Thirteen years ago, a University of Maryland study set up weak security on four Linux computers with Internet access and recorded what happened. The computers suffered cyberattacks every 39 seconds on average. That figure is still quoted today, although the frequency is surely much higher with the spread of brute force automated hacks. Since COVID-19 alone the FBI claims a 300% increase in reported cybercrimes.
Such a torrent of attack data would overwhelm any human brain, but it offers a nourishing diet for machine learning systems. They demand massive quantities of data to search for patterns, warning signs and likely consequences. Today’s cyberattackers have an ever-growing armoury of weapons and disguises, but they will increasingly be up against AI-informed defence forces that can move and learn as fast as they attack.
As so often happens in real life, the big step forward brings a whole new set of challenges. Historic data, once slumbering in archives, becomes a potential goldmine. The oil and gas industry is sitting on 50 years of detailed data on seismic activity that could transform earth science. The risk is not so much that the data mountain might be stolen – because its real value lies in its entirety – but the AI models mined from it could be vulnerable, especially at the network edge. What would be the black-market value of a fully developed self-driving car system? Whether hosted on-premises, in the cloud, or through a hybrid approach, it is imperative that customer information, business data and AI models be protected against a broad range of ever more sophisticated cyber-attacks.
Mauricio Sanchez leads the Dell’Oro Group network security research program, and Advanced Research reports on Secure Access Service Edge (SASE). He also has responsibility for coverage of next-generation networking architectures and services models. After a brief overview of leading edge research data on AI and security, he will chair a prestigious panel of industry experts to address emerging and anticipated security threats faced by enterprise CISOs.
Be forewarned. Become forearmed.