Abstract: Deepfake technology has grown significantly in recent years, posing serious challenges in digital security and misinformation. This research focuses on detecting deepfake audio using machine learning techniques by extracting key audio features such as Mel-Frequency Cepstral Coefficients (MFCCs), mel spectrograms, chroma features, zero-crossing rates, spectral centroid, and spectral flatness. A Flask-based web application is developed for real-time deepfake detection, allowing users to upload files and receive instant classification results. Our methodology involves data preprocessing, feature extraction, and similarity-based classification. The system demonstrates high accuracy in distinguishing real from fake audio, providing a valuable tool doe media forensics and digital security applications.
Keywords: Deepfake detection, Audio forensics, Feature extraction, Spectral analysis, Digital Security.
|
DOI:
10.17148/IJARCCE.2025.14159
