uitPDFXAI: Malicious PDF files detection with eXplainable Artificial Intelligence

There are many security incidents related to PDF (Portable Document Format) files. Many studies have been proposed to detect malicious PDF files. In this study, we propose uitPDFXAI system to detect malicious PDF files by leveraging the powerful machine learning algorithms XGBoost and AdaBoost. Utilizing the SHapley Additive exPlanations, an explainable Artificial Intelligence, we provide a detailed explanation of the contribution of features to the model’s predictions, enhancing the transparency and interpretability. Our experiments are conducted on the CIC-Evasive-PDFMal2022 dataset. Experimental results show that the proposed system achieves the highest accuracy of 99.55% when using XGBoost. When using AdaBoost the system’s accuracy is 98.90%. The results of this study can be used for future research in the field of malicious PDF file detection.