path: root/doc/draft-ietf-secsh-agent-01.txt

Network Working Group                                        Tatu Ylonen
INTERNET-DRAFT                                             Timo J. Rinne
draft-ietf-secsh-agent-01.txt                              Sami Lehtinen
Expires in six months                        SSH Communications Security
                                                       20 November, 2002



               Secure Shell Authentication Agent Protocol

Status of This Memo

This document is an Internet-Draft and is in full conformance
with all provisions of Section 10 of RFC2026.

Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups.  Note that
other groups may also distribute working documents as
Internet-Drafts.

Internet-Drafts are draft documents valid for a maximum of six
months and may be updated, replaced, or obsoleted by other
documents at any time.  It is inappropriate to use Internet-
Drafts as reference material or to cite them other than as
"work in progress."

The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt

The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.

Abstract

This document describes the Secure Shell authentication agent protocol
(i.e., the protocol used between a client requesting authentication and
the authentication agent).  This protocol usually runs in a machine-spe-
cific local channel or over a forwarded authentication channel.  It is
assumed that the channel is trusted, so no protection for the communica-
tions channel is provided by this protocol.

















Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 1]

INTERNET-DRAFT                                        20 November, 2002
 
Table of Contents

1.  Authentication Agent Protocol   . . . . . . . . . . . . . . . . .  2
  1.1.  Packet Format   . . . . . . . . . . . . . . . . . . . . . . .  2
  1.2.  Forwarding Notices  . . . . . . . . . . . . . . . . . . . . .  3
  1.3.  Requesting Version Number   . . . . . . . . . . . . . . . . .  3
  1.4.  Adding Keys to the Agent  . . . . . . . . . . . . . . . . . .  4
  1.5.  Deleting Keys from the Agent  . . . . . . . . . . . . . . . .  5
  1.6.  Deleting specific key from the Agent  . . . . . . . . . . . .  5
  1.7.  Listing the Keys that the Agent Can Use   . . . . . . . . . .  6
2.  Performing Private Key Operations   . . . . . . . . . . . . . . .  6
  2.1.  Signing   . . . . . . . . . . . . . . . . . . . . . . . . . .  7
  2.2.  Decrypting  . . . . . . . . . . . . . . . . . . . . . . . . .  7
  2.3.  Secure Shell Challenge-Response Authentication  . . . . . . .  7
3.  Administrative Messages   . . . . . . . . . . . . . . . . . . . .  7
  3.1.  Locking and unlocking the agent   . . . . . . . . . . . . . .  8
  3.2.  Miscellaneous Agent Commands  . . . . . . . . . . . . . . . .  8
4.  Agent Forwarding With Secure Shell  . . . . . . . . . . . . . . .  9
  4.1.  Requesting Agent Forwarding   . . . . . . . . . . . . . . . .  9
  4.2.  Agent Forwarding Channels   . . . . . . . . . . . . . . . . .  9
5.  Security Considerations   . . . . . . . . . . . . . . . . . . . .  9
6.  Intellectual Property   . . . . . . . . . . . . . . . . . . . . . 10
7.  Additional Information  . . . . . . . . . . . . . . . . . . . . . 10
8.  References  . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
9.  Address of Authors  . . . . . . . . . . . . . . . . . . . . . . . 10



1.  Authentication Agent Protocol

The authentication agent is a piece of software that runs in a user's
local workstation, laptop, or other trusted device.  It is used to
implement single sign-on.  It holds the user's private keys in its own
storage, and can perform requested operations using the private key.  It
allows the keys to be kept on a smartcard or other special hardware that
can perform cryptographic operations.

The authentication agent protocol is used to communicate between the
authentication agent and clients wanting to authenticate something or
wanting to perform private key operations.

The actual communication between the client and the agent happens using
a machine-dependent trusted communications channel.  This channel would
typically be a local socket, named pipe, or some kind of secure
messaging system that works inside the local machine.

