OpenSSH released, plays in sandbox

The latest version of OpenSSH was released overnight and introduces an experimental sandbox that all but eliminates pre-authentication attacks.

The tool set updated to 5.9 was a frontrunner in adoption of privilege separated architecture which helped mitigate compromises.  It meant network-facing code had reduced privileges or was placed into a chroot jail.

Google staffer and Melbourne-based OpenSSH developer, Damien Miller, said the new sandbox further restricted the code privileges.

“On a platform with a working sandbox implementation, it will no longer be able to open sockets of its own, fork new processes or attempt to use many system calls,” Miller said.

“This makes a pre-authentication compromise of a sshd (secure shell daemon) pretty much useless to an attacker.”

A pre-authentication attack would exploit the code and functionality of the server side of OpenSSH that was exposed before users log-in.

 “When considering the security of network applications, this is almost always the most sensitive code because anyone can access it and a successful compromise can be used to write fully automatic exploits [like] network worms.”

OpenSSH also supported Hash-based Message Authentication Code (HMAC) SHA256/512 within the SSH transport protocol to provide a choice of cryptographic options should one later prove broken.

The remaining bugfixes, available on the website, were numerous but not major.

Miller said the OpenSSH developers aimed to release updated versions about every three months.

“We prefer a steady stream of releases with fewer features to occasional ones that change everything. We also try to always be ready to release at a moment's notice, so if someone reports a security problem to us we can get a release out within days if necessary.”