Infinidat embeds cyber defenses into storage as AI expands the attack surface

The ability to recover critical information following a cyberattack forms the foundation of cyber resilience planning. As companies roll out agentic AI systems, the scope of what requires protection is growing significantly.

This widening footprint is forcing enterprises to rethink their approach to AI security and resilience. Large language models, smaller language models and the supporting software stack all present new vulnerabilities that attackers can exploit, according to Eric Herzog, chief marketing officer at Infinidat Ltd., a Lenovo company.

We would argue that … anything to do with AI dramatically expands the attack surface. Everything you're doing — the [large language models], the [small language models], that AI software stack you deploy — the cyber criminals are going to go after all of that. Instead of shrinking the attack surface, you've actually increased it.

Eric Herzog, Infinidat

Herzog shared his perspective during an exclusive conversation with theCUBE Research's Christophe Bertrand at the AI Cybersecurity and Resilience Summit, broadcast on theCUBE, SiliconANGLE Media's livestreaming platform. The discussion centered on recovery strategies and the role storage infrastructure should play in protecting AI systems.

Storage as a security layer

Infinidat's response to this challenge is InfiniSafe, a suite of security functions built directly into the company's InfuzeOS storage operating system. The offering provides immutable snapshots, logical air-gapping that isolates the management plane from the data plane, and a quarantined forensic space for analyzing potentially compromised data. Since roughly 90% of the most critical information at Fortune 500 enterprises resides on storage systems, this layer represents a natural place to establish protective measures, Herzog noted.

You tell it what you want to do, and the storage then responds to this data center-wide cybersecurity software or the SOC and automatically starts taking action. You don't want to waste a minute if you think you're having a cyberattack.

Eric Herzog, Infinidat

Detection capabilities follow next. The InfiniSafe Cyber Detection component leverages AI and machine learning to examine block-level data, file-level data and VMware data stores for unusual patterns, assigning each snapshot a risk rating of high, medium or low. Some organizations run detection scans on daily snapshots twice weekly, while others employ the tool only when they suspect a breach until locating a verified clean copy, Herzog explained. Infinidat guarantees recovery-time objectives of one minute or less for immutable snapshots on its InfiniBox platforms.

That guarantee is an RTO of 1 minute or less no matter how big the dataset size. Last July, we did one of our Experts Live events … If you were on that event, you would have seen us recover 9.8 petabytes in 5 seconds.

Eric Herzog, Infinidat

Threat actors frequently target backup copies and replicated data, which is why Infinidat partners with data protection vendors including Veeam Software Group GmbH and Commvault Systems Inc., Herzog explained. He cited PwC's most recent CEO survey, which identified cybersecurity as the top business risk, as evidence that securing board-level commitment requires demonstrating business impact.

https://www.youtube.com/embed/FLHmNpAtwlM?feature=oembed

What they should do is practically deploy enterprise-wide cyber storage, integrate it with the SOC or the [security incident and event management] or [security orchestration, automation and response] technology so it works together as a team. I would argue that from a big enterprise perspective, a cyberattack is a disaster.

Eric Herzog, Infinidat