SecureTalk is an AI-powered deepfake audio detection system developed to distinguish between genuine and manipulated audio recordings. The project addresses the growing threat of audio deepfakes, which can be used for identity impersonation, fraud, and misinformation.
The system utilizes machine learning techniques to analyze audio recordings and classify them as real or fake. Audio features such as MFCC, Chroma, Spectral Centroid, and RMS are extracted and used to train a Random Forest classifier. The developed model achieved an accuracy of 94%, demonstrating its effectiveness in detecting deepfake audio.
A user-friendly graphical interface was also developed, allowing users to upload audio files and receive instant predictions with confidence scores. The project contributes to enhancing cybersecurity, digital trust, and protection against audio-based fraud.
