Improvement of Electromagnetic Side-Channel Information Measurement Platform

Area of Science:

• Cybersecurity
• Embedded Systems
• Signal Processing

Background:

• Microcontroller (MCU) power-up electromagnetic radiation (EMR) patterns vary with executed instructions, posing a security risk for embedded systems and IoT devices.
• Current EMR pattern recognition accuracy is insufficient for robust security analysis.
• Understanding the relationship between EMR patterns and embedded system security is crucial.

Purpose of the Study:

• To propose and evaluate a novel platform designed to enhance EMR measurement and pattern recognition for embedded systems.
• To improve hardware-software interaction, automation, sampling rates, and reduce alignment issues in EMR data acquisition.
• To validate the platform's effectiveness on both microcontrollers (MCUs) and Field-Programmable Gate Array intellectual properties (FPGA-IPs).

Main Methods:

• Development of an integrated platform for seamless hardware-software interaction and automated EMR data acquisition.
• Implementation of higher sampling rates and precise positional alignment for improved measurement fidelity.
• Testing the platform with a microcontroller (MCU) and an FPGA-IP, utilizing similar neural network (NN) architectures for analysis.

Main Results:

• The new platform demonstrated improved top-1 EMR identification accuracy for the tested MCU compared to previous methods.
• Achieved the first known EMR identification for an FPGA-IP, showcasing the platform's broader applicability.
• The proposed platform facilitates EMR pattern measurement for neural network (NN) analysis across diverse embedded architectures.

Conclusions:

• The developed platform significantly enhances EMR measurement capabilities, leading to higher accuracy in identifying instruction-based EMR patterns.
• The platform's flexibility extends its utility from simple MCUs to complex FPGA-IPs, enabling system-level security verification.
• This research contributes to a deeper understanding of EMR pattern recognition for advancing embedded system security.