A Practical Methodology for ML-Based EM Side Channel Disassemblers

TL;DR

This paper presents a practical machine learning-based methodology for electromagnetic side channel disassemblers, achieving high accuracy in instruction recognition on embedded devices like Arduino UNO, enhancing security analysis capabilities.

Contribution

It introduces a new methodology combining electromagnetic trace collection and random forest models for instruction disassembly on embedded devices.

Findings

Achieved 88.69% instruction recognition accuracy on Arduino UNO

Improved over previous methods with 75.6% accuracy for more instructions

Demonstrated effectiveness of ML in EM side channel analysis

Abstract

Providing security guarantees for embedded devices with limited interface capabilities is an increasingly crucial task. Although these devices don't have traditional interfaces, they still generate unintentional electromagnetic signals that correlate with the instructions being executed. By collecting these traces using our methodology and leveraging a random forest algorithm to develop a machine learning model, we built an EM side channel based instruction level disassembler. The disassembler was tested on an Arduino UNO board, yielding an accuracy of 88.69% instruction recognition for traces from twelve instructions captured at a single location in the device; this is an improvement compared to the 75.6% (for twenty instructions) reported in previous similar work.