Using Neural Networks To Beat Hackers
Psynapse's Checkmate intrusion protection system conducts a real-time assessment of each visitor to a network, and if it sees behavior that indicates an attempted security breach, automatically terminates the intruder's access.
That's a big improvement over current monitoring systems, which generally only send alerts to network administrators in case of trouble, according to Psynapse's founder and CEO, Gary Jackson.
Those systems aren't accurate enough to use to automatically block suspicious visitors, says Jackson. The high number of false alarms means network managers can't rely on them to block hackers automatically without risking cutting off legitimate users.
Checkmate is different from existing network monitoring systems, which are generally rule-based systems or signature detection systems, according to Jackson. Those technologies are limited because they rely on recognition of known attack patterns and continual database updates, he says.
To create Checkmate, Jackson turned instead to the behavioral sciences, and based the system on a neural network. Using a process similar to that used to develop psychological assessments, Psynapse turned to computer security experts to identify the patterns of behavior that administrators should be concerned about.
"Once Checkmate has been trained," Jackson says, "it can generalize to situations it's never seen before."
According to the company, testing shows that Checkmate is as accurate as human experts in detecting intrusions -- and much faster. Compared to human experts, Checkmate "runs at blistering speeds, which means it can assess network users' intent in real time," says Jackson.
Checkmate, which is slated to become available in January, will be sold as a ready-to-use appliance, according to the company. Pricing is expected to start at about $30,000 per unit.