Automatically Detecting Voice Phishing: A Large Audio Model Approach

Benjamin M. Ampel, Sagar Samtani, Hsinchun Chen
Last updated on Oct 25, 2026
Source Document AIS eLibrary
In Plain Terms

Voice phishing (“vishing”) scams trick people over phone or audio channels and are hard to catch in real time. The authors built VishGPT, an AI system that listens to calls and flags vishing attempts as they happen, trained partly on synthetic data to overcome the shortage of real examples. It detected scams more accurately than existing methods (about 86% accuracy), offering a practical defense for users and security teams.

Watch a short video summary

Artifacts

Article

Related Papers

Cite this

APA Ampel, B. M., Samtani, S., & Chen, H. (2026). Automatically Detecting Voice Phishing: A Large Audio Model Approach. MIS Quarterly, 50(2), 527–556. https://misq.umn.edu/misq/article/50/2/527/3280/Automatically-Detecting-Voice-Phishing-A-Large
BibTeX
@article{ampel2026automatically,
  title={Automatically Detecting Voice Phishing: A Large Audio Model Approach},
  author={Benjamin M. Ampel and Sagar Samtani and Hsinchun Chen},
  journal={MIS Quarterly},
  year={2026},
  doi={10.25300/MISQ/2025/19532}
}
Voice Phishing Vishing Large Audio Models Deep Learning Cybersecurity