MMPD: A Novel Malicious PDF File Detector for Mobile Robots

Abstract

PDF is one of the most commonly used office tools, and can be used by robots to share documents with people. However, it has become the target of hackers to attack a variety of devices since it is cross-platform and can be embed with JavaScript code or URIs. Therefore, how to protect the platform security while robots parsing PDF files becomes a serious problem. In this paper, we design and realize a novel malicious PDF file detector, MMPD, for mobile robots based on deep learning. Experiments show that the F1-Score of MMPD can achieve up to 99.478%. In the meantime, the hardware resource usage is low and no significant performance reduction of the whole system is observed.