NSA wants to inject more science into computer security

Many information security professionals salivate over discovering a new attack shield, but not as many are interested or are skilled enough to apply the scientific method to their findings, said Tom Longstaff, technical director of the National Security Agency's systems behavior group. He is on a mission to promote scientific research into computer security for use in the government.

On March 7, the National Science Foundation called for papers for a June symposium on "moving target" research into dynamic defenses intended to constantly confuse intruders.

"I'm really keeping my fingers crossed that we can fill [a full day's worth of presentations]," said Longstaff, who is organizing the program with several other government and university experts.

The stumbling blocks to employing scientific principles for cybersecurity research include the time required to publish results capable of being replicated, a dearth of peer reviewers in the young field and accurate data capture, he said.

Data analysis is not just a problem for cyber academics. A September 2011 data collection foible at the CERN laboratory in Geneva nearly refuted Einstein's theory that nothing travels faster than the speed of light. CERN scientists reported they had identified a particle known as a neutrino that broke the light speed limit. But the journal Science last month largely debunked that discovery by finding an error in the experiment's setup: a faulty connection between a GPS unit and a computer.

Longstaff, who made his comments in a lecture at NSF's Arlington headquarters, said the physicists who mistakenly thought they had made a major breakthrough in reality performed a public service by clearly writing up and circulating their results so that others could refute them.

Cybersecurity peer reviewers are scarce, he said, because the United States has failed to train information security specialists to critically review their work.

To address the science's limitations, Longstaff suggested the government fund textbooks and document the knowledge of cyber professionals who know how to practice the scientific approach. Additionally, cybersecurity students should be trained in the scientific method, said Longstaff, who also serves as chairman of Johns Hopkins University's computer science, information assurance and information systems engineering programs.

But convincing computer whizzes to pursue research rather than incident response may be a tough sell, he said.

"There's nothing that I'm going to say to someone who wants to do innovation that's going to make them want to do science," Longstaff said in an interview after his remarks.