Protecting the Future
Security is an arms race, and not only the “good guys” like cybersecurity researchers have access to AI agents.
“Hacker activity opens up much more possibilities for attack than for defense, because an attacker can fail 1,000 times and still succeed once,” Stavrou said. “We’re opening Pandora’s box, trying to understand the impact of AI in cybersecurity.”
The goal of MATrEx is to scientifically evaluate the capabilities of red agents, develop new strategies and teach blue agents how to outwit attackers. MATrEx is designed to allow companies to run the programs in-house, so no protected data is leaked and all developed agent strategies remain contained. This is the first training environment that can be tested not only on traditional network-connected servers but also on 5G.
“No 5G system has ever been connected to a cybersecurity testbed with this level of emulation before,” Stavrou said. “5G networks are relatively new and typically confined to large enterprises. This allows us to see what happens when AI agents attack or defend that type of network and connection.”