Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
Over a third of IT leaders are worried their defenses aren’t ready for AI-powered attacks, according to a survey commissioned by password management firm Keeper Security.
The survey found 95% of respondents believe cyberattacks are “more sophisticated than ever” before. Some 35% identified AI-assisted attacks as one of the vectors they feel “ill-equipped to defeat”—along with deepfakes (30%), 5G network exploits (29%), cloud jacking (25%), and fileless attacks (23%).
“More often than not, IT directors and CISOs are not fully aware of the full scale in terms of how AI is being utilized,” Keeper Security CEO and co-founder Darren Guccione told IT Brew. He added that AI-assisted attacks hold the potential to be like “the difference between a screwdriver and a high-speed automated screwdriver, like an electric drill.”
IT services was the department respondents said was most frequently targeted, with 58% saying their IT units had experienced an attack. Financial operations came in second at 37%, while under a third said threat actors had gone after their supply chain management, data analysis and reporting, or research and development functions according to the survey.
Guccione explained that the relative prominence of IT services as a target is related to industry trends like consolidation and outsourcing of those functions to managed services providers (MSPs).
“The managed service providers are now the outsourced IT departments for basically a large portion of the SMB [small and medium business] channel, and also mid-market channel,” Guccione said. “Cyber criminals know [MSPs] have density because they manage remote data backup and recovery, they also are moving more into security best practices.”
“So they’re targets because they have density; they have a high number of end customers, which in turn have a much higher number of end users and points to attack,” he added.
Generative AI has to date largely been used to juice tactics like phishing and targeted impersonation attacks, as it makes it far easier to generate large numbers of customized messages (and minimize the impact of factors like language barriers). However, the large language models that power AI systems could themselves be vulnerable to attacks ranging from self-replicating worms to adversarial manipulation, as well as be used to automate attacks using known vulnerabilities.
Fortunately, Guccione said, AI-assisted attacks largely rely on known vectors, and cybersecurity leaders can rely on tried-and-tested techniques to minimize their impact.
“[Cyber criminals] don’t attack just the IT department anymore,” Guccione said. “They attack any particular endpoint that they can possibly get into, and then they’ll work in permeating at least horizontally and attack throughout an organization.”
“There’s endpoint security, there’s privileged access management, there’s enterprise password management, there’s secrets management for protecting infrastructure, there’s two-factor authentication,” he added. “It requires a multitude of different software technologies and a mindset.”