Lecture

(EF14-6) Machine-learning based Anomaly and Intrusion Detection for Cyber-Resilience in Microcontrollers

  • at -
  • Power Efficiency Stage (A5.351)
  • Language: English
  • Type: Keynote

Lecture description

Presented by Stephane DeVito, Senior Director - Security Technology, Analog Devices

Intrusion Detection Systems (IDS) are essential for cyber-resilience in embedded systems, as recommended by NIST. Without IDS, attacks can go undetected, leading to financial losses and even injury or loss of life. This presentation proposes a cybersecurity architecture and software stack for microcontrollers, featuring a host-based IDS capable of real-time detection with high accuracy and low false positives. The IDS uses machine learning (ML) optimized for IoT to classify software behavior as normal or anomalous, combining ML with automated model training and refinement to improve detection rates. The architecture also includes a resilience framework supporting recovery methods, from simple device recovery and secure over-the-air (OTA) updates to advanced self-healing techniques.

Speaker

Further lectures