious software,
"Trojans").Finally, attackers may also modify the
<KeyProvServerFinished> message. Replacing the <Mac>
element will only result in denial-of-service. Replacement of any
other element may cause the DSKPP client to associate, e.g., the
wrong service with the generated key. DSKPP SHOULD be run over a
transport providing confidentiality and integrity when this is a
concern.Message deletion will not cause any other harm than
denial-of-service, since a cryptographic module MUST NOT change its
state (i.e., "commit" to a generated key) until it receives the
final message from the DSKPP server and successfully has processed
that message, including validation of its MAC. A deleted
<KeyProvServerFinished> message will not cause the server to
end up in an inconsistent state vis-a-vis the cryptographic module
if the server implements the suggestions in .An active attacker may initiate a DSKPP run at any time, and
suggest any device identifier. DSKPP server implementations MAY
receive some protection against inadvertently initializing a key or
inadvertently replacing an existing key or assigning a key to a
cryptographic module by initializing the DSKPP run by use of the
<KeyProvTrigger>. The <AuthenticationData> element
allows the server to associate a DSKPP protocol run with, e.g., an
earlier user-authenticated session. The security of this method,
therefore, depends on the ability to protect the
<AuthenticationData> element in the DSKPP initialization
message. If an eavesdropper is able to capture this message, he may
race the legitimate user for a key initialization. DSKPP over a
transport providing confidentiality and integrity, coupled with the
recommendations in ,
is RECOMMENDED when this is a concern.Insertion of other messages into an existing protocol run is seen
as equivalent to modification of legitimately sent messages.During 4-pass DSKPP, attempts to replay a previously recorded
DSKPP message will be detected, as the use of nonces ensures that
both parties are live. For example, a DSKPP client knows that a
server it is communicating with is "live" since the server MUST
create a MAC on information sent by the client.The same is true for 2-pass DSKPP thanks to the requirement that
the client sends R in the <KeyProvClientHello> message and
that the server includes R in the MAC computation.An attacker may attempt to re-order 4-pass DSKPP messages but
this will be detected, as each message is of a unique type. Note:
Message re-ordering attacks cannot occur in 2-pass DSKPP since each
party sends at most one message each.In addition to other active attacks, an attacker posing as a
man-in-the-middle may be able to provide his own public key to the
DSKPP client. This threat and countermeasures to it are discussed in
. An attacker
posing as a man-in-the-middle may also be acting as a proxy and,
hence, may not interfere with DSKPP runs but still learn valuable
information; see .Passive attackers may eavesdrop on DSKPP runs to learn information
that later on may be used to impersonate users, mount active attacks,
etc.If DSKPP is not run over a transport providing confidentiality, a
passive attacker may learn:What cryptographic modules a particular user is in possession
ofThe identifiers of keys on those cryptographic modules and
other attributes pertaining to those keys, e.g., the lifetime of
the keysDSKPP versions and cryptographic algorithms supported by a
particular DSKPP client or serverAny value present in an <extension> that is part of
<KeyProvClientHello>Whenever the above is a concern, DSKPP SHOULD be run over a
transport providing confidentiality. If man-in-the-middle attacks for
the purposes described above are a concern, the transport SHOULD also
offer server-side authentication.An attacker with unlimited access to an initialized cryptographic
module may use the module as an "oracle" to pre-compute values that
later on may be used to impersonate the DSKPP server. contains a discussion of
this threat and steps RECOMMENDED to protect against it.Implementers SHOULD also be aware that cryptographic algorithms
become weaker with time. As new cryptographic techniques are developed
and computing performance improves, the work factor to break a
particular cryptographic algorithm will reduce. Therefore,
cryptographic algorithm implementations SHOULD be modular allowing new
algorithms to be readily inserted. That is, implementers SHOULD be
prepared to regularly update the algorithms in their
implementations.If keys generated in DSKPP will be associated with a particular
user at the DSKPP server (or a server trusted by, and communicating
with the DSKPP server), then in order to protect against threats where
an attacker replaces a client-provided encrypted R_C with his own R'C
(regardless of whether the public-key variation or the shared-secret
variation of DSKPP is employed to encrypt the client nonce), the
server SHOULD not commit to associate a generated K_TOKEN with the
