The Last Mile for Edge AI: Securing the Model Supply Chain
(Room Ballroom A)
05 Nov 25
12:15 PM
-
12:40 PM
Tracks:
Embedded AI: AI Processes
As AI models become the brains of our edge devices, from industrial robots to connected vehicles, they create a new supply chain crisis. The traditional security and DevOps practices that protect cloud software often break down during the "last mile": the journey from the build system to the device in the field. How can you trust an AI model controlling a critical system if its components are a black box and its chain of custody is broken just before deployment?
This session tackles this challenge head-on. We will present a practical framework for extending a secure software supply chain to the edge, one specifically designed to handle the unique complexity of AI/ML models. Attendees will learn how to:
Architect a unified developer-to-device pipeline that treats AI models as verifiable, traceable software packages, not just opaque files.
Deconstruct models into their core components (code, data, dependencies) to manage vulnerabilities and link them to a comprehensive SBOM.
Implement automated security gates to scan models for risks before they are deployed to your IoT fleet.
Establish a robust, at-scale process for deploying, monitoring, and rolling back models on heterogeneous edge devices.