Secure Shell Transport Model for SNMP

This memo describes a Transport Model for the Simple Network Management Protocol, using the Secure Shell protocol (SSH) within a transport subsystem . The transport model specified in this memo is referred to as the Secure Shell Transport Model (SSHTM). This memo also defines a portion of the Management Information Base (MIB) for use with network management protocols in TCP/IP based internets. In particular it defines objects for monitoring and managing the Secure Shell Transport Model for SNMP. It is important to understand the SNMP architecture and the terminology of the architecture to understand where the Transport Model described in this memo fits into the architecture and interacts with other subsystems within the architecture.

For a detailed overview of the documents that describe the current Internet-Standard Management Framework, please refer to section 7 of RFC 3410 . Managed objects are accessed via a virtual information store, termed the Management Information Base or MIB. MIB objects are generally accessed through the Simple Network Management Protocol (SNMP). Objects in the MIB are defined using the mechanisms defined in the Structure of Management Information (SMI). This memo specifies a MIB module that is compliant to the SMIv2, which is described in STD 58, RFC 2578 , STD 58, RFC 2579 and STD 58, RFC 2580 .

For consistency with SNMP-related specifications, this document favors terminology as defined in STD62 rather than favoring terminology that is consistent with non-SNMP specifications. This is consistent with the IESG decision to not require the SNMPv3 terminology be modified to match the usage of other non-SNMP specifications when SNMPv3 was advanced to Full Standard. Authentication in this document typically refers to the English meaning of "serving to prove the authenticity of" the message, not data source authentication or peer identity authentication. The terms "manager" and "agent" are not used in this document, because in the RFC 3411 architecture , all SNMP entities have the capability of acting in either manager or agent or in both roles depending on the SNMP application types supported in the implementation. Where distinction is required, the application names of Command Generator, Command Responder, Notification Originator, Notification Receiver, and Proxy Forwarder are used. See "SNMP Applications" for further information. Throughout this document, the terms "client" and "server" are used to refer to the two ends of the SSH transport connection. The client actively opens the SSH connection, and the server passively listens for the incoming SSH connection. Either SNMP entity may act as client or as server, as discussed further below. The User-Based Security Model (USM) is a mandatory-to-implement Security Model in STD 62. While SSH and USM frequently refer to a user, the terminology preferred in RFC3411 and in this memo is "principal". A principal is the "who" on whose behalf services are provided or processing takes place. A principal can be, among other things, an individual acting in a particular role; a set of individuals, with each acting in a particular role; an application or a set of applications, or a combination of these within an administrative domain. 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 . Non uppercased versions of the keywords should be read as in normal English. They will usually, but not always, be used in a context relating to compatibility with the RFC3411 architecture or the subsystem defined here, but which might have no impact on on-the-wire compatibility. These terms are used as guidance for designers of proposed IETF models to make the designs compatible with RFC3411 subsystems and Abstract Service Interfaces (ASIs) (see section 3.2). Implementers are free to implement differently. Some usages of these lowercase terms are simply normal English usage.

The reader is expected to have read and understood the description of the SNMP architecture, as defined in , and the Transport Subsystem architecture extension specified in "Transport Subsystem for the Simple Network Management Protocol" . This memo describes the Secure Shell Transport Model for SNMP, a specific SNMP transport model to be used within the SNMP transport subsystem to provide authentication, encryption, and integrity checking of SNMP messages. In keeping with the RFC 3411 design decision to use self-contained documents, this document defines the elements of procedure and associated MIB module objects which are needed for processing the Secure Shell Transport Model for SNMP. This modularity of specification is not meant to be interpreted as imposing any specific requirements on implementation.