given cryptographic module until the user simultaneously has proven
both possession of the device that hosts the cryptographic module
containing K_TOKEN and some out-of-band provided authenticating
information (e.g., an authentication code). For example, if the
cryptographic module is a one-time password token, the user could be
required to authenticate with both a one-time password generated by
the cryptographic module and an out-of-band provided authentication
code in order to have the server "commit" to the generated OTP value
for the given user. Preferably, the user SHOULD perform this operation
from another host than the one used to initialize keys on the
cryptographic module, in order to minimize the risk of malicious
software on the client interfering with the process.Note: This scenario, wherein the attacker replaces a
client-provided R_C with his own R'C, does not apply to 2-pass DSKPP
as the client does not provide any entropy to K_TOKEN. The attack as
such (and its countermeasures) still applies to 2-pass DSKPP, however,
as it essentially is a man-in-the-middle attack.Another threat arises when an attacker is able to trick a user to
authenticate to the attacker rather than to the legitimate service
before the DSKPP protocol run. If successful, the attacker will then
be able to impersonate the user towards the legitimate service, and
subsequently receive a valid DSKPP trigger. If the public-key variant
of DSKPP is used, this may result in the attacker being able to (after
a successful DSKPP protocol run) impersonate the user. Ordinary
precautions MUST, therefore, be in place to ensure that users
authenticate only to legitimate services.In 4-pass DSKPP, both the client and the server provide
randomizing material to K_TOKEN, in a manner that allows both
parties to verify that they did contribute to the resulting key. In
the 2-pass DSKPP version defined herein, only the server contributes
to the entropy of K_TOKEN. This means that a broken or compromised
(pseudo-)random number generator in the server may cause more damage
than it would in the 4-pass variant. Server implementations SHOULD
therefore take extreme care to ensure that this situation does not
occur.4-pass DSKPP servers provide key confirmation through the MAC on
R_C in the <KeyProvServerFinished> message. In the 2-pass
DSKPP variant described herein, key confirmation is provided by the
MAC including R, using K_MAC.DSKPP servers MUST authenticate themselves whenever a successful
DSKPP 2-pass protocol run would result in an existing K_TOKEN being
replaced by a K_TOKEN', or else a denial-of-service attack where an
unauthorized DSKPP server replaces a K_TOKEN with another key would
be possible. In 2-pass DSKPP, servers authenticate by including the
AuthenticationDataType extension containing a MAC as described in
for two-pass DSKPP.A DSKPP server MUST authenticate a client to ensure that K_TOKEN
is delivered to the intended device. The following measures SHOULD
be considered:When an Authentication Code is used for client
authentication, a password dictionary attack on the
authentication data is possible.The length of the Authentication Code when used over a
non-secure channel SHOULD be longer than what is used over a
secure channel. When a device, e.g., some mobile phones with
small screens, cannot handle a long Authentication Code in a
user-friendly manner, DSKPP SHOULD rely on a secure channel for
communication.In the case that a non-secure channel has to be used, the
Authentication Code SHOULD be sent to the server MAC'd as
specified in . The
Authentication Code and nonce value MUST be strong enough to
prevent offline brute-force recovery of the Authentication Code
from the HMAC data. Given that the nonce value is sent in
plaintext format over a non-secure transport, the cryptographic
strength of the Authentication Data depends more on the quality
of the Authentication Code.When the Authentication Code is sent from the DSKPP server to
the device in a DSKPP initialization trigger message, an
eavesdropper may be able to capture this message and race the
legitimate user for a key initialization. To prevent this, the
transport layer used to send the DSKPP trigger MUST provide
confidentiality and integrity, e.g. a secure browser
session.Three key protection methods are defined for the different usages
of 2-pass DSKPP, which MUST be supported by a key package format,
such as and . Therefore, key protection in the
two-pass DSKPP is dependent upon the security of the key package
format selected for a protocol run. Some considerations for the
Passphrase-Based Key Wrap method follow.The passphrase-based key wrap method SHOULD depend upon the
PBKDF2 function from to generate an
encryption key from a passphrase and salt string. It is important to
note that passphrase-based encryption is generally limited in the
security that it provides despite the use of salt and iteration
count in PBKDF2 to increase the complexity of attack.
