Talk Title:

Deep Learning-based Denoising of TEMPEST Images for Efficient Optical Character Recognition

Abstract:

The present work shows the application of deep learning models to the denoising of video frames retrieved from electromagnetic emanations from remote video interfaces. It has been demonstrated that the cables of video interfaces like VGA or HDMI, produce unintended emanations and that these emanations can be received and processed to reconstruct the video frames displayed on the external monitor. However, the reconstructed frames are noisy, making it difficult to recover any useful information. By applying deep learning models to denoise, deblur, and interpret the images, information can be interpreted.

Speaker Bio:

Santiago is a Senior RF and Electronics Engineer at the Directed Energy Research Centre from the Technology and Innovation Institute, in Abu Dhabi, UAE, where he researches about trend technological topics related to the applicability of computer science methods to radiofrequency systems. His research areas are the application of machine learning techniques to the detection, classification, and identification of RF emanations, analysis of security weaknesses due to RF unintended emanations, and application of computational science algorithms as an innovative way with which some RF aspects can be analyzed. He obtained his master's degree in Computer Science at the National University of Colombia in 2020.

Juan Galvis is a Senior Systems Integration Engineer at the Directed Energy Research Centre from the Technology and Innovation Institute, in Abu Dhabi, UAE. He holds an honors degree in Mechatronics from the Universidad Nacional de Colombia in Bogotá, Colombia. His research topics include Autonomous Systems, Deep Learning applied to computer vision and signal processing, Convolutional Neural Networks, Reinforcement Learning and Embedded Systems.