Version 3 of the Simple Network Management Protocol (SNMPv3) added security to the protocol. The User-based Security Model (USM) was designed to be independent of other existing security infrastructures, to ensure it could function when third party authentication services were not available, such as in a broken network. As a result, USM utilizes a separate user and key management infrastructure. Operators have reported that having to deploy another user and key management infrastructure in order to use SNMPv3 is a reason for not deploying SNMPv3. This memo describes a transport model that will make use of the existing and commonly deployed Secure Shell security infrastructure. This transport model is designed to meet the security and operational needs of network administrators, maximize usability in operational environments to achieve high deployment success and at the same time minimize implementation and deployment costs to minimize deployment time. This document addresses the requirement for the SSH client to authenticate the SSH server, for the SSH server to authenticate the SSH client, and describes how SNMP can make use of the authenticated identities in authorization policies for data access, in a manner that is independent of any specific access control model. This document addresses the requirement to utilize client authentication and key exchange methods which support different security infrastructures and provide different security properties. This document describes how to use client authentication as described in "SSH Authentication Protocol" . The SSH Transport Model should work with any of the ssh-userauth methods including the "publickey", "password", "hostbased", "none", "keyboard-interactive", "gssapi-with-mic", ."gssapi-keyex", "gssapi", and "external-keyx" (see http://www.iana.org/assignments/ssh-parameters). The use of the "none" authentication method is NOT RECOMMENDED, as described in Security Considerations. Local accounts may be supported through the use of the publickey, hostbased or password methods. The password method allows for integration with deployed password infrastructure such as AAA servers using the RADIUS protocol . The SSH Transport Model SHOULD be able to take advantage of future defined ssh-userauth methods, such as those that might make use of X.509 certificate credentials. It is desirable to use mechanisms that could unify the approach for administrative security for SNMPv3 and Command Line interfaces (CLI) and other management interfaces. The use of security services provided by Secure Shell is the approach commonly used for the CLI, and is the approach being adopted for use with NETCONF . This memo describes a method for invoking and running the SNMP protocol within a Secure Shell (SSH) session as an SSH subsystem. This memo describes how SNMP can be used within a Secure Shell (SSH) session, using the SSH connection protocol over the SSH transport protocol, using SSH user-auth for authentication. There are a number of challenges to be addressed to map Secure Shell authentication method parameters into the SNMP architecture so that SNMP continues to work without any surprises. These are discussed in detail below.

The design of this SNMP Transport Model is influenced by the following constraints: In times of network stress, the transport protocol and its underlying security mechanisms SHOULD NOT depend upon the ready availability of other network services (e.g., Network Time Protocol (NTP) or AAA protocols). When the network is not under stress, the transport model and its underlying security mechanisms MAY depend upon the ready availability of other network services. It may not be possible for the transport model to determine when the network is under stress. A transport model SHOULD NOT require changes to the SNMP architecture. A transport model SHOULD NOT require changes to the underlying security protocol.

SSH is a protocol for secure remote login and other secure network services over an insecure network. It consists of three major protocol components, and add-on methods for user authentication: The Transport Layer Protocol provides server authentication, and message confidentiality and integrity. It may optionally also provide compression. The transport layer will typically be run over a TCP/IP connection, but might also be used on top of any other reliable data stream. The User Authentication Protocol authenticates the client-side principal to the server. It runs over the transport layer protocol. The Connection Protocol multiplexes the encrypted tunnel into several logical channels. It runs over the transport after successfully authenticating the principal. Generic Message Exchange Authentication is a general purpose authentication method for the SSH protocol, suitable for interactive authentications where the authentication data should be entered via a keyboard Generic Security Service Application Program Interface (GSS-API) Authentication and Key Exchange for the Secure Shell (SSH) Protocol describes methods for using the GSS-API for authentication and key exchange in SSH. It defines an SSH user authentication method that uses a specified GSS-API mechanism to authenticate a user, and a family of SSH key exchange methods that use GSS-API to authenticate a Diffie-Hellman key exchange. The client sends a service request once a secure transport layer connection has been established. A second service request is sent after client authentication is complete. This allows new protocols to be defined and coexist with the protocols listed above. The connection protocol provides channels that can be used for a wide range of purposes. Standard methods are provided for setting up secure interactive shell sessions and for forwarding ("tunneling") arbitrary TCP/IP ports and X11 connections.

A transport model is a component of the Transport Subsystem within the SNMP architecture. The SSH Transport Model thus fits between the underlying SSH transport layer and the message dispatcher . The SSH Transport Model will establish a channel between itself and the SSH Transport Model of another SNMP engine. The sending transport model passes unencrypted messages from the dispatcher to SSH to be encrypted, and the receiving transport model accepts decrypted incoming messages from SSH and passes them to the dispatcher. After an SSH Transport model channel is established, then SNMP messages can conceptually be sent through the channel from one SNMP message dispatcher to another SNMP message dispatcher. Multiple SNMP messages MAY be passed through the same channel. The SSH Transport Model of an SNMP engine will perform the translation between SSH-specific security parameters and SNMP-specific, model-independent parameters.

