AI Based – Deepfake Voice Detection System

Abstract

The rise of artificial intelligence has enabled the creation of deepfake technologies that can generate highly realistic synthetic voices. While these technologies have positive applications in entertainment and accessibility, they also pose serious threats such as impersonation, misinformation, and voice-based fraud. Detecting fake or manipulated audio has therefore become an essential area of research in artificial intelligence and cybersecurity. This project presents a general fake audio detection system capable of identifying not only AI-generated deepfake voices but also manipulated real recordings such as splicing, pitch modification, and voice cloning. Unlike existing models that are limited to specific languages or static audio files, the proposed system supports multilingual audio, including major Indian languages, and is designed to work in real-time, even during live calls.