An information entropy enhanced CNN-LSTM analysis framework for electromagnetic leakage detection of display device

Video Display Units (VDUs) serve as ubiquitous interfaces between humans and machines; however, their unintended electromagnetic emissions during signal transmission pose significant risks of information leakage. Traditional methods for leakage assessment are inherently subjective, as they rely on expert interpretation or screen content reconstruction. This study systematically examines the temporal and statistical characteristics of electromagnetic leakage signals from VGA cables, employing differential information entropy to quantify the features of these information-bearing emissions. We further propose an attention-based CNN-LSTM model enhanced with information entropy, designed to identify leakage-sensitive electromagnetic signals ("red signals"), without requiring reconstruction of the original screen content. Experiments conducted with real VDUs demonstrate the model's high accuracy and robustness. The proposed framework serves as an effective tool for characterizing electromagnetic signals and assessing leakage risks associated with display devices, offering a data-driven approach to evaluating the electromagnetic security of information technology equipment.