Maj. Gen. Steven Smith, who heads the Army Cyber Directorate, said that putting such software in place is now one of his main priorities, the Army Times recently reported.
Such software would create benchmarks of normal behavior, then watch for any activity that looked abnormal. "So I'm on the South American desk, doing intelligence work, and all of a sudden I start going around to China, let's say," said Smith. "That might be an anomaly, it might be justified, but I would sure like to know that and let someone make a decision, almost at the speed of thought." He said his desired system would record downloads, Web search queries, and complete keystrokes.
Information security experts said that what the Army is proposing is likely possible. "We're verging on the capability of being able to handle, from a technological basis, that quantity of data," said Scott Crawford, managing research director for Enterprise Management Associates, via phone. "But can you really automate the process without being overwhelmed by false positives, or stumped by false negatives?"
[ Agency officials are struggling with many issues, but Security Is Top Concern Of Federal CIOs. ]
The Army's search for new monitoring tools is part of a broader Pentagon push to help detect when any bad actors--domestic or foreign--are accessing military or government networks. But the impetus for this particular wave of improvements can be traced directly to Army private Bradley Manning, the former intelligence analyst who's accused of copying sensitive State Department cables and almost 500,000 battlefield reports from Afghanistan and Iraq onto a recordable CD, then releasing them to WikiLeaks. He's likewise accused of leaking U.S. helicopter gunship footage, which WikiLeaks released under the banner of "Collateral Murder."
Obviously, the stolen data didn't make the State Department or Army look good. Likewise, WikiLeaks ultimately released the cables in unredacted form, which the U.S. government said put at least 100 confidential diplomatic sources at risk.
But questions remain about whether the Army's plan to analyze keystrokes to spot malicious insiders would be affordable, feasible, or even help prevent the next big breach. As an anonymous information assurance engineer posted to a related DataBreaches.net discussion, technology is only a first step. Someone still has to investigate potentially malicious behavior, and that requires substantial time and effort. "I see a large stream of data and an overwhelmed staff who eventually cannot keep up. Heck, just look at most places that can't even keep up to look at event logs. Now an additional layer of burden is brought upon an already overworked staff," said the engineer.
Indeed, security experts have suggested that the Army's plan "is going to take an army of people to run this stuff, deploy it, analyze the data, and act on it effectively," said Crawford at Enterprise Management Associates. "And that's true, to a point. But big data platforms are designed to handle this--though the analytic techniques may still be playing catch-up."
To be clear, he said that the analytic techniques are good enough, but using them in the big data way that the Army is proposing will require advancing the state of the art. Still, when it comes to crunching big data sets, "these are things that the Hadoop file system and MapReduce are specifically designed for," Crawford said. "They may be able to analyze this in hours instead of days. But is that fast enough?"
Hacktivist and cybercriminal threats concern IT teams most, our first Federal Government Cybersecurity Survey reveals. Here's how they're fighting back. Also in the new, all-digital Top Federal IT Threats issue of InformqtionWeek Government: Why federal efforts to cut IT costs don't go far enough, and how the State Department is enhancing security. (Free registration required.)