Use of LSTM Machine Learning Integrated with Explainable AI (x-AI) in Face Deep-fake Detection

This paper presents a novel approach for detecting face deepfakes using Long Short-Term Memory (LSTM) networks integrated with Explainable AI (x-AI) techniques. With the rise of synthetic media generated by AI, particularly deepfake videos and images, there is an urgent need for robust detection systems. Our work utilizes the temporal capabilities of LSTM networks to learn temporal facial patterns, while incorporating explainable methods such as SHAP and LIME to offer transparency in decision-making. We evaluate our system on the DeepFake Detection Challenge (DFDC) dataset, showing significant improvement in detection accuracy and interpretability compared to baseline models.