Nvidia has launched a new software framework to secure AI infrastructure during runtime. Doca Argus detects threats in real-time without impacting performance and integrates with existing security systems.
On the sidelines of the RSA conference, Nvidia launches a new security solution for AI: Doca Argus. The Doca Argus framework runs on Nvidia BlueField network cards and analyzes AI workloads in real-time. According to Nvidia, the system uses advanced memory forensics and offers up to a thousand times faster detection than traditional agentless solutions.
The software runs completely separate from the host machine, keeping it out of reach of potential attackers, even if criminals manage to take over a system. Thanks to this architecture, Argus works without additional integrations or dependencies on the host, increasing security and efficiency in container-based and multi-tenant environments.
Security professionals can connect Doca Argus to existing SIEM, SOAR, and XDR platforms. This allows organizations to maintain a central overview of threats and respond automatically. The system filters out false alerts and focuses on known threats, helping to reduce alert fatigue. However, the security solution’s focus is exclusively on AI workloads.
Collaboration with Cisco
Cisco is working with Nvidia on a reference architecture for a so-called Secure AI Factory. AI Factory is the name Jensen Huang coined for a system or cluster that runs AI workloads. The architecture aims to help organizations set up their AI infrastructure more securely and easily. According to Cisco, this approach builds in security from the start, rather than adding it afterwards. The company also launched AI security solutions at RSA.
read also
Cisco introduces AI Defense for securing AI applications
Doca Argus is part of Nvidia’s broader cybersecurity platform. It combines BlueField’s data security with Doca Argus and the Morpheus framework from Nvidia AI Enterprise. The combination enables AI systems to autonomously respond to threats. According to Nvidia, this is necessary as companies increasingly use agent-based AI. Nvidia positions the new platform as a way to secure AI environments at scale without sacrificing performance.