Next‐Gen Deepfake Detection: ResNet ‐Swish‐ BiLSTM Model to Deliver Superior Accuracy in Visual Forensics
Mithun Palani, Anitha Perumalsamy · Security and Privacy · 2025
ABSTRACT The rise of deepfakes and propaganda media, characterized by both facial and voice manipulations, poses a significant threat to society by spreading misinformation. This study proposes an advanced AI‐based model for detecting such deceptive content, leveraging deep learning (DL) techniques. The model integrates a hybrid architecture combining ResNet, Swish activation, and BiLSTM networks to perform robust and accurate analysis of deepfake attributes, including voice alterations, face swapping, emotion manipulation, and lip‐syncing. Using benchmark datasets such as FaceForensics++ (FF++) and the deepfake detection challenge (DFDC), the model achieves an impressive 96% accuracy on FF++ and 78% accuracy on combined FF++ and DFDC datasets. The results demonstrate the model's superior performance compared to existing approaches, highlighting its potential in combating deepfake‐driven misinformation on social media platforms.