Early Risk Prediction in Real Calls
arXiv · 2026
Predicting requests for sensitive information from a call's first lines.
Problem
Can the opening lines of a real inbound call predict that the caller will later ask for sensitive information?
My contribution
- Analyzed 10,211 real inbound calls and 330,956 transcribed turns to predict later sensitive-information requests.
Results
- 0.87 ROC-AUC of a bag-of-words baseline from the first 8 caller lines, on 1,276 calls held out by phone number
Paper
Anatomy of a Scam Call: What 10,000 real scam and spam calls reveal about how phone scammers operate
arXiv:2608.24127 (2026)
* Equal contribution; listed in no particular order. † Corresponding author.
@misc{traister2026anatomyscamcall10000,
title={Anatomy of a Scam Call: What 10,000 real scam and spam calls reveal about how phone scammers operate},
author={Ethan Traister and Ankit Raj and Jiaqi Gan and Xingyu Shen and Tyler Wu and Yuchen Zhou and Tommy Duong and Kidus Zewde and Siying Chen and Simiao Ren},
year={2026},
eprint={2608.24127},
archivePrefix={arXiv},
primaryClass={cs.CR},
url={https://arxiv.org/abs/2608.24127},
}