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.
Make sure you check that bot’s work!
When it comes to security, organizations are embracing AI’s ability to expand their capabilities—but the human element is still important.
Ganesh Pai, founder and CEO of security visibility company Uptycs, told IT Brew at this year’s RSA Conference that for all the integration of AI into security systems, the technology still needs the human touch.
“Traditionally, if you train a machine with a picture of 1,000 cats and the next picture I show you, [you ask it] is this a cat or not? That’s more straightforward,” Pai said. “But in security, there isn’t such a thing to say, ‘Here [are] 1,000 things and I’ll show you the next one—is it similar?’”
Caution ahead. There’s an inherent risk when you operationalize security controls, Pai added. Models can leak data, or be trained incorrectly—there are many areas where it’s unclear how to responsibly leverage AI.
Security benefits of AI are easy to assess. The technology allows for expanded anomaly detection, making finding that needle in the haystack a breeze. Augmentation of generative AI on cybersecurity, though, is a bit more difficult, he noted.
“The fact that you can augment what the model predicts by looking at additional context probably is a path forward which will provide good value,” Pai said. “Now, the other challenge…is that reducing false positives and doing detections with high fidelity is harder.”
Options everywhere. As IT Brew reported in April, security professionals are enthusiastic about implementing AI in their systems, but haven’t quite come to a consensus on what that implementation will look like. Anton Chuvakin, Google Cloud’s security advisor at the office of the CISO, told IT Brew at the time that because security professionals have been working on integrating AI and security for some time, “the knowledge of what use cases are robust is about to emerge.”
For Petros Efstathopoulos, VP of research at the RSA Conference, a prescriptive approach to the different roles of humans and AI systems is essential, especially as these systems get stronger and more mature.
“The systems that have to do with incident detection or putting events together and correlating them to form a broader incident or paint a bigger picture of the attack—they’re pretty mature, and we couldn’t do that as humans,” Efstathopoulos said.