Implementations SHOULD therefore take additional measures to
strengthen the security of the passphrase-based key wrap method. The
following measures SHOULD be considered where applicable:The passphrase is the same as the one-time password component
of the authentication code (see ) for a description of the AC
format). The passphrase SHOULD be selected well, and usage
guidelines such as the ones in
SHOULD be taken into account.A different passphrase SHOULD be used for every key
initialization wherever possible (the use of a global passphrase
for a batch of cryptographic modules SHOULD be avoided, for
example). One way to achieve this is to use randomly-generated
passphrases.The passphrase SHOULD be protected well if stored on the
server and/or on the cryptographic module and SHOULD be
delivered to the device's user using secure methods.User per-authentication SHOULD be implemented to ensure that
K_TOKEN is not delivered to a rogue recipient.The iteration count in PBKDF2 SHOULD be high to impose more
work for an attacker using brute-force methods (see for recommendations). However, it MUST
be noted that the higher the count, the more work is required on
the legitimate cryptographic module to decrypt the newly
delivered K_TOKEN. Servers MAY use relatively low iteration
counts to accommodate devices with limited processing power such
as some PDA and cell phones when other security measures are
implemented and the security of the passphrase-based key wrap
method is not weakened.Transport level security (e.g. TLS) SHOULD be used where
possible to protect a two-pass protocol run. Transport level
security provides a second layer of protection for the newly
generated K_TOKEN.The DSKPP protocol is mostly meant for machine-to-machine
communications; as such, most of its elements are tokens not meant for
direct human consumption. If these tokens are presented to the end user,
some localization may need to occur. DSKPP exchanges information using
XML. All XML processors are required to understand UTF-8 and UTF-16
encoding, and therefore all DSKPP clients and servers MUST understand
UTF-8 and UTF-16 encoded XML. Additionally, DSKPP servers and clients
MUST NOT encode XML with encodings other than UTF-8 or UTF-16.This document requires several IANA registrations, detailed
below.This section registers a new XML namespace,
"urn:ietf:params:xml:ns:keyprov:dskpp:1.0" per the guidelines in :urn:ietf:params:xml:ns:keyprov:dskpp:1.0IETF,
KEYPROV Working Group (keyprov@ietf.org), Andrea Doherty
(andrea.doherty@rsa.com)This section registers an XML schema as per the guidelines in .urn:ietf:params:xml:ns:keyprov:dskpp:1.0IETF,
KEYPROV Working Group (keyprov@ietf.org), Andrea Doherty
(andrea.doherty@rsa.com)The XML for this
schema can be found as the entirety of of this document.This section registers the "application/dskpp+xml" MIME type:ietf-types@iana.orgRegistration of MIME media type
application/dskpp+xmlapplicationdskpp+xml(none)charsetIndicates the character encoding of enclosed XML.
Default is UTF-8.Uses XML, which can employ
8-bit characters, depending on the character encoding used. See
, Section 3.2.This content type is
designed to carry protocol data related to key management.
