New Algorithm Lets SCADA Devices Detect, Deflect Attacks

Researchers have built a prototype that lets SCADA devices police one another in order to catch and cut off a fellow power plant or factory floor device that has been compromised.

The so-called secure distributed control methodology outfits SCADA systems, such as robots or PLCs, with embedded software that uses a specially created algorithm to detect devices behaving badly. The software, which was developed by researchers at NC State University with funding from the National Science Foundation, detects and then isolates a neighboring device that has been compromised.

It uses a reputation manager for the devices, so that if one robot or PLC starts doing something it's not supposed to do -- or exceeds a certain threshold such as improperly accelerating or slowing its speed -- fellow robots or other devices detect the uncharacteristic behavior, sound an alarm, and cut it off from their operations to minimize or stop any damage.

This peer-level SCADA security would augment existing and emerging SCADA security products and features, the researchers say. The algorithm could be added into existing software in control systems, with some minor coding modifications, according to the researchers.

"Commercial SCADA security uses a police car and travels and monitors the area. Ours is more like a community [neighborhood] watch," says Mo-Yuen Chow, a professor of electrical and computer engineering at NC State and co-author of the research. "Each of the devices watch each other and talk to [their] neighbor."

It's a next-generation security technology for the new normal of assuming you've been breached and focusing on minimizing any damage. "Our [technology] assumes the attack is already [occurring] and the device is already compromised," says Wente Zeng, an NC State Ph.D. student who worked on the prototype. "After that, it [focuses on] how can we still make sure the rest of the system can work well" and uninterrupted, he says.

[Industrial control systems vendors are starting to patch security bugs, but actually installing the fixes can invite more trouble. See The SCADA Patch Problem.]

Each local SCADA device monitors the others so that if one device behaves abnormally, the others shut down its communications, Zeng explains. "So we can isolate the attack from the system."

The researchers ran a simulation with robots outfitted with the embedded software and controller. "If one robot is compromised, it will affect other robots, so some of them would go to the wrong place," Zeng says. "With our code, each is monitoring each other, so if this robot behaves weirdly," it is cut off, he says. "There's a controller on the robot ... and they talk to each other with the simple algorithm."

The researchers say they plan to patent the algorithm and explore commercialization prospects for the technology.

But distinguishing between normal and abnormal behavior isn't always so straightforward, and sophisticated attackers could find ways to taint the information in some way, according to some security experts.

The NC State researchers will present their technical paper, "Convergence and Recovery Analysis of the Secure Distributed Control Methodology for D-NCS" (PDF), later this month at the IEEE International Symposium on Industrial Electronics in Taipei, Taiwan.

Have a comment on this story? Please click "Add Your Comment" below. If you'd like to contact Dark Reading's editors directly, send us a message.