The protocol works by the client sending requests to the agent, and the
agent responding to these requests.

1.1.  Packet Format

All messages passed to/from the authentication agent have the following
format:


Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 2]

INTERNET-DRAFT                                        20 November, 2002
 
  uint32           length
  byte             type
  data[length -1]  data payload

The following packet types are currently defined:

   /* Messages sent by the client. */
   #define SSH_AGENT_REQUEST_VERSION                        1
   #define SSH_AGENT_ADD_KEY                              202
   #define SSH_AGENT_DELETE_ALL_KEYS                      203
   #define SSH_AGENT_LIST_KEYS                            204
   #define SSH_AGENT_PRIVATE_KEY_OP                       205
   #define SSH_AGENT_FORWARDING_NOTICE                    206
   #define SSH_AGENT_DELETE_KEY                           207
   #define SSH_AGENT_LOCK                                 208
   #define SSH_AGENT_UNLOCK                               209
   #define SSH_AGENT_PING                                 212
   #define SSH_AGENT_RANDOM                               213

   /* Messages sent by the agent. */
   #define SSH_AGENT_SUCCESS                              101
   #define SSH_AGENT_FAILURE                              102
   #define SSH_AGENT_VERSION_RESPONSE                     103
   #define SSH_AGENT_KEY_LIST                             104
   #define SSH_AGENT_OPERATION_COMPLETE                   105
   #define SSH_AGENT_RANDOM_DATA                          106
   #define SSH_AGENT_ALIVE                                150

1.2.  Forwarding Notices

If the agent connection is forwarded through intermediate hosts (using
the SSH Connection Protocol agent forwarding feature (described in
Section ``Agent Forwarding With Secure Shell'' of this document), or
some other means), each intermediate node (Secure Shell client) should
insert the following message into the agent channel before forwarding
any other messages.  The real agent will then receive these messages in
sequence the nearest node first, and can determine whether the
connection is from a local machine and if not, can log the path where
the connection came from.  These messages must be wrapped in the
appropriate header.

  byte      SSH_AGENT_FORWARDING_NOTICE
  string    remote host name  (as typed by the user, preferably)
  string    remote host ip
  uint32    remote host port

1.3.  Requesting Version Number

When the client opens a connection, it must send the following message
to the server.  This must be the first message sent.  The real agent
will receive this after zero or more forwarding notice messages.
  byte      SSH_AGENT_REQUEST_VERSION
  string    version string of the application sending the request


Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 3]

INTERNET-DRAFT                                        20 November, 2002
 
            (optional)

If the agent follows this protocol, it will respond with

  byte      SSH_AGENT_VERSION_RESPONSE
  uint32    version number, 2 for this protocol

If the version number request is ever sent to the Secure Shell 1.x
agent, it will interpret it as a request to list identities.  It will
then respond with a message whose first byte is 2.  This can be used to
determine the version of the agent if compatibility with Secure Shell
1.x is desired.

If the version string query arrives without trailing string identifying
the client software version, it can be translated list identities
request sent by Secure Shell 1.x and handled accordingly.  If agent
software does not support the agent protocol of Secure Shell 1.x, it MAY
also interpret this query as valid SSH_AGENT_REQUEST_VERSION packet.

1.4.  Adding Keys to the Agent

The client can add a new private key to the agent with the following
message.

  byte      SSH_AGENT_ADD_KEY
  string    private key blob with empty passphrase
  string    public key and/or certificates for it
  string    description of the key
  ... 0, 1 or several constraints follow

All constraints are pairs of following format:

  byte      SSH_AGENT_CONSTRAINT_*
  variable  argument for the constraint

The type of the argument is dependent on the constraint type.  Following
constraint types are currently defined:

   /* Constraints 50-99 have a uint32 argument */

   /* Argument is uint32 defining key expiration time-out in
      seconds. After this timeout expires, the key can't be used.
      0 == no timeout */
   #define SSH_AGENT_CONSTRAINT_TIMEOUT                 50