Security mechanisms are built into the protocol to ensure that
various threats are dealt with.This content type
provides a basis for a protocol.RFC XXXX [NOTE TO
IANA/RFC-EDITOR: Please replace XXXX with the RFC number for this
specification.]Protocol for
key exchange.Magic Number(s): (none)File extension(s): .xmlsMacintosh File Type Code(s): (none)Andrea Doherty (andrea.doherty@rsa.com)LIMITED USEThe IETFThis media type is a
specialization of application/xml ,
and many of the considerations described there also apply to
application/dskpp+xml.This section registers status codes included in each DSKPP response
message. The status codes are defined in the schema in the
<StatusCode> type definition contained in the XML schema in
. The following summarizes the
registry:KEYPROV
DSKPP Registries, Status codes for DSKPPRFC XXXX
[NOTE TO IANA/RFC-EDITOR: Please replace XXXX with the RFC number
for this specification.]Following the policies outlined in , the IANA policy for assigning new values
for the status codes for DSKPP MUST be "Specification Required"
and their meanings MUST be documented in an RFC or in some other
permanent and readily available reference, in sufficient detail
that interoperability between independent implementations is
possible. No mechanism to mark entries as "deprecated" is
envisioned. It is possible to delete or update entries from the
registry.IETF,
KEYPROV working group (keyprov@ietf.org),Andrea Doherty (andrea.doherty@rsa.com)RSA and RSA Security are registered trademarks or trademarks of RSA
Security Inc. in the United States and/or other countries. The names of
other products and services mentioned may be the trademarks of their
respective owners.This work is based on information contained in , authored by Magnus Nystrom, with enhancements
borrowed from an individual Internet-Draft co-authored by Mingliang Pei
and Salah Machani (e.g., User Authentication, and support for multiple
key package formats).We would like to thank Philip Hoyer for his work in aligning DSKPP
and PSKC schemas.We would also like to thank Hannes Tschofenig and Phillip
Hallam-Baker for their draft reviews, feedback, and text
contributions.We would like to thank the following for review of previous DSKPP
document versions:Dr. Ulrike Meyer (Review June 2007)Niklas Neumann (Review June 2007)Shuh Chang (Review June 2007)Hannes Tschofenig (Review June 2007 and again in August
2007)Sean Turner (Reviews August 2007 and again in July 2008)John Linn (Review August 2007)Philip Hoyer (Review September 2007)Thomas Roessler (Review November 2007)Lakshminath Dondeti (Comments December 2007)Pasi Eronen (Comments December 2007)Phillip Hallam-Baker (Review and Edits November 2008 and again in
January 2009)We would also like to thank the following for their input to selected
design aspects of the DSKPP protocol:Anders Rundgren (Key Package Format and Client Authentication
Data)Thomas Roessler (HTTP Binding)Hannes Tschofenig (HTTP Binding)Phillip Hallam-Baker (Registry for Algorithms)Finally, we would like to thank Robert Griffin for opening
communication channels for us with the IEEE P1619.3 Key Management
Group, and facilitating our groups in staying informed of potential
areas (esp. key provisioning and global key identifiers of
collaboration) of collaboration.Secure Hash StandardNational Institute of Standards and
TechnologySpecification for the Advanced Encryption Standard
(AES)National Institute of Standards and
TechnologyRSA Cryptography StandardRSA LaboratoriesPassword-Based Cryptography StandardRSA LaboratoriesXML Schema for PKCS #5 Version 2.0RSA LaboratoriesPortable Symmetric Key ContainerHMAC: Keyed-Hashing for Message AuthenticationKey words for use in RFCs to Indicate Requirement
LevelsUTF-8, a transformation format of ISO10646Internet X.509 Public Key Infrastructure Certificate
Management Protocol (CMP)This document describes the Internet X.509 Public Key
Infrastructure (PKI) Certificate Management Protocol (CMP).
Protocol messages are defined for X.509v3 certificate creation and
management. CMP provides on-line interactions between PKI
components, including an exchange between a Certification
Authority (CA) and a client system. [STANDARDS TRACK]Certificate Management over CMS (CMC)This document defines the base syntax for CMC, a Certificate
Management protocol using the Cryptographic Message Syntax (CMS).
This protocol addresses two immediate needs within the Internet
Public Key Infrastructure (PKI) community:</t><t> 1.
