Real-time Korean voice phishing detection based on machine learning approaches

Minyoung Lee, Eunil Park

Abstract: Voice phishing, or vishing, is a phishing phone call in which an attacker lures receivers into providing personal their information. Damage from vishing is a serious problem worldwide and is increasing in frequency. Therefore, this study is aimed at detecting vishing in real time. Owing to the absence of research on spam detection using low-resource languages, we detect vishing in the Korean language using basic machine-learning models. We collected actual vishing damage data and converted the voice files into text to achieve spam detection using natural language processing techniques. The focus is on determining whether vishing can be rapidly detected, rather than model development. Based on the results, we suggest that vishing can be detected in real time and requires only a short training time when using machine learning models.