The Secure Shell Transport Model provides protection against the threats identified by the RFC 3411 architecture : Modification of Information - SSH provides for verification that the contents of each message has not been modified during its transmission through the network, by digitally signing each SSH packet. Masquerade - SSH provides for verification of the identity of the SSH server and the identity of the SSH client. SSH provides for verification of the identity of the SSH server through the SSH Transport Protocol server authentication . This allows an operator or management station to ensure the authenticity of the SNMP engine that provides MIB data. SSH provides a number of mechanisms for verification of the identity of the SSH client-side principal, using the Secure Shell Authentication Protocol . These include public key, password and host-based mechanisms. This allows the SNMP access control subsystem to ensure that only authorized principals have access to potentially sensitive data. Verification of client's principal identity is important for use with the SNMP access control subsystem, to ensure that only authorized principals have access to potentially sensitive data. The SSH user identity is provided to the transport model, so it can be used to map to an SNMP model-independent securityName for use with SNMP access control and notification configuration. (The identity may undergo various transforms before it maps to the securityName.) Message Stream Modification - SSH protects against malicious re-ordering or replaying of messages within a single SSH session by using sequence numbers and integrity checks. SSH protects against replay of messages across SSH sessions by ensuring that the cryptographic keys used for encryption and integrity checks are generated afresh for each session. Disclosure - SSH provides protection against the disclosure of information to unauthorized recipients or eavesdroppers by allowing for encryption of all traffic between SNMP engines.

The RFC 3411 architecture recognizes three levels of security: - without authentication and without privacy (noAuthNoPriv) - with authentication but without privacy (authNoPriv) - with authentication and with privacy (authPriv) The Secure Shell protocol provides support for encryption and data integrity. While it is technically possible to support no authentication and no encryption in SSH it is NOT RECOMMENDED by . The SSH Transport Model determines from SSH the identity of the authenticated principal, and the type and address associated with an incoming message, and provides this information to SSH for an outgoing message. The SSH transport layer algorithms used to provide authentication, data integrity and encryption SHOULD NOT be exposed to the SSH Transport Model layer. The SNMPv3 WG deliberately avoided this and settled for an assertion by the security model that the requirements of securityLevel were met. The SSH Transport Model has no mechanisms by which it can test whether an underlying SSH connection provides auth or priv, so the SSH Transport Model trusts that the underlying SSH connection has been properly configured to support authPriv security characteristics. An SSH Transport Model-compliant implementation MUST use an SSH connection that provides authentication, data integrity and encryption that meets the highest level of SNMP security (authPriv). Outgoing messages specified with a securityLevel of noAuthNoPriv or authNoPriv, are actually sent by the SSH Transport Model with authPriv-level protection. The security protocols used in the Secure Shell Authentication Protocol and the Secure Shell Transport Layer Protocol are considered acceptably secure at the time of writing. However, the procedures allow for new authentication and privacy methods to be specified at a future time if the need arises.

The SSH Transport Model should support any server or client authentication mechanism supported by SSH. This includes the three authentication methods described in the SSH Authentication Protocol document - publickey, password, and host-based - and keyboard interactive and others. The password authentication mechanism allows for integration with deployed password based infrastructure. It is possible to hand a password to a service such as RADIUS or Diameter for validation. The validation could be done using the user-name and user-password attributes. It is also possible to use a different password validation protocol such as CHAP or digest authentication to integrate with RADIUS or Diameter. At some point in the processing, these mechanisms require the password be made available as clear text on the device that is authenticating the password which might introduce threats to the authentication infrastructure. GSSKeyex provides a framework for the addition of client authentication mechanisms which support different security infrastructures and provide different security properties. Additional authentication mechanisms, such as one that supports X.509 certificates, may be added to SSH in the future.

This document describes the use of an SSH subsystem for SNMP to make SNMP usage distinct from other usages. An SSH subsystem of type "snmp" is opened by the SSH Transport Model during the elements of procedure for an outgoing SNMP message. Since the sender of a message initiates the creation of an SSH session if needed, the SSH session will already exist for an incoming message or the incoming message would never reach the SSH Transport Model. Implementations may choose to instantiate SSH sessions in anticipation of outgoing messages. This approach might be useful to ensure that an SSH session to a given target can be established before it becomes important to send a message over the SSH session. Of course, there is no guarantee that a pre-established session will still be valid when needed. SSH sessions are uniquely identified within the SSH Transport Model by the combination of tmTransportAddress and tmSecurityName associated with each session. Because naming policies might differ between administrative domains, many SSH client software packages support a user@hostname:port addressing syntax that operators can use to align non-equivalent account names. The SnmpSSHAddress Textual Convention echos this common SSH notation. When this notation is used in an SnmpSSHAddress, the SSH connection should be established with a SSH user name matching the "user" portion of the notation when establishing a session with the remote SSH server. The user name must be encoded in UTF-8 (per RFC4252 ). The "user" portion may or may not match the tmSecurityName parameter passed from the security model. If no "user@" portion is specified in the SnmpSSHAddress, then the SSH connection should be established using the tmSecurityName as the SSH user name when establishing a session with the remote SSH server. The SnmpSSHAddress and tmSecurityName associated with an SSH session MUST remain constant during the life of the session. Different SnmpSSHAddress values (with different hostnames, "user@" prefix names and/or port numbers) will each result in individual SSH sessions.

