EchoShield: A Deepfake Audio Detection System Using BI-LSTM

Deepfake audio generated using advanced artificial intelligence models has become a growing concern in areas such as cybersecurity, misinformation, financial fraud, and identity impersonation. Detecting artificially generated speech is therefore an important research challenge. This paper presents EchoShield, a deepfake audio detection system based on a Bidirectional Long Short-Term Memory (BI-LSTM) neural network. The system processes speech signals through audio preprocessing and Mel-Frequency Cepstral Coefficient (MFCC) feature extraction to capture relevant acoustic patterns. These features are then analyzed by the BI-LSTM model to distinguish between authentic and AI-generated speech. The proposed framework includes a complete pipeline consisting of audio preprocessing, feature extraction, model training, and real-time prediction. The model is trained and evaluated on publicly available deepfake audio datasets. In addition, a Flask-based web interface is implemented to allow users to upload audio files and receive real-time authenticity predictions. EchoShield demonstrates the potential of deep learning techniques for improving audio forensics and strengthening defenses against deepfake-based attacks. The system can support applications in digital security, media verification, and automated content moderation.