A Study on Analysis and Discriminant Model of Cybercrime Data
The Korean Data Analysis Society, Geun-A Kim, Sang-Tae Han · The Korean Data Analysis Society · 2025
In this study, CONCOR analysis and emotional analysis were conducted based on a total of 21,831 IN body texts of knowledge about voice phishing, smishing, romance scams, and body cam phishing among new crime types using TEXTOM, a big data analysis platform. As a result of emotional analysis, voice phishing mainly used words to reassure victims. Smishing has many negative words such as rejection and anger, and through content sharing, behavioral patterns to prevent or be careful of other damage are identified. Unlike the previous two crimes, body cam phishing and romance scams contain many words from emotional vocabulary dictionaries even though there is relatively little data. It is judged that there were many emotional exchanges in the damage process written in the text, and the author suffered the most emotionally. A large number of articles are asking for help, and it seems that they borrowed anonymity because it is difficult to reveal or receive help compared to other crimes. In addition, the performance of the voice phishing discrimination model was compared with CNN, LSTM, and DistilBERT models using the Python program. As a result of the analysis, the LSTM model, which is strong in learning patterns and sequential dependence learning that have different meanings depending on the context, showed the best results.