Plagiarism Detection in Texts Obfuscated with Homoglyphs

Homoglyphs can be used for disguising plagiarized text by replacing letters in source texts with visually identical letters from other scripts. Most current plagiarism detection systems are not able to detect plagiarism when text has been obfuscated using homoglyphs. In this work, we present two alternative approaches for detecting plagiarism in homoglyph obfuscated texts. The first approach utilizes the Unicode list of confusables to replace homoglyphs with visually identical letters, while the second approach uses a similarity score computed using normalized hamming distance to match homoglyph obfuscated words with source words. Empirical testing on datasets from PAN-2015 shows that both approaches perform equally well for plagiarism detection in homoglyph obfuscated texts.