Researchers Reveal Holes in Grid
New research into security weaknesses in a popular secure remote access technology highlights the vulnerability of large-scale computing environments such as grids and supercomputing clusters to a potentially crippling form of attack.
The SSH (Secure Shell) technology is used by security administrators and other technical users to connect to remote machines via a secure, encrypted tunnel.
The system is used widely in university and research networks, and security researchers at the Massachusetts Institute of Technology last week published a paper showing how a simple worm could grab SSH user credentials from one machine and move rapidly among any number of connected systems, causing whats known as a "cascade failure."
This type of attack would be especially damaging in a grid or cluster environment, in which hundreds or thousands of high-powered computers are linked together and shared by multiple users.
Many industry heavyweights, including Sun Microsystems Inc., Oracle Corp. and IBM, are pushing grid or on-demand computing initiatives, although little has been said about the security of such architectures.
The researchers, from MITs Computer Science and Artificial Intelligence Lab and Lincoln Laboratory, found that malicious hackers could use an SSH file called "known_hosts" to gather lists of remote systems visited by other users via SSH.
Information in known_hosts files is not usually encrypted or otherwise protected from viewing. While the data stored in the files doesnt give a hacker access to those systems, it could be a road map to other servers running SSH, said Will Stockwell, one of the papers authors and an undergraduate researcher at CSAIL, in Cambridge, Mass.
For example, known_hosts data, when coupled with a critical and widespread SSH vulnerability, could allow a virus or worm to quickly find and try to infect other systems that it knows run SSH, Stockwell said.
Even though SSH has been widely deployed for years, the security holes raised by the MIT researchers escaped notice, experts agree.
"Nobody realized they were there. Security involves someone saying, You can do this," said Bruce Schneier, chief technology officer of Counterpane Internet Security Inc., of Cupertino, Calif., and a widely respected cryptographer.
And the MIT research suggests that known_hosts could be a gold mine for a virus or worm author. The team from MIT collected more than 8,000 unique addresses from known_hosts files on 92 systems, and researchers believe that hackers could harvest many more with unfettered access to machines on which SSH is used to manage other systems, according to the MIT report.
Such a worm is still hypothetical, but experts say that an attacker last year used a compromised version of SSH to gather user credentials and break into sensitive machines at Stanford University; The National Supercomputing Center; and the TeraGrid, a network of supercomputing centers.
The supercomputing community is also trying to learn from last years hacks. Leading scientists from labs across the world gathered for the first-ever International Workshop on Cluster Security at the IEEE CCGrid conference in Cardiff, Wales, last week to discuss short- and long-term fixes for security issues, according to Bill Yurcik, senior security engineer at the National Center for Supercomputing Applications at the University of Illinois, in Urbana-Champaign.
The interconnected nature of supercomputing and cluster environments means an SSH worm that could traverse known_hosts files could create a cascade failure, quickly moving between linked research networks, according to a presentation given at the conference by Stuart Schechter of Lincoln Lab, another author of the SSH paper.