Securing Edge AI with Containers: Eurotech’s Approach

Eurotech, a prominent technology company based in Italy, is at the forefront of addressing the challenges of AI security at the edge through innovative software container technology.

Recently, the company received accolades for its AI and learning hardware at the prestigious Embedded World exhibition. Marco Carrer, the Chief Technology Officer at Eurotech, shed light on their groundbreaking achievement with the ReliaCOR 44-11.

“The design of the 44-11 industrial PC was driven by the convergence of IoT and AI, focusing on workload consolidation at the edge,” Carrer explained. “Even when operating at the edge, industrial PCs require effective management, and device management poses a typical IoT challenge,” he added.

The 4411 model is equipped with either an Intel 13th Gen i7 or i9 processor along with an Nvidia GPU. Carrer emphasized the importance of selecting the right device for edge AI applications, highlighting Eurotech's diverse portfolio that includes a fanless embedded version with the Nvidia Orin processor and the Orin AGX for enhanced performance.

One of the standout features of the 4411 is its ability to streamline the transition from concept to full operation without the need for intricate coding. The system supports the loading and testing of multiple models sequentially, while offering efficient monitoring, management, and updating capabilities.

Central to the success of Eurotech's AI box is its robust hardware and software security measures, facilitated by software containers. The company's hardware boasts the ISO 62443-4-2 L2 security certification, making it the first industrial PC to achieve this recognition.

“Our middleware serves as both an agent and an application framework, functioning as an orchestrator for container workloads,” Carrer explained, underscoring the critical role played by security in the overall design.

With Ubuntu Linux support guaranteed for a decade, the AI box stands out for its container framework, leveraging DeepStream—a Gstreamer version by Nvidia—to seamlessly integrate with various cameras. This integration allows for easy configuration of cameras by adjusting parameters within the container.

For applications involving multiple cameras in edge systems, the 4411 excels in optimizing data flow efficiency. By leveraging the MQTT IoT protocol and Nvidia's Triton AI software framework, the system ensures rapid data processing and analysis.

“The container captures images, while the MQTT broker serves as a notification sink. With Nvidia Triton pre-installed, the system determines the appropriate models for tasks such as object detection and defect identification,” Carrer elaborated.

Furthermore, Eurotech prioritizes data security by implementing root file system encryption and secure boot mechanisms to safeguard against unauthorized access. The company is actively exploring advanced encryption techniques, such as kernel RAM encryption, to enhance data protection.

Looking ahead, Eurotech is exploring innovative solutions, such as zero-touch provisioning for AWS, to bolster device security. By leveraging the Trusted Platform Module (TPM) for device identity management, the company aims to fortify its systems against potential threats.

In the realm of AI hardware development, Nvidia's upcoming system on chip, Thor, promises significant performance enhancements through the adoption of the Blackwell GPU architecture. The transition from 8-bit integer mode (INT8) to 4-bit (INT4) is expected to deliver notable performance gains in AI workloads.

As Eurotech continues to push the boundaries of AI security and edge computing, their commitment to innovation and excellence positions them as a key player in the rapidly evolving technology landscape.