   /* Argument is uint32 defining the number of operations that can
      be performed with this key.  0xffffffff == no limit */
   #define SSH_AGENT_CONSTRAINT_USE_LIMIT               51

   /* Argument is uint32 defining the number of forwarding steps that
      this key can be forwarded.  0xffffffff == no limit */
   #define SSH_AGENT_CONSTRAINT_FORWARDING_STEPS        52



Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 4]

INTERNET-DRAFT                                        20 November, 2002
 
   /* Constraints 100-149 have a string argument */

   /* Argument is string defining the allowed forwarding steps for
      this key. XXX define this. */
   #define SSH_AGENT_CONSTRAINT_FORWARDING_PATH        100

   /* Constraints 150-199 have a boolean argument */

   /* Argument is a boolean telling whether the key can be used
      in Secure Shell 1.x compatibility operations. */

   #define SSH_AGENT_CONSTRAINT_SSH1_COMPAT            150

   /* Argument is a boolean telling whether operations performed
      with this key should  be confirmed interactively by the user
      or not. */
   #define SSH_AGENT_CONSTRAINT_NEED_USER_VERIFICATION 151

Message can contain zero, one or multiple constraints.

If the operation is successful, the agent will respond with the
following message.

   byte      SSH_AGENT_SUCCESS

If the operation fails for some reason, the following message will be
returned instead.

   byte      SSH_AGENT_FAILURE
   uint32    error code

The error code is one of the following:

   #define SSH_AGENT_ERROR_TIMEOUT               1
   #define SSH_AGENT_ERROR_KEY_NOT_FOUND         2
   #define SSH_AGENT_ERROR_DECRYPT_FAILED        3
   #define SSH_AGENT_ERROR_SIZE_ERROR            4
   #define SSH_AGENT_ERROR_KEY_NOT_SUITABLE      5
   #define SSH_AGENT_ERROR_DENIED                6
   #define SSH_AGENT_ERROR_FAILURE               7
   #define SSH_AGENT_ERROR_UNSUPPORTED_OP        8

1.5.  Deleting Keys from the Agent

All keys that are in possession of the agent can be deleted with the
following message.  (The client is allowed to ignore this for some keys
if desired.)

   byte      SSH_AGENT_DELETE_ALL_KEYS

The agent responds with either SSH_AGENT_SUCCESS or SSH_AGENT_FAILURE.




Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 5]

INTERNET-DRAFT                                        20 November, 2002
 
1.6.  Deleting specific key from the Agent

The client can delete a specific key with given public key with
following message.

   byte      SSH_AGENT_DELETE_KEY
   string    public key and/or certificates for it
   string    description of the key

The agent responds with either SSH_AGENT_SUCCESS or SSH_AGENT_FAILURE.

1.7.  Listing the Keys that the Agent Can Use

The following message requests a list of all keys that the agent can
use.

   byte      SSH_AGENT_LIST_KEYS

The agent will respond with the following message.

   byte      SSH_AGENT_KEY_LIST
   uint32    number_of_keys
   repeats number_of_keys times:
    string    public key blob or certificates
    string    description

2.  Performing Private Key Operations

The real purpose of the agent is to perform private key operations.
Such operations are performed with the following message.

   byte      SSH_AGENT_PRIVATE_KEY_OP
   string    operation name
   string    key or certificates, as returned in SSH_AGENT_KEY_LIST
   ... operation-specific data follows

The operation to be performed is identified by a name (string).  Custom
operations can be added by suffixing the operation name by the fully
qualified domain name of the person/organization adding the new
operation.

When the operation is complete, the agent will respond with either
SSH_AGENT_FAILURE or with the following message if the operation is
successful:

   byte      SSH_AGENT_OPERATION_COMPLETE
   string    resulting data

If an operation is attempted that is not supported by the agent, the
agent will respond with SSH_AGENT_FAILURE with error code set to
SSH_AGENT_ERROR_UNSUPPORTED_OP.

