Detecting Malicious PDF Files Using Semi-Supervised Learning Method
Download as PDF
DOI: 10.23977/acsat.2017.1001
Author(s)
Feng Di, Yu Min, Wang Yongjian, Liu Chao, Ma Chunguang
Corresponding Author
Min Yu
ABSTRACT
With the increase in popularity of Portable Document Format (PDF) documents and increasing vulnerability of PDF users, effective detection of malicious PDF documents has become as a more and more significant issue. In this paper, we proposed a way to detect malicious PDF files by using semi-supervise learning method. Compare with previous studies, this method not only improve detection accuracy and generalization ability by combining with three different classifiers, but also effectively utilize the abundant unlabeled PDF files to retrain classifiers and update module by selecting the “useful” files from unlabeled test set.
KEYWORDS
malicious PDF files, malicious JavaScript, semi-supervised learning