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

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.