The standard operations are defined below.


Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 6]

INTERNET-DRAFT                                        20 November, 2002
 
2.1.  Signing

The agent can be used to create a digital signature using a key held by
the agent.  The operation name is "sign", and data in is a hash
(suitable for the key) that is to be signed.  This normally performs the
raw private key operation, without hashing data first.  The resulting
data will be a binary representation of the output of the private key
operation.  The exact details of the operations to be performed depend
on the key being used.

The operation-specific data has the following format:

   string    data to be signed

Alternatively, it is possible to give the actual data to be signed to
the agent.  This is done using the operation "hash-and-sign".  This is
otherwise equal, but performs key-dependent hashing before signing.

If the requested operation is not legal for the key, SSH_AGENT_FAILURE
will be returned with error code set to
SSH_AGENT_ERROR_KEY_NOT_SUITABLE.

2.2.  Decrypting

The agent can be used to decrypt a public key encrypted message with the
operation "decrypt".  This takes in raw public-key encrypted data, and
returns the resulting decrypted data.

This may also fail.  If the requested operation is not legal for the
key, error code is set to SSH_AGENT_ERROR_KEY_NOT_SUITABLE.

The operation-specific data has the following format:

   string    data to be decrypted

2.3.  Secure Shell Challenge-Response Authentication

Performs Secure Shell challenge-response authentication.  This operation
has the name "ssh1-challenge-response".

This operation works by first decrypting the challenge, then computing
MD5 of the concatenation of the decrypted challenge and the session id
(in this order), and returns the resulting 16 byte hash.  The operation-
specific data is in the following format:

   string    challenge encrypted using the public key
   string    session id

Normally, the length of the challenge before encryption will be 32 bytes
and the length of the session id 16 bytes.  The length of the encrypted
challenge depends on the key and algorithm used.




Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 7]

INTERNET-DRAFT                                        20 November, 2002
 
3.  Administrative Messages

There are also a number of messages that are only used to administer the
agent.  These might e.g. be used by a user interface for the agent.  The
agent should only allow these messages from local connection (i.e., if
no forwarding notice messages were received before the version number
request).

3.1.  Locking and unlocking the agent

The agent can be temporarily locked by message:

   byte      SSH_AGENT_LOCK
   string    locking password

The agent responds with either SSH_AGENT_SUCCESS or SSH_AGENT_FAILURE.
Particularily SSH_AGENT_FAILURE is sent, if agent is already locked.
After this message, agent responds to all commands with
SSH_AGENT_FAILURE until it receives a following command.

   byte      SSH_AGENT_UNLOCK
   string    locking password

The agent responds with either SSH_AGENT_SUCCESS or SSH_AGENT_FAILURE.
Particularily SSH_AGENT_FAILURE is sent, if agent is not locked or if
the submitted password does not match with one given with SSH_AGENT_LOCK
message.

3.2.  Miscellaneous Agent Commands

   byte      SSH_AGENT_PING
   ... arbitrary padding data

Any agent or client receiving this message, should respond with

   byte      SSH_AGENT_ALIVE
   ... padding data from the SSH_AGENT_PING request

where the padding data is identical to the data sent with
SSH_AGENT_PING.

   byte      SSH_AGENT_RANDOM
   uint32    the length of the requested random buffer

Client can request random data from the agent by this message.  Agent
responds either with SSH_AGENT_RANDOM_DATA or SSH_AGENT_FAILURE message.

   byte      SSH_AGENT_RANDOM_DATA
   string    random data

This message is a successful response to SSH_AGENT_RANDOM message.
Message contains the random string of requested length.



Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 8]

INTERNET-DRAFT                                        20 November, 2002
 
4.  Agent Forwarding With Secure Shell

The agent connection is typically forwarded over a Secure Shell
connection. This requires small additions to the SSH Connection Protocol
[SSH-CONN].

