Nvidia thinks AI is the new Zero Trust frontier

Hologram fingerprint hovering above a smartphone
(Image credit: Shutterstock / Marko Aliaksandr)

Computing giant Nvidia has developed a new AI framework that it hopes will eventually replace passwords and multi-factor authentication as the most common way of securing digital identities in the hybrid workplace.

The framework, which will first become available on January 23, 2023 as part of end-to-end workflow application Nvidia LaunchPad, looks to build on the company’s digital fingerprinting research by using graphical processing units (GPUs) to train deep learning models on user habits.

These habits, such as typing speed, typo frequency, and “what services [users] use and how they use them”, are the next development in Zero Trust network environments, although Nvidia recognizes that cybersecurity is an “ever-evolving digital wall.”

Nvidia digital fingerprinting framework

Nvidia’s plan is to assign a deep learning model for “every account, server, application, and device on the network”, creating a ‘personality’ of sorts for each person or thing with a presence on a company network. IT staff assigned to monitor them can then tweak the models to send alerts for specific kinds of activity.

The Director of Cybersecurity Engineering and R&D at Nvidia, Bartley Richardson, recognises that the framework has a long way to go, given that, when he first pitched the idea to his technology leads in February 2022, he was told that the idea was “crazy” and “computationally impossible”. At the time, GPUs had no role in cybersecurity.

However Daniel Rohrer, Nvidia’s Vice President for Software Product Security, said  at a GTC talk in September 2022 that a tech industry-wide lack of computational resources to train these models and store the data was fuelling Nvidia’s desire to combine cybersecurity and AI into one fingerprinting solution.

Fast forward to October 2022, and Nvidia had its AI-powered digital fingerprinting framework deployed across its global networks on four Nvidia A100 Tensor Core GPUs.

It claims that, with the framework in place, IT teams may have their workloads cut drastically, having to deal with only “8-10 [security] incidents to investigate daily”. 

That’s a drastic reduction, and a robust defence against intrusions, when the 2022 Verizon Data Breach Investigations Report claimed that 40% of all data compromises involved stolen credentials.

Following the Nvidia Launchpad trial, the company will make the AI framework available as part of the digital fingerprinting AI workflow in Nvidia AI Enterprise 3.0

But before then, and even before we know whether using AI to power digital fingerprinting is even feasible long-term, Jason Recla, Nvidia’s Senior Director of Information Security, has some advice for IT admins looking to help create a longstanding tech environment designed to support new, AI-driven solutions.

“Get up to speed on AI. Start investing in AI engineering and data science skills — that’s the biggest thing.”

Luke Hughes
Staff Writer

 Luke Hughes holds the role of Staff Writer at TechRadar Pro, producing news, features and deals content across topics ranging from computing to cloud services, cybersecurity, data privacy and business software.