Machine Learning Based Audio Analysis to Detect Fraud Call in 4G-5G Wireless Networks

Telephony fraud is a growing global problem, and carriers are struggling to find solutions. Most carriers focus on caller ID verification, but this fails to catch phishing calls that use legitimate phone numbers. To address this, fraudsters are increasingly making calls from abroad using VoIP gateways. However, these international calls are transcoded, which generally reduces call quality. We have developed an echo analysis and machine learning (ML) based algorithm that reveals the true quality of the call, enabling real-time detection of long-distance transcoded calls and alerting users. Our algorithm is based on a comprehensive analysis of typical fraudulent calls. The Machine Learning system can detect foreign VoIP or GoIP calls with 95.8% accuracy. By detecting these calls, we can help users avoid becoming victims of telephony fraud.