What Our Machines Dont Know

Asimov's laws of robotics are a warning to IT security practitioners.

My middle-school sons both want to see the movie, "I, Robot," when it debuts next month. Fortunately, theres time enough between the last day of school and the first day of the movies opening to enforce our family rule: If you want to see the movie, you have to read the book first. And every technically literate person ought to read "I, Robot"—if only for the insights it offers into debugging the complex behaviors that can arise from just a few simple rules.

I thought of this and other robot stories of Isaac Asimov when I saw this mornings story in USA Today about Lancope Inc.s oddly named "StealthWatch+Therminator" intrusion detection system. "Therminator," it turns out, comes from the products use of algorithms drawn from thermodynamics—and first developed by a mathematician at the National Security Agency—to identify unusual behaviors that depart from "network equilibrium."

Whats the connection with Asimovs Laws of Robotics? The answer is not in the collected stories of "I, Robot," but in one of Asimovs robot novels, "The Naked Sun." Investigating a murder, Asimovs fictional detective Elijah Baley declares that "The First Law of Robotics has been deliberately misquoted": It should not be said, he warns his listener, that "A robot may not harm a human being." Rather, he asserts that the law should be stated as "A robot may do nothing that, to its knowledge, will harm a human being." The difference is one that allows a robot to be used, for example, to administer poison to a person: The person who gives the deceptive orders to the robot is the murderer, but the supposed safety of all interactions between robots and people is nonetheless compromised.

Its all very well, then, for Lancope to greet me at breakfast this morning with its joint announcement that Check Point Software Technologies Ltd. has certified StealthWatch for automatic, integrated operation with Check Points VPN/firewall systems. Reducing the time between behavior and response is critical to the pursuit of zero-day protection of enterprise systems. I actually wrote "critical to assuring zero-day protection" in the preceding sentence, then changed the verb to what you just read, because "assure" is not a word that ought to be used in this context. That is, in fact, my whole point here.

A network that protects its users from every net-borne threat, all the time, is just as elusive a goal as a robot that can do no harm to human beings. Its a good sign that security practitioners are exploring other domains, like thermodynamics or biology, for useful insights and algorithms. Its good news that cyber-crime losses are falling for the third straight year, as estimated by the Computer Security Institute and the FBI.

But as the late Isaac Asimov warned us almost 50 years ago, we cant ever expect machines to be as devious as people.

Tell me what devious people have done to you lately at peter_coffee@ziffdavis.com.