Much of the concern over the hacking of autonomous vehicles has centered on the development of passenger cars. But it's just as likely to occur in an industrial setting. It's not a paranoid fantasy to imagine a rogue actor gaining control of a robotic arm or vehicle within the plant or warehouse, causing severe damage to products or even human workers. Even more likely is the possibility of countless "smart" devices being hacked for proprietary information or commercial sabotage. Welcome to the Internet of (Vulnerable) Things.
The fact that such an incident has yet to happen doesn’t mean it never will, says Jeff Christensen, vice president of product with Seegrid, the maker of self-driving vision-guided vehicles. He says cyberthieves could potentially use their power to wreak havoc in the plant, or steal product within distribution centers.
The most serious threat is, of course, to the human worker. “At the vehicle level, first and foremost, safety is always the number-one driver for every decision that you make,” Christensen says.
One vital preventive measure is to make sure that the control systems for autonomous equipment are isolated from any network that can be hacked, especially from remote locations. Safety devices should feature built-in redundancy and be hard-wired to the braking system. “They don’t even go through the software that navigates the truck,” Christensen says. “If somebody walks in front of the truck, it will stop whether the software is compromised or not.”
The next step is to design the autonomous vehicle so that it lacks a “joystick mode” by which it could be remotely piloted. The navigation system should be installed on the vehicle itself. The truck must have the capability to validate any commands that it’s given, in order to ensure that they fall within the scope of its operation. The actions required to carry out a command need to be facilitated solely by onboard systems.
Here’s one variant of modern-day networking that shouldn’t necessarily be placed in the cloud. From an applications perspective, control systems should generally remain within the local network firewall, and housed on-premise. Says Christensen: “Network security for the facility should be leveraged for industrial automation systems just like everything else.”
Industrial robots should be equipped with similar onboard control systems. Safety features for devices such as manipulator arms and spot welders should be separated from the device’s execution software. In the case of caged robots, a physical fence serves as the safety barrier. But with the increasing deployment of robots alongside human workers, designers must take extra care to build units that incorporate robust fail-safe features.
The application of drones in manufacturing and distribution environments raises yet another safety and security threat, Christensen notes. Merely cutting the motor in response to an attempted hack is no solution; the device could cause serious injury to workers when it crashes to the ground. Commercial drones need to be equipped with an automatic “fly home” mechanism that kicks in when they travel past their range of remote control, or become unsure of what they’re supposed to be doing.
The easiest access for hackers targeting supply-chain operations arises from the surge in mobile and connected devices. They are growing “exponentially,” says Christensen, and operators must be aware of the high potential for infiltration. “Every one of those is a new door or window in your house,” he says.
When it comes to connected devices, the likelihood of penetration by hackers is virtually 100 percent. As security experts are fond of saying these days, it’s not a matter of whether; it’s a matter of when. All devices must be designed with the same fallback level of protection, Christensen says, so that any compromised unit will be immediately disabled and become incapable of sending malicious messages to autonomous robots or vehicles.
Connected devices are, of course, a necessity in today’s industrial world, a reality that requires operators to fight potential hackers both offensively and defensively. Offensive means tightening down the firewall and monitoring all devices on a constant basis. Defensive means having in place effective measures to deploy when the inevitable attack occurs.
“Eventually companies will miss,” says Christensen. “So we need defensive measures. Every device needs to be verifying its own command before executing.”
Next: Cybersecurity and machine learning.