Real Time Deepfake Voice Detection Using Machine Learning

Abstract

Deepfake voice technology has rapidly evolved with advancements in artificial intelligence, enabling the generation of highly realistic synthetic speech that mimics a person’s tone, emotion, and speaking style. Although this technology has beneficial applications in entertainment and virtual assistants, it also poses serious threats such as fraud, identity theft, misinformation, and financial scams. This paper presents a Real-Time Deepfake Voice Detection System using Machine Learning techniques to distinguish between genuine human speech and AI-generated synthetic audio. The proposed system uses Mel Frequency Cepstral Coefficients (MFCC) for feature extraction and applies Support Vector Machine (SVM) and Random Forest classifiers for classification. The system includes preprocessing techniques such as noise removal and signal normalization to enhance accuracy. Experimental results demonstrate that the proposed model achieves high detection accuracy while maintaining low latency suitable for real-time applications. The developed system provides a reliable solution for detecting deepfake audio and contributes toward improving digital security and trust in voice-based communication systems.