TeleAntiFraud-28k introduces a novel audio-text dataset specifically designed to enhance telecom fraud detection through multimodal analysis.

• First open-source dataset combining speech audio with reasoning-oriented text analysis
• Uses privacy-preserving techniques to generate high-quality training data
• Enables more robust automated fraud detection systems through multimodal learning
• Addresses critical security challenges in telecommunications

This research significantly advances cybersecurity capabilities in the telecom sector by providing the foundation for more accurate, reasoning-based fraud detection systems that can analyze both audio and textual patterns in conversations.

TeleAntiFraud-28k: An Audio-Text Slow-Thinking Dataset for Telecom Fraud Detection