The need for an interface to public key certification products and
services based on CMS and PKCS #10 (Public Key Cryptography
Standard), and</t><t> 2. The need for a PKI enrollment
protocol for encryption only keys due to algorithm or hardware
design.</t><t> CMC also requires the use of the
transport document and the requirements usage document along with
this document for a full definition. [STANDARDS TRACK]Unicode Normalization FormsUNICODE ConsortiumUNICODE ConsortiumXML Signature Syntax and ProcessingW3CXML Encryption Syntax and ProcessingW3CPKCS #11 Mechanisms for the Cryptographic Token Key
Initialization ProtocolRSA LaboratoriesFrequently Asked Questions About Today's CryptographyRSA LaboratoriesISO Information Processing Systems - Data Communication -
High-Level Data Link Control Procedure - Frame StructurePassword UsageNational Institute of Standards and
TechnologyRecommendations for Block Cipher Modes of Operation: The CMAC
Mode for AuthenticationInternational Organization for
StandardizationRecommendation for Key Management - Part I: General
(Revised)National Institute of Standards and
TechnologyCryptographic Token Interface StandardRSA LaboratoriesPersonal Information Exchange Syntax StandardUniform Resource Identifiers (URI): Generic SyntaxHypertext Transfer Protocol -- HTTP/1.1XML Media TypesInternet X.509 Public Key Infrastructure Certificate and
Certificate Revocation List (CRL) ProfileNISTMicrosoftTrinity College DublinEntrustVigil SecurityNISTIANA Considerations for RADIUSVeriSignThe IETF XML RegistryVeriSignCryptographic Token Key Initialization Protocol
(CT-KIP)RSA, The Security Division of EMCSymmetric Key Package Content TypeNamespaces in XMLW3CDSKPP is expected to be used to provision symmetric keys to
cryptographic modules in a number of different scenarios, each with its
own special requirements.The usual scenario is that a cryptographic module makes a request
for a symmetric key from a provisioning server that is located on the
local network or somewhere on the Internet. Depending upon the
deployment scenario, the provisioning server may generate a new key
on-the-fly or use a pre-generated key, e.g., one provided by a legacy
back-end issuance server. The provisioning server assigns a unique key
ID to the symmetric key and provisions it to the cryptographic
module.A cryptographic module makes multiple requests for symmetric keys
from the same provisioning server. The symmetric keys need not be of
the same type, i.e., the keys may be used with different symmetric key
cryptographic algorithms, including one-time password authentication
algorithms, and the AES encryption algorithm.In some deployment scenarios, a key issuer may rely on a third
party provisioning service. In this case, the issuer directs
provisioning requests from the cryptographic module to the
provisioning service. As such, it is the responsibility of the issuer
to authenticate the user through some out-of-band means before
granting him rights to acquire keys. Once the issuer has granted those
rights, the issuer provides an authentication code to the user and
makes it available to the provisioning service, so that the user can
prove that he is authorized to acquire keys.An issuer may provide a time-limited authentication code to a user
during registration, which the user will input into the cryptographic
module to authenticate themselves with the provisioning server. The
server will allow a key to be provisioned to the cryptographic module
hosted by the user's device when user authentication is required only
if the user inputs a valid authentication code within the fixed time
period established by the issuer.A cryptographic module requests renewal of the symmetric key
material attached to a key ID, as opposed to keeping the key value
constant and refreshing the metadata. Such a need may occur in the
case when a user wants to upgrade her device that houses the
cryptographic module or when a key has expired. When a user uses the
same cryptographic module to, for example, perform strong
authentication at multiple Web login sites, keeping the same key ID
removes the need for the user to register a new key ID at each
site.This scenario represents a special case of symmetric key renewal in
which a local administrator can authenticate the user procedurally
before initiating the provisioning process. It also allows for a
device issuer to pre-load a key onto a cryptographic module with a
restriction that the key is replaced with a new key prior to use of
the cryptographic module. Another variation of this scenario is the
organization who recycles devices. In this case, a key issuer would
provision a new symmetric key to a cryptographic module hosted on a
device that was previously owned by another user.Note that this usage scenario is essentially the same as the
previous scenario wherein the same key ID is used for renewal.A cryptographic module is loaded onto a smart card after the card
is issued to a user. The symmetric key for the cryptographic module
will then be provisioned using a secure channel mechanism present in
many smart card platforms. This allows a direct secure channel to be
established between the smart card chip and the provisioning server.
