Nvidia leverages its GPUs to spot network attacks

Nvidia's AI-powered technology detects unusual behavior and associates it with users to prevent insider attacks and protect digital credentials.Californian Nvidia announced yesterday that its latest security offering is now available in a test environment. In this digital lab, users can try out AI-powered technology (Morpheus) designed by the Santa Clara provider to monitor individual user accounts and spot potentially dangerous behavior. According to the company, the idea is to harness the large amounts of data that most companies compile about login and data access events on their systems, and use it to train an AI with the ability monitor user accounts and identify those that deviate from their usual practices. According to Nvidia, the system would be able to identify true "high-risk" events and could save security teams from combing through millions of events each week. This technology is nothing new, Darktrace has been offering a similar solution for several years now. Nvidia simply exploits a little more the computing resources of its accelerators to highlight security with digital simulation.A "LEGO kit" of the digital footprintIn an official blog post, Nvidia said its solution is a kind of "Lego kit" of the digital footprint, in the sense that it allows users to customize the solution to best suit their needs. Justin Boitano, Nvidia's vice president for enterprise and edge computing, said in a meeting that the system can both reduce workloads and achieve faster identification of potential wrongdoers. “This fingerprinting API workflow allows security companies to detect threats instantly,” he said. “This means that identity attacks, which are the most common form of corporate security attacks where bad actors seek to steal employee credentials via social engineering, or phishing attempts, can be quickly mitigated and detected instantly,” he added.According to Nvidia, this AI-backed security solution is just the first in a long series. “In the future, more granular user profiles, which may include data as granular as typing speed and accuracy, could be used to identify suspicious behavior,” the company said in its blog post. . “Network event data is an extremely valuable resource for building AI models that harden networks, but no one wants to share detailed information about real users and break-ins,” the vendor wrote. "Synthetic data, generated by a variant of the digital fingerprint, could fill this gap, letting users create what they need to meet their use case." The fingerprinting system is currently available on Nvidia's AI Enterprise platform, and the test module - which allows users to explore a real deployment on a sample architecture hosted at Equinix and interact with vendor experts - has been available through the company's LaunchPad website since yesterday.