• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.34 no.3
• /
• pp.439-449
• /
• 2024

Recent rapid advancements in voice generation technology have enabled the natural synthesis of voices using text alone. However, this progress has led to an increase in malicious activities, such as voice phishing (voishing), where generated voices are exploited for criminal purposes. Numerous models have been developed to detect the presence of synthesized voices, typically by extracting features from the voice and using these features to determine the likelihood of voice generation.This paper proposes a new model for extracting voice features to address misuse cases arising from generated voices. It utilizes a deep learning-based audio codec model and the pre-trained natural language processing model BERT to extract novel voice features. To assess the suitability of the proposed voice feature extraction model for voice detection, four generated voice detection models were created using the extracted features, and performance evaluations were conducted. For performance comparison, three voice detection models based on Deepfeature proposed in previous studies were evaluated against other models in terms of accuracy and EER. The model proposed in this paper achieved an accuracy of 88.08%and a low EER of 11.79%, outperforming the existing models. These results confirm that the voice feature extraction method introduced in this paper can be an effective tool for distinguishing between generated and real voices.

• Information Systems Review
• /
• v.20 no.1
• /
• pp.61-80
• /
• 2018

Individuals often receive smishing campaigns (mobile phishing messages), which they treat as spam. Thus, firms should understand how their customers distinguish their promotion messages from smishing. However, only a few studies examined this important issue. The present study employs the elaboration likelihood model to develop research hypotheses on the relationship between message cue and message credibility. The message cue in this study is classified as content cue, which is found in the content of promotion messages, and as leakage cue, which is found in peripheral information in the message. Leakage cue includes orthography (inclusion of special characters)and an abbreviated link sent by a faithless sender. We also propose that contextualization has a moderating effect on the relationship between content cue and credibility. We conducted a survey experiment to examine the effect of message cues on message credibility in the context of respondents receiving discount coupons through mobile messages. The result of data analysis based on 166 responses suggests that leakage cue had a negative effect on message credibility. A message with defective content cue has a marginally negative effect on message credibility. In particular, defective content cue in a high-contextual message has a strong negative impact on message credibility. This effect was not observed in low-contextual messages. Moreover, message credibility is significantly low regardless of the degree of contextualization if there is a leakage cue in the message. Our findings suggest that mobile promotion messages should be customized for message receivers and should have no leakage cues.