For example, the card commands (i.e., Application Protocol Data Units,
or APDUs) are encrypted with a pre-issued card manufacturer's key and
sent directly to the smart card chip, allowing secure post-issuance
in-the-field provisioning. This secure flow can pass Transport Layer
Security (TLS) and other transport security boundaries.Note that two pre-conditions for this usage scenario are for the
protocol to be tunneled and the provisioning server to know the
correct pre-established manufacturer's key.In this scenario, transport layer security does not provide
end-to-end protection of keying material transported from the
provisioning server to the cryptographic module. For example, TLS may
terminate at an application hosted on a PC rather than at the
cryptographic module (i.e., the endpoint) located on a data storage
device. Mutually authenticated key agreement provides end-to-end
protection, which TLS cannot provide.This appendix contains example messages that illustrate parameters,
encoding, and semantics in four-and two- pass DSKPP exchanges. The
examples are written using XML, and are syntactically correct. MAC and
cipher values are fictitious however.This message contains the nonce chosen by the cryptographic
module, R_C, encrypted by the specified encryption key and
encryption algorithm.The client indicates support for all the Key Transport, Key Wrap,
and Passphrase-Based Key Wrap key protection methods:In this example, the server responds to the previous request by
returning a key package in which the provisioning key was encrypted
using the Key Transport key protection method..The client sends a request that specifies a shared key to protect
the K_TOKEN, and the server responds using the Key Wrap key
protection method. Authentication data in this example is based on
an authentication code rather than a device certificate.In this example, the server responds to the previous request by
returning a key package in which the provisioning key was encrypted
using the Key Wrap key protection method.The client sends a request similar to that in with authentication data based on
an authentication code, and the server responds using the
Passphrase-Based Key Wrap method to encrypt the provisioning key
(note that the encryption is derived from the password component of
the authentication code). The authentication data is set in clear
text when it is sent over a secure transport channel such as
TLS.In this example, the server responds to the previous request by
returning a key package in which the provisioning key was encrypted
using the Passphrase-Based Key Wrap key protection method.A DSKPP client that needs to communicate with a connected
cryptographic module to perform a DSKPP exchange MAY use PKCS #11 as a programming interface.When performing 4-pass DSKPP with a cryptographic module using the
PKCS #11 programming interface, the procedure described in , Appendix B, is RECOMMENDED.A suggested procedure to perform 2-pass DSKPP with a cryptographic
module through the PKCS #11 interface using the mechanisms defined in
is as follows:On the client side, The client selects a suitable slot and token (e.g., through
use of the <DeviceIdentifier> or the
<PlatformInfo> element of the DSKPP trigger
message).A nonce R is generated, e.g. by calling C_SeedRandom and
C_GenerateRandom.The client sends its first message to the server, including
the nonce R.On the server side, A generic key K_PROV = K_TOKEN | K_MAC (where '|' denotes
concatenation) is generated, e.g. by calling C_GenerateKey
(using key type CKK_GENERIC_SECRET). The template for K_PROV
MUST allow it to be exported (but only in wrapped form, i.e.
CKA_SENSITIVE MUST be set to CK_TRUE and CKA_EXTRACTABLE MUST
also be set to CK_TRUE), and also to be used for further key
derivation. From K, a token key K_TOKEN of suitable type is
derived by calling C_DeriveKey using the PKCS #11 mechanism
CKM_EXTRACT_KEY_FROM_KEY and setting the CK_EXTRACT_PARAMS to
the first bit of the generic secret key (i.e. set to 0).
