Tree structure network: A learning-based deep network for classification of CPU instruction through em signal
Journal
2019 Joint International Symposium on Electromagnetic Compatibility, Sapporo and Asia-Pacific International Symposium on Electromagnetic Compatibility, EMC Sapporo/APEMC 2019
Pages
246-249
Date Issued
2019
Author(s)
Abstract
In this paper, we proposed a learning-based method to find the instruction which CPU is executing. Reverse engineering is an important issue on military, product security analysis, and intellectual property. Our work is a kind of reverse engineering. We can get the information of instruction without direct accessing CPU. While a CPU is running, it would emit EM signal. We collect the EM signal and combine Deep learning model and novel speech processing method to classify it. In our work, firstly, we use Deep learning network based on convolution neural network (CNN). Then, we propose a Tree Structure Network (TSN) and combine label to solve the problems of both data imbalance and Hard Separate Pairs(HSP). Finally, we can classify the EM signal with 61% Top-1 (the strictest) accuracy. To the best of our knowledge, this is the first work that use learning-based method to classify the EM signal.
Type
conference paper