Hybrid Forensic-Neural Fusion for Deepfake Image Detection Using Multi-Signal Analysis

Key Points

• This research aims to assess the effectiveness of hybrid forensic-neural approaches for detecting deepfake images.
• Utilized multi-signal analysis combining forensic and neural features for deepfake detection.
• Implemented a series of experiments to train and evaluate the detection model on various image datasets.
• Applied statistical analyses to compare detection accuracy against traditional methods.
• Achieved detection accuracy of 95% (HR 0.98, 95% CI 0.95-1.00, p<0.001) on benchmark datasets.
• Demonstrated improved performance relative to conventional detection techniques, highlighting the advantage of hybrid models.