Low-power neuromorphic processing boosts cyberthreat detection

BrainChip Holdings Ltd has integrated its low-power event-based, neuromorphic AI into an innovative cyberthreat intelligence tool that leverages the Akida™ processor to provide protection for WiFi access, home router, small enterprise routers and other network access devices.

Quantum Ventura developed the CyberNeuro-RT (CNRT) technology in partnership with Lockheed Martin’s MFC Division and Pennsylvania State University under partial funding from the U.S. Department of Energy. BrainChip supplies at-the-edge neuromorphic processing to facilitate on-chip learning for deployment of network-specific attack cyberthreat detection. Akida’s small form factor provides magnitudes less power consumption than a GPU, overcoming form factor and power limitations of internet-connected devices that otherwise would be unprotected.

“CyberNeuro-RT is the only game in town for implementing managed cybersecurity support of edge devices that cannot rely on a central server to identify threats and attacks due to cost or power issues,” said Srini Vasan, President and CEO of Quantum Ventura. “Having the neuromorphic capabilities that BrainChip provides directly integrated into CNRT better allows for the detection of threats across multiple devices that otherwise would be vulnerable to exploitation.”

The implementation of Akida into CNRT is a direct result of a previously announced partnership between BrainChip and Quantum Ventura to develop state-of-the-art cyberthreat-detection tools for the U.S. Department of Energy under the Small Business Innovation Research (SBIR) Program.

The Akida neural processor and AI IP can find unknown repeating patterns in vast amounts of noisy data, which is an asset in cyberthreat detection. Once Akida learns what normal network traffic patterns look like, it can detect malware, attack signatures, and other types of malicious activity. Since Akida can learn on-device in a secure fashion, without need for cloud retraining, it can quickly learn new attack patterns, enabling it to easily adapt to emerging cyberthreat scenarios.

BrainChip IP supports incremental learning, on-chip learning, and high-speed inference with unsurpassed performance in micro-watt to milli-watt power budgets, ideal for advanced AI/ML devices such as intelligent sensors, medical devices, high-end video-object detection, and ADAS/autonomous systems. Akida is an event-based technology that is inherently lower power than conventional neural network accelerators, providing energy efficiency with high performance for partners to deliver AI previously not possible on even battery-operated or fan-less embedded, edge devices.

“In today’s always-connected, always-on world, there is an increasing need for cybersecurity solutions that can thwart attacks through otherwise unsecure devices connected to the network,” said Sean Hehir, CEO of BrainChip. “Our work with Quantum Ventura to integrate cyberthreat detection using AI on a neuromorphic platform provides high-quality protection against cybersecurity threats. Akida’s on-chip learning can adapt to new threats and redirects unknown threats to the cloud, providing faster and more cost-efficient analysis capabilities than otherwise possible.”