For incoming messages, SSH-specific security parameters are translated by the transport model into security parameters independent of the transport and security models. The transport model accepts messages from the SSH subsystem, and records the transport-related and SSH-security-related information, including the authenticated identity, in a cache referenced by tmStateReference, and passes the WholeMsg and the tmStateReference to the dispatcher using the receiveMessage() ASI (Application Service Interface). For outgoing messages, the transport model takes input provided by the dispatcher in the sendMessage() ASI. The SSH Transport Model converts that information into suitable security parameters for SSH, establishes sessions as needed, and passes messages to the SSH subsystem for sending.

SSH connections may be initiated by command generators or by notification originators. Command generators are frequently operated by a human, but notification originators are usually unmanned automated processes. As a result, it may be necessary to provision authentication credentials on the SNMP engine containing the notification originator, or use a third party key provider such as Kerberos, so the engine can successfully authenticate to an engine containing a notification receiver. The targets to whom notifications or proxy requests should be sent is typically determined and configured by a network administrator. The SNMP-NOTIFICATION-MIB contains a list of targets to which notifications should be sent. The SNMP-TARGET-MIB module contains objects for defining these management targets, including transport domains and addresses and security parameters, for applications such as notification generators and proxy forwarders. For the SSH Transport Model, transport type and address are configured in the snmpTargetAddrTable, and the securityName, and securityLevel parameters are configured in the snmpTargetParamsTable. The default approach is for an administrator to statically preconfigure this information to identify the targets authorized to receive notifications or received proxied messages. Local access control processing needs to be performed by a Notification Originator before notifications are actually sent and this processing is done using the configured securityName. An important characteristic of this is that authorization is done prior to determining if the connection can succeed. Thus the locally configured securityName is entirely trusted within the Notification Originator. The SNMP-TARGET-MIB and NOTIFICATION-MIB MIB modules may be configured using SNMP or other implementation-dependent mechanisms, such as CLI scripting or loading a configuration file. It may be necessary to provide additional implementation-specific configuration of SSH parameters.

When performing SNMP processing, there are two levels of state information that may need to be retained: the immediate state linking a request-response pair, and potentially longer-term state relating to transport and security. "Transport Subsystem for the Simple Network Management Protocol" defines general requirements for caches and references. This document defines additional cache requirements related to the Secure Shell Transport Model.

The Secure Shell Transport Model has specific responsibilities regarding the cached information. See the Elements of Procedure in for detailed processing instructions on the use of the tmStateReference fields by the SSH Transport Model.

The tmSecurityName MUST be a human-readable name (in snmpAdminString format) representing the identity that has been set according to the procedures in . The tmSecurityName MUST be constant for all traffic passing through an SSHTM session. Messages MUST NOT be sent through an existing SSH session that was established using a different tmSecurityName. On the SSH server side of a connection: The tmSecurityName should be the SSH user name. How the SSH user name is extracted from the SSH layer is implementation-dependent. The SSH protocol is not always clear on whether the user name field must be filled in, so for some implementations, such as those using GSSAPI authentication, it may be necessary to use a mapping algorithm to transform an SSH identity to a tmSecurityName, or to transform a tmSecurityName to an SSH identity. In other cases the user name may not be verified by the server, so for these implementations, it may be necessary to obtain the user name from other credentials exchanged during the SSH exchange. On the SSH client side of a connection: The tmSecurityName is presented to the SSH Transport Model by the application (possibly because of configuration specified in the SNMP-TARGET-MIB). The securityName MAY be derived from the tmSecurityName by a security model, and MAY be used to configure notifications and access controls in MIB modules. Transport models SHOULD generate a predictable tmSecurityName so operators will know what to use when configuring MIB modules that use securityNames derived from tmSecurityNames.