Likewise, a MAC key K_MAC is derived from K_PROV by calling
C_DeriveKey using the CKM_EXTRACT_KEY_FROM_KEY mechanism, this
time setting CK_EXTRACT_PARAMS to the length of K_PROV (in
bits) divided by two.The server wraps K_PROV with either the public key of the
DSKPP client or device, the pre-shared secret key, or the
derived shared secret key by using C_WrapKey. If use of the
DSKPP key wrap algorithm has been negotiated then the
CKM_KIP_WRAP mechanism MUST be used to wrap K. When calling
C_WrapKey, the hKey handle in the CK_KIP_PARAMS structure MUST
be set to NULL_PTR. The pSeed parameter in the CK_KIP_PARAMS
structure MUST point to the nonce R provided by the DSKPP
client, and the ulSeedLen parameter MUST indicate the length
of R. The hWrappingKey parameter in the call to C_WrapKey MUST
be set to refer to the key wrapping key.Next, the server needs to calculate a MAC using K_MAC. If
use of the DSKPP MAC algorithm has been negotiated, then the
MAC is calculated by calling C_SignInit with the CKM_KIP_MAC
mechanism followed by a call to C_Sign. In the call to
C_SignInit, K_MAC MUST be the signature key, the hKey
parameter in the CK_KIP_PARAMS structure MUST be set to
NULL_PTR, the pSeed parameter of the CT_KIP_PARAMS structure
MUST be set to NULL_PTR, and the ulSeedLen parameter MUST be
set to zero. In the call to C_Sign, the pData parameter MUST
be set to the concatenation of the string ServerID and the
nonce R, and the ulDataLen parameter MUST be set to the length
of the concatenated string. The desired length of the MAC MUST
be specified through the pulSignatureLen parameter and MUST be
set to the length of R.If the server also needs to authenticate its message (due
to an existing K_TOKEN being replaced), the server MUST
calculate a second MAC. Again, if use of the DSKPP MAC
algorithm has been negotiated, then the MAC is calculated by
calling C_SignInit with the CKM_KIP_MAC mechanism followed by
a call to C_Sign. In this call to C_SignInit, the K_MAC'
existing before this DSKPP protocol run MUST be the signature
key (the implementation may specify K_MAC' to be the value of
the K_TOKEN that is being replaced, or a version of K_MAC from
the previous protocol run), the hKey parameter in the
CK_KIP_PARAMS structure MUST be set to NULL, the pSeed
parameter of the CT_KIP_PARAMS structure MUST be set to
NULL_PTR, and the ulSeedLen parameter MUST be set to zero. In
the call to C_Sign, the pData parameter MUST be set to the
concatenation of the string ServerID and the nonce R, and the
ulDataLen parameter MUST be set to the length of concatenated
string. The desired length of the MAC MUST be specified
through the pulSignatureLen parameter and MUST be set to the
length of R.The server sends its message to the client, including the
wrapped key K_TOKEN, the MAC and possibly also the
authenticating MAC.On the client side, The client calls C_UnwrapKey to receive a handle to K.
After this, the client calls C_DeriveKey twice: Once to derive
K_TOKEN and once to derive K_MAC. The client MUST use the same
mechanism (CKM_EXTRACT_KEY_FROM_KEY) and the same mechanism
parameters as used by the server above. When calling
C_UnwrapKey and C_DeriveKey, the pTemplate parameter MUST be
used to set additional key attributes in accordance with local
policy and as negotiated and expressed in the protocol. In
particular, the value of the <KeyID> element in the
server's response message MAY be used as CKA_ID for K_TOKEN.
The key K_PROV MUST be destroyed after deriving K_TOKEN and
K_MAC.The MAC is verified in a reciprocal fashion as it was
generated by the server. If use of the CKM_KIP_MAC mechanism
has been negotiated, then in the call to C_VerifyInit, the
hKey parameter in the CK_KIP_PARAMS structure MUST be set to
NULL_PTR, the pSeed parameter MUST be set to NULL_PTR, and
ulSeedLen MUST be set to 0. The hKey parameter of C_VerifyInit
MUST refer to K_MAC. In the call to C_Verify, pData MUST be
set to the concatenation of the string ServerID and the nonce
R, and the ulDataLen parameter MUST be set to the length of
the concatenated string, pSignature to the MAC value received
from the server, and ulSignatureLen to the length of the MAC.
