10. Security Considerations

The Dispatcher coordinates the processing of messages to provide a level of security for management messages and to direct the SNMP PDUs to the proper SNMP application(s).

A Message Processing Model, and in particular the v3MP defined in this document, interacts as part of the Message Processing with Security Models in the Security Subsystem via the abstract service interface primitives defined in [RFC3411] and elaborated above.

The level of security actually provided is primarily determined by the specific Security Model implementation(s) and the specific SNMP application implementation(s) incorporated into this framework. Applications have access to data which is not secured. Applications should take reasonable steps to protect the data from disclosure, and when they send data across the network, they should obey the securityLevel and call upon the services of an Access Control Model as they apply access control.

The values for the msgID element used in communication between SNMP entities MUST be chosen to avoid replay attacks. The values do not need to be unpredictable; it is sufficient that they not repeat.

When exchanges are carried out over an insecure network, there is an open opportunity for a third party to spoof or replay messages when any message of an exchange is given at the security level of noAuthNoPriv. For most exchanges, all messages exist at the same security level. In the case where the final message is an Internal Class PDU, this message may be delivered at a level of noAuthNoPriv or authNoPriv, independent of the security level of the preceding messages. Internal Class PDUs delivered at the level of authNoPriv are not considered to pose a security hazard. Internal Class PDUs delivered at the security level of noAuthNoPriv open a window of opportunity for spoofing or replay attacks. If the receiver of such messages is aware of these risks, the use of such unauthenticated messages is acceptable and may provide a useful function for discovering engine IDs or for detecting misconfiguration at remote nodes.

This document also contains a MIB definition module. None of the objects defined is writable, and the information they represent is not deemed to be particularly sensitive. However, if they are deemed sensitive in a particular environment, access to them should be restricted through the use of appropriately configured Security and Access Control models.