ABSTRACT
Owing to the spread of the COVID-19 virus, the online meeting system has become popular. From the security point of view, the protection against information leakage is important, as confidential documents are often displayed on a screen to share the information with all participants through the screen sharing function. Some participants may capture their screen to store the displayed documents in their local devices. In this study, we focus on the filtering process and lossy compression applied to the video delivered over an online meeting system, and investigate the identification of screenshot images using deep learning techniques to analyze the distortion caused by such operations. In our experimental results for Zoom applications, we can obtain more than 92.5% classification accuracy even if the captured image is intentionally edited to remove the traces of screen capture. © 2022, IFIP International Federation for Information Processing.