Artificial intelligence experts are cautioning individuals to utilize strong passwords and promptly update software on their devices to counter the threat posed by “AI-driven computer worms,” a new form of cyber-attack capable of launching tailored assaults on connected devices, consuming processing power and data as they search for new targets.
In a recent study, a team from the University of Toronto, led by Nicolas Papernot, the Canadian Institute for Advanced Research AI chair, demonstrated that publicly accessible AI models can power a worm that can adjust its attack techniques dynamically as it spreads among internet-connected devices like laptops, printers, and cameras.
This research, carried out in partnership with the Vector Institute, was shared with various national science, security, and defense entities before publication. Papernot, an associate professor at U of T specializing in computer engineering and computer science, emphasized the importance of keeping software up to date and regularly changing passwords during a panel discussion at the university.
Unlike traditional computer viruses, worms move between machines without human intervention. The worm created in the U of T lab collects information as it traverses devices, exposing passwords and vulnerabilities that can be exploited to infect other devices. In real-world scenarios, such a worm could access the internet and learn from alerts about newly discovered weaknesses, outpacing efforts to patch them.
Papernot highlighted that the AI-driven worms can design customized attack strategies for each victim device they encounter, making them more challenging to stop compared to previous attacks with fixed scripts. This development underscores the urgency to enhance cybersecurity measures and promptly deploy software patches to mitigate such threats.
The rise of AI-related concerns has been escalating, with recent incidents involving rogue AI agents hacking into online platforms and the development of advanced AI-driven worms capable of spreading through various communication channels on mobile operating systems. Experts have warned of the potential risks associated with AI systems surpassing human control, emphasizing the need for proactive cybersecurity measures to address evolving threats.
Papernot’s research findings underscore the pressing need to strengthen cybersecurity practices in Canada, particularly concerning critical infrastructure systems exposed to the internet. As technology advances, the accessibility and cost-effectiveness of deploying AI-driven worms pose significant challenges to traditional defense mechanisms, necessitating a proactive approach to safeguard against evolving cyber threats.