The tmSessionID MUST be recorded per message at the time of receipt. When tmSameSecurity is set, the recorded tmSessionID can be used to determine whether the SSH session available for sending a corresponding outgoing message is the same SSH session as was used when receiving the incoming message (e.g., a response to a request).

The per-session state that is referenced by tmStateReference may be saved across multiple messages in a Local Configuration Datastore. Additional session/connection state information might also be stored in a Local Configuration Datastore.

Abstract service interfaces have been defined by and further augmented by to describe the conceptual data flows between the various subsystems within an SNMP entity. The Secure Shell Transport Model uses some of these conceptual data flows when communicating between subsystems. To simplify the elements of procedure, the release of state information is not always explicitly specified. As a general rule, if state information is available when a message gets discarded, the message-state information should also be released, and if state information is available when a session is closed, the session state information should also be released. An error indication in statusInformation will typically include the OID and value for an incremented error counter. This may be accompanied by the requested securityLevel, and the tmStateReference. Per-message context information is not accessible to Transport Models, so for the returned counter OID and value, contextEngine would be set to the local value of snmpEngineID, and contextName to the default context for error counters.

The SSH Transport Model queries the SSH engine, in an implementation-dependent manner, to determine the address the message originated from, the user name authenticated by SSH, and a session identifier. Determine the tmTransportAddress to be associated with the incoming message: If this is a client-side SSH session, then the tmTransportAddress is set to the tmTransportAddress used to establish the session. It MUST exactly include any "user@" prefix associated with the address provided to the openSession() ASI. If this is a server-side SSH session and this is the first message received over the session, then the tmTransportAddress is set to the address the message originated from, determined in an implementation-dependent way. This value MUST be constant for the entire SSH session and future messages received MUST result in the tmTransportAddress being set to the same value. If this is a server-side SSH session and this is not the first message received over the session, then the tmTransportAddress is set to the previously established tmTransportAddress for the session (the value from step B. determined from a previous incoming message). Determine the tmSecurityName to be associated with the incoming message: If this is a client-side SSH session, then the tmSecurityName MUST be set to the tmSecurityName used to establish the session. If this is a server-side SSH session and this is the first message received over the session, then the tmSecurityName is set to the SSH user name. How the SSH user name is extracted from the SSH layer is implementation-dependent. This value MUST be constant for the entire SSH session and future messages received MUST result in the tmSecurityName being set to the same value. If this is a server-side SSH session and this is not the first message received over the session, then the tmSecurityName is set to the previously established tmSecurityName for the session (the value from step B. determined from a previous incoming message). Create a tmStateReference cache for subsequent reference to the information. tmTransportDomain = snmpSSHDomain tmTransportAddress = the derived tmTransportAddress from step 2. tmSecurityName = The derived tmSecurityName from step 3. tmTransportSecurityLevel = "authPriv" (authentication and confidentiality MUST be used to comply with this transport model.) tmSessionID = an implementation-dependent value that can be used to detect when a session has closed and been replaced by another session. The value in tmStateReference MUST uniquely identify the session over which the message was received. This session identifier MUST NOT be reused until there are no references to it remaining. Then the Transport model passes the message to the Dispatcher using the following ASI:

The Dispatcher passes the information to the Transport Model using the ASI defined in the transport subsystem:

The SSH Transport Model performs the following tasks. If tmStateReference does not refer to a cache containing values for tmTransportDomain, tmTransportAddress, tmSecurityName, tmRequestedSecurityLevel, and tmSameSecurity, then increment the snmpSshtmSessionInvalidCaches counter, discard the message and return the error indication in the statusInformation. Processing of this message stops. Extract the tmTransportDomain, tmTransportAddress, tmSecurityName, tmRequestedSecurityLevel, tmSameSecurity, and tmSessionID from the tmStateReference. Identify an SSH session to send the messages over: If tmSameSecurity is true and there is no existing session with a matching tmSessionID, tmSecurityName and tmTransportAddress, then increment the snmpSshtmSessionNoSessions counter, discard the message and return the error indication in the statusInformation. Processing of this message stops. If there is a session with a matching tmSessionID, tmTransportAddress and tmSecurityName, then select that session. If there is a session that matches the tmTransportAddress and tmSecurityName, then select that session. If the above steps failed to select a session to use, then call openSession() with the tmStateReference as a parameter. If openSession fails, then discard the message, release tmStateReference and pass the error indication returned by openSession back to the calling module. Processing stops for this message. If openSession succeeds, then record the destTransportDomain, destTransportAddress, tmSecurityname and tmSessionID in an implementation-dependent manner. This will be needed when processing an incoming message. Pass the wholeMessage to SSH for encapsulation as data in an SSH message over the identified SSH session. Any necessary additional SSH-specific parameters should be provided in an implementation-dependent manner.

