Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

"It Warned Me Just at the Right Moment": Exploring LLM-based Real-time Detection of Phone Scams

Zitong Shen, Sineng Yan, Youqian Zhang, Xiapu Luo, Grace Ngai, Eugene Yujun Fu

TL;DR

The paper tackles the problem of real-time detection of phone scams by leveraging large language models to intervene during ongoing calls. It presents a three-component system (Scammer, User, LLM-based Detector) and uses turn-by-turn prompts (binary RT and UNC) to classify or defer judgments, evaluated on real and synthetic Chinese datasets across multiple LLMs. Results show high recall in real-time detection but variable precision, with the UNC prompt improving precision at the cost of recall and timeliness, highlighting important trade-offs for practical deployment. Overall, the work demonstrates the potential of LLM-based real-time interventions to reduce scam harm while underscoring challenges related to false positives, context, and user trust, and outlines avenues for future research on privacy and user-centered notification design.

Abstract

Despite living in the era of the internet, phone-based scams remain one of the most prevalent forms of scams. These scams aim to exploit victims for financial gain, causing both monetary losses and psychological distress. While governments, industries, and academia have actively introduced various countermeasures, scammers also continue to evolve their tactics, making phone scams a persistent threat. To combat these increasingly sophisticated scams, detection technologies must also advance. In this work, we propose a framework for modeling scam calls and introduce an LLM-based real-time detection approach, which assesses fraudulent intent in conversations, further providing immediate warnings to users to mitigate harm. Through experiments, we evaluate the method's performance and analyze key factors influencing its effectiveness. This analysis enables us to refine the method to improve precision while exploring the trade-off between recall and timeliness, paving the way for future directions in this critical area of research.

"It Warned Me Just at the Right Moment": Exploring LLM-based Real-time Detection of Phone Scams

TL;DR

The paper tackles the problem of real-time detection of phone scams by leveraging large language models to intervene during ongoing calls. It presents a three-component system (Scammer, User, LLM-based Detector) and uses turn-by-turn prompts (binary RT and UNC) to classify or defer judgments, evaluated on real and synthetic Chinese datasets across multiple LLMs. Results show high recall in real-time detection but variable precision, with the UNC prompt improving precision at the cost of recall and timeliness, highlighting important trade-offs for practical deployment. Overall, the work demonstrates the potential of LLM-based real-time interventions to reduce scam harm while underscoring challenges related to false positives, context, and user trust, and outlines avenues for future research on privacy and user-centered notification design.

Abstract

Despite living in the era of the internet, phone-based scams remain one of the most prevalent forms of scams. These scams aim to exploit victims for financial gain, causing both monetary losses and psychological distress. While governments, industries, and academia have actively introduced various countermeasures, scammers also continue to evolve their tactics, making phone scams a persistent threat. To combat these increasingly sophisticated scams, detection technologies must also advance. In this work, we propose a framework for modeling scam calls and introduce an LLM-based real-time detection approach, which assesses fraudulent intent in conversations, further providing immediate warnings to users to mitigate harm. Through experiments, we evaluate the method's performance and analyze key factors influencing its effectiveness. This analysis enables us to refine the method to improve precision while exploring the trade-off between recall and timeliness, paving the way for future directions in this critical area of research.

Paper Structure

This paper contains 22 sections, 4 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. SYSTEM MODELING AND DETECTION METHOD
  3. System Model
  4. Scammer
  5. User
  6. LLM-based Detector
  7. Detection Method
  8. Binary Classification Prompt (RT)
  9. Prompt with Uncertain Option (UNC)
  10. DATASETS, LLM MODELS, and METRICS
  11. Datasets
  12. Authentic Dataset (Real)
  13. Synthetic Dataset (Syn.)
  14. LLM Models
  15. Metrics
  16. ...and 7 more sections

Figures (4)

  • Figure 1: Taxonomy of Phone Scam Prevention Methods: A Temporal Perspective.
  • Figure 2: The system consists of three components, including "Scammer", "User", and "LLM-based Detector".
  • Figure 3: Analysis of false positive scenarios in fraud detection systems
  • Figure 4: Visualization of fraud detection scenarios in an ongoing normal phone call and a scam phone call. Suspicious patterns are highlighted.