4.1.  Requesting Agent Forwarding

Agent forwarding may be requested for a session by sending

  byte      SSH_MSG_CHANNEL_REQUEST
  uint32    recipient channel
  string    "auth-agent-req"
  boolean   want reply

This will, on success, create an agent listener to the remote end.

4.2.  Agent Forwarding Channels

When a connection comes to the forwarded agent listener, a channel is
opened to forward the connection to the other side.

  byte      SSH_MSG_CHANNEL_OPEN
  string    "auth-agent"
  uint32    sender channel
  uint32    initial window size
  uint32    maximum packet size

Implementations MUST reject these messages unless they have previously
requested agent forwarding.

Forwarded agent channels are independent of any sessions, and closing a
session channel does not in any way imply that forwarded connections
should be closed.

5.  Security Considerations

The authentication agent is used to control security-sensitive
operations, and is used to implement single sign-on.

Anyone with access to the authentication agent can perform private key
operations with the agent.  This is a power equivalent to possession of
the private key as long as the connection to the key is maintained.  It
is not possible to retrieve the key from the agent.

It is recommended that agent implementations allow and perform some form
of logging and access control.  This access control may utilize
information about the path through which the connection was received (as
collected with SSH_AGENT_FORWARDING_NOTICE messages; however, the path
is reliable only up to and including the first unreliable machine.).
Implementations should also allow restricting the operations that can be
performed with keys - e.g., limiting them to challenge-response only.



Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                    [page 9]

INTERNET-DRAFT                                        20 November, 2002
 
One should note that a local superuser will be able to obtain access to
agents running on the local machine.  This cannot be prevented; in most
operating systems, a user with sufficient privileges will be able to
read the keys from the physical memory.

The authentication agent should not be run or forwarded to machine whose
integrity is not trusted, as security on such machines might be
compromised and might allow an attacker to obtain unauthorized access to
the agent.

6.  Intellectual Property

The IETF takes no position regarding the validity or scope of any
intellectual property or other rights that might be claimed to pertain
to the implementation or use of the technology described in this
document or the extent to which any license under such rights might or
might not be available; neither does it represent that it has made any
effort to identify any such rights.  Information on the IETF's
procedures with respect to rights in standards-track and standards-
related documentation can be found in BCP-11.  Copies of claims of
rights made available for publication and any assurances of licenses to
be made available, or the result of an attempt made to obtain a general
license or permission for the use of such proprietary rights by
implementers or users of this specification can be obtained from the
IETF Secretariat.

The IETF has been notified of intellectual property rights claimed in
regard to some or all of the specification contained in this document.
For more information consult the online list of claimed rights.

7.  Additional Information

The current document editor is: Sami Lehtinen <sjl@ssh.com>.  Comments
on this Internet-Draft should be sent to the IETF SECSH working group,
details at: http://ietf.org/html.charters/secsh-charter.html

8.  References

[SECSH-CONNECT] Ylonen, T., et al: "Secure Shell Connection Protocol",
Internet-Draft, draft-ietf-secsh-connect-16.txt

9.  Address of Authors

   Tatu Ylonen
    SSH Communications Security Corp
    Fredrikinkatu 42
    FIN-00100 HELSINKI
    Finland
    E-mail: ylo@ssh.com

   Timo J. Rinne
    SSH Communications Security Corp
    Fredrikinkatu 42


Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                   [page 10]

INTERNET-DRAFT                                        20 November, 2002
 
    FIN-00100 HELSINKI
    Finland
    E-mail: tri@ssh.com

   Sami Lehtinen
    SSH Communications Security Corp
    Fredrikinkatu 42
    FIN-00100 HELSINKI
    Finland
    E-mail: sjl@ssh.com












































Tatu Ylonen, Timo J. Rinne and Sami Lehtinen                   [page 11]