Automatic plagiarism detection in obfuscated text

Abstract

Plagiarism is a serious problem in education, research, publishing and other fields. Automatic plagiarism detection systems are crucial for ensuring the integrity and genuineness of intellectual work. There are different types of plagiarism, such as copy–paste, obfuscation and translation. In particular, obfuscated text is one of the hardest types of plagiarism to detect. In this paper, we propose an automatic plagiarism detection system for obfuscated text based on a support vector machine classifier that exploits a set of lexical, syntactic and semantic features. We evaluated the performance of the proposed system on benchmark English and Arabic corpora made available by the PAN Workshop series: PAN 2012, PAN 2013, PAN 2014 and PAN@FIRE2015. We also compared the performance of our system to the performances of other systems that participated in the PAN competitions. The obtained results show that our system had the best performance in terms of the F-measure on the PAN 2012 and on the PAN@FIRE2015 obfuscated sub-corpora, was among the top four on the PAN 2013 corpus and was among the top two on the PAN 2014 corpus.