The Secure Shell Transport Model provides the following abstract service interface (ASI) to describe the data passed between the SSH Transport Model and the SSH service. It is an implementation decision how such data is passed.

The following describes the procedure to follow to establish a session between a client and server to run SNMP over SSH. This process is used by any SNMP engine establishing a session for subsequent use. This will be done automatically for an SNMP application that initiates a transaction, such as a Command Generator or a Notification Originator or a Proxy Forwarder. Increment the snmpSshtmSessionOpens counter. Using tmTransportAddress, the client will establish an SSH transport connection using the SSH transport protocol, authenticate the server, and exchange keys for message integrity and encryption. The transportAddress associated with a session MUST remain constant during the lifetime of the SSH session. Implementations may need to cache the transportAddress passed to the openSession API for later use when performing incoming message processing (see section ). To authenticate the server, the client usually stores (tmTransportAddress, server host public key) pairs in an implementation-dependent manner. The other parameters of the transport connection are provided in an implementation-dependent manner. If the attempt to establish a connection is unsuccessful, or server authentication fails, then snmpSshtmSessionOpenErrors is incremented, an openSession error indication is returned, and openSession processing stops. The client will then invoke an SSH authentication service to authenticate the principal, such as that described in the SSH authentication protocol . If the tmTransportAddress field contains a user-name followed by an '@' character (US-ASCII 0x40), that user-name string should be presented to the ssh server as the "user name" for user authentication purposes. If there is no user-name in the tmTransportAddress then the tmSecurityName should be used as the user-name. The credentials used to authenticate the SSH principal are determined in an implementation-dependent manner. In an implementation-specific manner, invoke the SSH user authentication service using the calculated user-name. If the user authentication is unsuccessful, then the transport connection is closed, the snmpSshtmSessionUserAuthFailures counter is incremented, an error indication is returned to the calling module, and processing stops for this message. The client should invoke the "ssh-connection" service (also known as the SSH connection protocol ), and request a channel of type "session". If unsuccessful, the transport connection is closed, the snmpSshtmSessionNoChannels counter is incremented, an error indication is returned to the calling module, and processing stops for this message. The client invokes "snmp" as an SSH subsystem, as indicated in the "subsystem" parameter. If unsuccessful, the transport connection is closed, the snmpSshtmSessionNoSubsystems counter is incremented, an error indication is returned to the calling module, and processing stops for this message. In order to allow SNMP traffic to be easily identified and filtered by firewalls and other network devices, servers associated with SNMP entities using the Secure Shell Transport Model MUST default to providing access to the "snmp" SSH subsystem if the SSH session is established using the IANA-assigned TCP ports (YYY and ZZZ). Servers SHOULD be configurable to allow access to the SNMP SSH subsystem over other ports. Set tmSessionID in the tmStateReference cache to an implementation-dependent value to identify the session. The tmSecurityName used to establish the SSH session must be the only tmSecurityName used with the session. Incoming messages for the session MUST be associated with this tmSecurityName value. How this is accomplished is implementation-dependent.

The Secure Shell Transport Model provides the following ASI to close a session:

The following describes the procedure to follow to close a session between a client and server. This process is followed by any SNMP engine to close an SSH session. It is implementation-dependent when a session should be closed. The calling code should release the associated tmStateReference. Increment the snmpSshtmSessionCloses counter. If there is no session corresponding to tmSessionID, then closeSession processing is completed. Have SSH close the session associated with tmSessionID.

This MIB module provides management of the Secure Shell Transport Model. It defines an OID to identify the SNMP-over-SSH transport domain, a textual convention for SSH Addresses and several statistics counters.

Objects in this MIB module are arranged into subtrees. Each subtree is organized as a set of related objects. The overall structure and assignment of objects to their subtrees, and the intended purpose of each subtree, is shown below.

