Voice Recognition and Document Classification-Based Data Analysis for Voice Phishing Detection

Abstract

Phishing crime has become a serious issueworldwide. Damagescaused by phishing have been increasing continuously ever since the first phishing attacks occurred. Voice phishing, in which criminals impersonate financial institutions over the telephonein order to damage consumers, account for the majority of such attacks.This study aimed to convert phishing sound source files to text files through voice recognition and to classify and evaluate whether such texts can be judged as voice phishing. From the proposed methodology, it was confirmed that the Doc2Vec embedding method and the similarity determination method performed better than the methods used in previous studies. Through this, the proposed methodology confirmed that voice phishing can be judged by document data that are textualized by voice recognition for voice phishing sound sources.