If the MAC does not verify the protocol session ends with a
failure. The token MUST be constructed to not "commit" to the
new K_TOKEN or the new K_MAC unless the MAC verifies.If an authenticating MAC was received (REQUIRED if the new
K_TOKEN will replace an existing key on the token), then it is
verified in a similar vein but using the K_MAC' associated
with this server and existing before the protocol run (the
implementation may specify K_MAC' to be the value of the
K_TOKEN that is being replaced, or a version of K_MAC from the
previous protocol run). Again, if the MAC does not verify the
protocol session ends with a failure, and the token MUST be
constructed no to "commit" to the new K_TOKEN or the new K_MAC
unless the MAC verifies.This example appendix defines DSKPP-PRF in terms of AES and HMAC .For cryptographic modules supporting this realization of
DSKPP-PRF, the following URL MAY be used to identify this algorithm
in DSKPP:http://www.ietf.org/keyprov/dskpp#dskpp-prf-aes-128When this URL is used to identify the encryption algorithm, the
method for encryption of R_C values described in MUST be used.DSKPP-PRF-AES (k, s, dsLen)Input:Encryption key to useOctet string consisting of randomizing material.
The length of the string s is sLen.Desired length of the outputOutput:A pseudorandom string, dsLen-octets longSteps:Let bLen be the output block size of AES in octets:bLen = (AES output block length in
octets)(normally, bLen = 16)If dsLen > (2**32 - 1) * bLen, output "derived data too
long" and stopLet n be the number of bLen-octet blocks in the output data,
rounding up, and let j be the number of octets in the last
block:n = CEILING( dsLen / bLen)j = dsLen - (n - 1) * bLenFor each block of the pseudorandom string DS, apply the
function F defined below to the key k, the string s and the
block index to compute the block:B1 = F
(k, s, 1) ,B2 = F (k, s, 2) ,...Bn = F (k, s,
n)The function F is defined in terms of the CMAC construction from
, using AES as the block
cipher:F (k, s, i) = CMAC-AES (k, INT (i)
|| s)where INT (i) is a four-octet encoding
of the integer i, most significant octet first, and the output
length of CMAC is set to bLen.Concatenate
the blocks and extract the first dsLen octets to product the desired
data string DS:DS = B1 || B2 || ... ||
Bn<0..j-1>Output the derived data
DS.If we assume that dsLen = 16, then:n = 16 / 16 = 1j = 16 - (1 - 1) * 16 = 16DS = B1 = F (k, s, 1) = CMAC-AES (k, INT (1) || s)For cryptographic modules supporting this realization of
DSKPP-PRF, the following URL MAY be used to identify this algorithm
in DSKPP:http://www.ietf.org/keyprov/dskpp#dskpp-prf-sha256When this URL is used to identify the encryption algorithm to
use, the method for encryption of R_C values described in MUST be used.DSKPP-PRF-SHA256 (k, s, dsLen)Input:Encryption key to useOctet string consisting of randomizing material.
The length of the string s is sLen.Desired length of the outputOutput:A pseudorandom string, dsLen-octets longSteps:Let bLen be the output size of SHA-256 in octets of (no truncation is done on the HMAC
output):bLen = 32(normally, bLen = 16)If dsLen > (2**32 - 1) * bLen, output "derived data too
long" and stopLet n be the number of bLen-octet blocks in the output data,
rounding up, and let j be the number of octets in the last
block:n = CEILING( dsLen / bLen)j = dsLen - (n - 1) * bLenFor each block of the pseudorandom string DS, apply the
function F defined below to the key k, the string s and the
block index to compute the block:B1 = F
(k, s, 1),B2 = F (k, s, 2),...Bn = F (k, s,
n)The function F is defined in terms of the HMAC construction from
, using SHA-256 as the digest
algorithm:F (k, s, i) = HMAC-SHA256 (k, INT
(i) || s)where INT (i) is a four-octet
encoding of the integer i, most significant octet first, and the
output length of HMAC is set to bLen.Concatenate the blocks and extract the first dsLen
octets to product the desired data string DS:DS = B1 || B2 || ... || Bn<0..j-1>Output the derived data DS.If we assume that sLen = 256 (two 128-octet long values) and
dsLen = 16, then:n = CEILING( 16 / 32 ) = 1j = 16 - (1 - 1) * 32 = 16B1 = F (k, s, 1) = HMAC-SHA256 (k, INT (1) || s)DS = B1<0 ... 15>That is, the result will be the first 16 octets of the HMAC
output.