Generic and Common Textual Conventions used in this document can be found summarized at http://www.ops.ietf.org/mib-common-tcs.html

Some management objects defined in other MIB modules are applicable to an entity implementing the SSH Transport Model. In particular, it is assumed that an entity implementing the SNMP-SSH-TM-MIB will implement the SNMPv2-MIB , and the SNMP-FRAMEWORK-MIB . It is expected that an entity implementing this MIB will also support the Transport Security Model , and therefore implement the SNMP-TSM-MIB. This MIB module is for monitoring SSH Transport Model information.

The following MIB module imports items from , , . This MIB module also references , , and This document uses TDomain textual conventions for the SNMP-internal MIB modules defined here for compatibility with the RFC3413 MIB modules and the RFC3411 Abstract Service Interfaces.

The SSH Transport Model will likely not work in conditions where remote access to the CLI has stopped working. The SSH Transport Model assumes that TCP and IP continue to operate correctly between the communicating nodes. Failures in either node, death of the deamon serving the communication, routing problems in the network between, firewalls that block the traffic, and other problems can prevent the SSH Transport Model from working. In situations where management access has to be very reliable, operators should consider mitigating measures. These measures may include dedicated management-only networks, point-to-point links, and the ability to use alternate protocols and transports. To have SNMP properly utilize the security services provided by SSH, the SSH Transport Model MUST be used with a Security Model that knows how to process a tmStateReference, such as the Transport Security Model for SNMP . If the SSH Transport Model is configured to utilize AAA services, operators should consider configuring support for local authentication mechanisms, such as local passwords, so SNMP can continue operating during times of network stress. The SSH protocol has its own window mechanism, defined in RFC 4254. The SSH specifications leave it open when window adjustments messages should be created, and some implementations send these whenever received data has been passed to the application. There are noticeable bandwidth and processing overheads to handling such window adjustment messages, which can be avoided by sending them less frequently. The SSH protocol requires the execution of CPU intensive calculations to establish a session key during session establishment. This means that short lived sessions become computationally expensive compared to USM, which does not have a notion of a session key. Other transport security protocols such as TLS support a session resumption feature that allows reusing a cached session key. Such a mechanism does not exist for SSH and thus SNMP applications should keep SSH sessions for longer time periods. To initiate SSH connections, an entity must be configured with SSH client credentials plus information to authenticate the server. While hosts are often configured to be SSH clients, most internetworking devices are not. To send notifications over SSHTM, the internetworking device will need to be configured as an SSH client. How this credential configuration is done is implementation and deployment specific.

This document describes a transport model that permits SNMP to utilize SSH security services. The security threats and how the SSH Transport Model mitigates those threats is covered in detail throughout this memo. The SSH Transport Model relies on SSH mutual authentication, binding of keys, confidentiality and integrity. Any authentication method that meets the requirements of the SSH architecture will provide the properties of mutual authentication and binding of keys. SSHv2 provides Perfect Forward Security (PFS) for encryption keys. PFS is a major design goal of SSH, and any well-designed keyex algorithm will provide it. The security implications of using SSH are covered in . The SSH Transport Model has no way to verify that server authentication was performed, to learn the host's public key in advance, or verify that the correct key is being used. The SSH Transport Model simply trusts that these are properly configured by the implementer and deployer. SSH provides the "none" userauth method. The SSH Transport Model MUST NOT be used with an SSH connection with the "none" userauth method. While SSH does support turning off confidentiality and integrity, they MUST NOT be turned off when used with the SSH Transport Model. The SSH protocol is not always clear on whether the user name field must be filled in, so for some implementations, such as those using GSSAPI authentication, it may be necessary to use a mapping algorithm to transform an SSH identity to a tmSecurityName, or to transform a tmSecurityName to an SSH identity. In other cases the user name may not be verified by the server, so for these implementations, it may be necessary to obtain the user name from other credentials exchanged during the SSH exchange.

Most key exchange algorithms are able to authenticate the SSH server's identity to the client. However, for the common case of DH signed by public keys, this requires the client to know the host's public key a priori and to verify that the correct key is being used. If this step is skipped, then authentication of the SSH server to the SSH client is not done. Data confidentiality and data integrity protection to the server still exist, but these are of dubious value when an attacker can insert himself between the client and the real SSH server. Note that some userauth methods may defend against this situation, but many of the common ones (including password and keyboard-interactive) do not, and in fact depend on the fact that the server's identity has been verified (so passwords are not disclosed to an attacker). SSH MUST NOT be configured to skip public key verification for use with the SSH Transport Model.

