Real-Time Deepfake Audio Detection Using Machine Learning and SVM

Abstract

With the increasing prevalence of AI-generated deepfake voices, ensuring the authenticity of speech has become a critical challenge. Our project focuses on developing an AI-driven fake voice detection system using a Support Vector Machine (SVM) model to distinguish between real and synthesized voices. The system extracts key audio features such as Mel-Frequency Cepstral Coefficients (MFCCs), pitch, and spectral properties to analyze voice patterns. These features are then processed using an SVM classifier, which effectively categorizes the input as either genuine or fake based on trained datasets. The proposed solution enhances voice security in applications like banking, virtual assistants, and fraud prevention