There is a vulnerability in the SSH protocol that can simplify brute force attacks against passwords typed within an existing SSH session.
Researchers at the University of California at Berkeley have determined that by monitoring the delays between SSH packets transmitted across the network, it is possible to make educated guesses about the keystrokes typed by the user. This vulnerability relies primarily upon the fact that in interactive SSH sessions, each keystroke made by the user causes the SSH client to transmit one IP packet to the SSH server. Similarly, as the remote server echoes the typed characters back to the user, the SSH server sends individual IP packets back to the SSH client.
This behavior (which is common to many terminal programs) creates identifiable patterns in the packet data that yield information about the user's activities. For example, as described in the Berkeley paper, if a system administrator logs into a remote Unix system and types the su command to become the superuser, the process will look similar to this:
In addition to these limitations, there are a few other factors to consider when judging the severity of this vulnerability. First, this vulnerability does not rely upon a cryptographic flaw; it is based upon a statistical analysis of network traffic that is independent of cipher or SSH protocol version. Second, it is easy to misinterpret this vulnerability as a weakness in the initial SSH login authentication and to conclude that using strong SSH authentication (e.g. RSA) is a defense against this problem. In fact, the initial authentication method is irrelevant. This technique analyzes traffic generated after the initial authentication, and passwords are just one possible type of data that could be analyzed and exposed.
This vulnerability reduces the number of guesses needed to perform brute-force attacks against passwords typed in an existing SSH session.
The CERT/CC is currently unaware of a practical solution to this problem.
Mask echo behavior in terminal programs
This vulnerability was discovered independently by Solar Designer and Dug Song of The Openwall Project as well as Dawn Xiaodong Song, David Wagner, and Xuqing Tian of the University of California at Berkeley; the CERT/CC thanks both parties for their research and analysis.
This document was written by Jeffrey Lanza.
|Date First Published:||2001-09-28|
|Date Last Updated:||2001-12-14 21:12 UTC|