As humanoid robots enter the mainstream, security pros flag the risk of botnets on legs

Interview Imagine botnets in physical form and you’ve got a pretty good idea of what could go wrong with the influx of AI-infused humanoid robots expected to integrate into society over the next few decades.

Morgan Stanley recently predicted robot revenue could surpass $5 trillion by 2050, and firms including Unitree Robotics, Agility Robotics, and Engineered Art, along with major carmakers including BMW,  Toyota, Tesla, and Hyundai (which also owns Boston Dynamics), are all developing humanoid robots.

These bipedal machines and their four-legged friends are slated for use on the battlefield and in law enforcement operations, in labs and academia, plus warehouses and manufacturing facilities. As such, building security into these robots “is imperative,” Joseph Rooke, risk insights director at Recorded Future’s Insikt Group, told The Register.

With 3 billion of these robots in use by 2060, according to BofA Global Research’s forecast, the worst-case scenario from a security perspective would be “an I, Robot situation with no coming back,” Rooke said in an interview. “Botnets in a physical form is now.”

In a report published Tuesday, Rooke outlines the potential fallout of not taking these precautionary measures immediately, and predicts a whole new industry dedicated to securing these robots will likely emerge in the next decade.

This isn’t just the stuff of Hollywood or sky-is-falling security vendors either.

Researchers in late September published technical details about a working proof-of-concept hack that exploited multiple security flaws – hardcoded cryptographic keys, trivial authentication bypass, and unsanitized command injection – in Unitree Robotics’ Bluetooth Low Energy (BLE) Wi-Fi configuration interface used by the Chinese company’s humanoid robots.

“What makes this particularly concerning is that it’s completely wormable – infected robots can automatically compromise other robots in BLE range,” wrote Víctor Mayoral-Vilches, Andreas Makris, and Kevin Finisterre. “This vulnerability allows the attacker to completely take over the device.”

Three factors combine to form the “perfect storm” for these embodied AI systems, according to Rooke.

“First: we already use robotics, we’ve been using robotics in factories for a long time, and now we’re developing walking ones,” he said. 

You’ve got robotics, you’ve got AI, and now you’ve got the need. It’s the perfect storm for this to just skyrocket

“The second thing is, we’ve got the means for these embodied things to learn, so their sensors and their hearing devices can now learn the same way that you train an AI in a server box where it’s contained.”

The third factor has to do with the global population decline combined with an aging population, which is expected to lead to labor shortages and many countries turning to technology in general – robots in particular – to boost productivity.

“You’ve got robotics, you’ve got AI, and now you’ve got the need,” Rooke said. “It’s the perfect storm for this to just skyrocket.”

Like with any internet-connected devices, robots are vulnerable to cyberattacks in the form of hijacking, data leaks, and long-term espionage and IP theft, putting not only the users of these machines at risk but also the companies (and their suppliers) that make humanoid robots.

While Recorded Future hasn’t publicly reported any robotics firm breaches, its threat hunters have been tracking RedNovember (Microsoft calls this crew Storm-2077) and other Chinese government-sponsored spy groups that target defense, electronics, and manufacturing companies.

“I don’t have specific evidence of them going after robotics, although absolutely it’s their MO,”  Rooke said. “If your sector is listed in China’s 15th five-year plan, you’re basically being given a heads-up.” 

The most recent five-year plan puts a big emphasis on AI and smart robots and their role in economic growth.

However, as Russia (with limited success thus far) and other nations put more resources into humanoid robot R&D, “espionage makes sense in terms of the dangers that the actual robots themselves face,” according to Rooke.

Along similar lines, if your company supplies critical components to a firm developing advanced robots, consider your firm at risk of being targeted in a larger supply chain, he added.

There’s also potential security risks for consumers as home-helper humanoid robots like Neo hit the market, and to that end Rooke says it’s likely to be the consumer’s responsibility to update the robots’ software, similar to what happens after someone buys a laptop. “It’s my job to update my Mac, or I am liable for problems if I get hacked because I haven’t updated it,” he explained. “The same thing could happen with some of these robots.”

Additionally, he expects to see a new crop of companies specifically designed to secure humanoid robots emerge – so put that on your RSAC bingo card for 2030 and beyond. These machines are essentially IoT devices, Rooke said.

But they’ve also got embedded AI systems. And unlike other big, heavy IoT machines – like a connected refrigerator that tells you when you’re running low on milk – these robots walk on two feet, and come equipped with “eyes” and “ears” that presumably could be co-opted for all manner of nefarious purposes.

Rooke says it’s pretty speculative at this point what humanoid robot security startups will look like, but he has one guess: “IoT on steroids.” ®

READ MORE HERE