Machine Learning for Side-Channel Disassembly

Code execution on processors results in side effects such as electromagnetic emanations. Monitoring these effects can provide information about the code execution. However, extracting the information from the collected data is not a straightforward task and the information can be affected by noise and signal collection conditions. Therefore, in this project, we investigate the use of machine learning models for sidechannel disassembly of instructions running on a processor. Building on previous work, we investigate the implementation of a hierarchical model to recognize the instructions and their operands with high accuracy. Our approach provides results close to simulation, highlighting the validity of our proposed approach.