European academics and researchers are already working on securing 6G communications in anticipation of a widening attack surface, increased interconnectivity, and classical nation-state threats.
Did you know that 6G is already a thing? Word on the street is that it might get rolled out globally around 2030. In anticipation of that near future, 19 organizations have signed onto the “Shield-6G” project — a European Union (EU)-funded venture to develop 6G cybersecurity. The ultimate goal is to develop a cyber threat intelligence platform across network operators, securing the future of mobile communications by the time it arrives.
“6G is way more complex than 5G, because it manages more devices, and there’s more automation, and with automation there come problems,” says Bart Siniarski, director at MBP Network Technology, a Shield-6G member organization. “The attack surface will be extended by a couple of magnitudes. So when we step into 6G — which is, to me, 5G on AI steroids — there are going to be problems at the beginning. And then over time [the goal is that] we are comfortable using 6G in critical infrastructure like hospitals or factories or maybe in shipping and militaries.”
What’s Going to Change with 6G
Siniarksi points to the Internet of Things (IoT) and the industrial IT as drivers for this next generation of wireless networking. “There’s going to be a huge number of interconnected devices [in the near future] — like cars that need to talk to each other, need to talk to us, need to talk to the roadside infrastructure, plus smart factories, smart homes, everything smart-plus-whatever,” he explains. Some governments and industries also anticipate a significant rise in embodied AI in the 2030s, plus other new categories of computation.
6G, then, isn’t so much about streaming Netflix faster, or in higher definition. It’s about managing a wave of new connected devices, the many ways in which they interconnect, the computational load from widespread artificial intelligence (AI) adoption, and the new demands of emerging technologies, like remote surgeries that require ultra-low latency and zero data loss or corruption. To account for all that, besides improving on capacity and latency, 6G is being conceived of as AI-centric.
In short, Siniarski says, “When I explain 6G to my colleagues and friends, I basically say: It’s 5G with AI on top.”
What 6G Security Looks Like
Shield-6G will incorporate some traditional security measures, like an extensive use of honeypots to trap and study threat actors. The focus of 6G, though, will be on a few key areas of emerging AI security, like AI-driven threat detection and response, and testing AI security controls in digital twin environments.
One major hurdle will be the sheer degree of fragmentation among 6G providers. “[Mobile] networks are quite fragmented, so the security systems for those networks are very fragmented,” Siniarski explains. “The moment I send a picture on my phone, it has to travel through a number of base stations, a number of servers, a number of databases, perhaps some third-party software. And that is very problematic because our data, first of all, can leak at any stage. You want to have the whole picture of how this data travels from A to B. And we’re worried about the fact that the security systems watch very narrowly sliced parts of the network.”
Besides providing even security across many stakeholders, Shield-6G has a dual aim of maintaining privacy in data processing. Traditionally, AI models collate and train on data indiscriminately, no matter how many sources the data comes from. In this case, Siniarski asks, “If you want to train a model using sensitive data — which is something you need to do from time to time [in 6G] — then how do you do that collaboratively without exposing this data? The answer is federated learning.”
In a federated model, different parties to an AI system feed training data to an AI in total isolation. “Essentially, I train my piece, you train your piece. Partner number three trains their piece, and then we put together one big model without seeing each other’s data,” Siniarski explains, calling federated learning the “centerpiece” of Shield-6G.
Lastly, he highlights the project’s emphasis on building explainable AI systems. “When we flag something as a security threat or a potential attack, we want to have reasonable explanations for why the model has made such a decision to, for example, block a particular connection,” Siniarski says. “Perhaps it’s just a spike in the use of a particular website. Or maybe it’s an abnormal number of connections, but it’s not poisonous in any way. So we have to make those security decisions easy to understand by humans.”
Siniarski acknowledges that some folks might be reticent about just how deeply and broadly AI will be integrated both into 6G networks themselves, and the defenses built on top. “I think people do not have to worry as much as it sounds [like they otherwise should], because there’s going to be humans involved, and we advocate for this so much. We always try to put a human in the loop so these decisions can be informed.”
Comments are closed