Analysis of Information Leakage from MCU using Neural Network

Abstract

Electromagnetic leakage from an operating microcontroller unit (MCU) can be intercepted and analyzed to deduce the instructions being processed. This is helpful to understanding eavesdropping and to later protecting against it. In this work, harnessing a deep neural network, we analyze the massive electromagnetic (EM) leakage information to extract the instructions run in a microcontroller unit (MCU). An EM leakage acquisition environment is built, and the leakage signal of an MCU is collected. A multi-layer convolutional neural network is constructed for the side channel analysis and identification. The recognition accuracy is over 95% in the training phase, and more than 75% in the prediction phase. This experiment proves that the electromagnetic leakage emitted can be collected by appropriate methods and analyzed by deep learning technology. It can effectively deduce the instructions running in an MCU, which lays a foundation for the protection against eavesdropping.