SNMP Notifications are authorized to be sent to a receiver based on the securityName used by the notification originator's SNMP engine. This authorization is performed before the message is actually sent and before the credentials of the remote receiver have been verified. It is thus critical that the credentials presented by a notification receiver MUST match the expected value(s) for a given transport address and that ownership of the credentials MUST be properly cryptographically verified.

If a "user@" prefix is used within a SnmpSSHAddress value to specify a SSH user name to use for authentication then the key presented to the remote entity MUST be the key expected by the server for the "user". This may be different than a locally cached key identified by the securityName value.

The User Based Security Model employed symmetric cryptography and user naming conventions. SSH employs an asymmetric cryptographic and naming model. Unlike USM, cryptographic keys will be different on both sides of the SSH connection. Both sides are responsible for verifying that the remote entity presents the right key. The optional "user@" prefix component of the SnmpSSHAddress Textual Convention allows the client SNMP stack to associate the connection with a securityName that may be different than the SSH user name presented to the SSH server.

SSH provides the "none" MAC algorithm, which would allow you to turn off data integrity while maintaining confidentiality. However, if you do this, then an attacker may be able to modify the data in flight, which means you effectively have no authentication. SSH MUST NOT be configured using the "none" MAC algorithm for use with the SSH Transport Model.

The SNMPv1 and SNMPv2c message processing described in RFC3584 (BCP 74) always select the SNMPv1 or SNMPv2c Security Models respectively. Both of these, and the User-based Security Model typically used with SNMPv3, derive the securityName and securityLevel from the SNMP message received, even when the message was received over a secure transport. Access control decisions are therefore made based on the contents of the SNMP message, rather than using the authenticated identity and securityLevel provided by the SSH Transport Model.

There are no management objects defined in this MIB module that have a MAX-ACCESS clause of read-write and/or read-create. So, if this MIB module is implemented correctly, then there is no risk that an intruder can alter or create any management objects of this MIB module via direct SNMP SET operations. Some of the readable objects in this MIB module (i.e., objects with a MAX-ACCESS other than not-accessible) may be considered sensitive or vulnerable in some network environments. It is thus important to control even GET and/or NOTIFY access to these objects and possibly to even encrypt the values of these objects when sending them over the network via SNMP. These are the tables and objects and their sensitivity/vulnerability: The information in the snmpSshtmSession group is generated locally when a client session is being opened or closed. This information can reflect the configured capabilities of a remote SSH server, which could be helpful to an attacker for focusing an attack. SNMP versions prior to SNMPv3 did not include adequate security. Even if the network itself is secure (for example by using IPSec or SSH), even then, there is no control as to who on the secure network is allowed to access and GET/SET (read/change/create/delete) the objects in this MIB module. It is RECOMMENDED that implementers consider the security features as provided by the SNMPv3 framework (see section 8), including full support for cryptographic mechanisms for authentication and privacy, such as those found in the User-based Security Model , the Transport Security Model and the SSH Transport Model described in this document. Further, deployment of SNMP versions prior to SNMPv3 is NOT RECOMMENDED. Instead, it is RECOMMENDED to deploy SNMPv3 and to enable cryptographic security. It is then a customer/operator responsibility to ensure that the SNMP entity giving access to an instance of this MIB module is properly configured to give access to the objects only to those principals (users) that have legitimate rights to indeed GET or SET (change/create/delete) them.

IANA is requested to assign: two TCP registered port numbers in the http://www.iana.org/assignments/port-numbers registry which will be the default ports for SNMP over an SSH Transport Model as defined in this document, and SNMP over an SSH Transport Model for notifications as defined in this document. It would be good if the assigned keywords were "snmpssh" and "snmpssh-trap" and the port numbers were x161 and x162 (where x represents a number, such as '5' or '6', so that we end up with ports 5161 and 5162, or 6161 and 6162). an SMI number under mib-2, for the MIB module in this document, an SMI number under snmpDomains, for the snmpSSHDomain, "ssh" as the corresponding prefix for the snmpSSHDomain in the SNMP Transport Model registry; defined in "snmp" as an SSH Subsystem Name in the http://www.iana.org/assignments/ssh-parameters registry. -- note to RFC editor -- Please replace YYY and ZZZ in this document with the assigned port numbers and remove this note.

The editors would like to thank Jeffrey Hutzelman for sharing his SSH insights, and Dave Shield for an outstanding job wordsmithing the existing document to improve organization and clarity. Additionally, helpful document reviews were received from: Juergen Schoenwaelder.