This paper proposes a novel blockchain-based deepfake detection method using federated learning (FL) and deep learning models. The approach aims to preserve data source anonymity while improving the accuracy of deepfake detection. The method combines the strengths of SegCaps and convolutional neural networks (CNNs) for enhanced image feature extraction, followed by capsule network (CN) training to improve generalization. A novel data normalization technique is introduced to address data heterogeneity from diverse global sources. Transfer learning (TL) and preprocessing methods are also employed to enhance deep learning performance. The system leverages blockchain and FL to enable collaborative global model training while maintaining data confidentiality. Extensive experiments show that the proposed method achieves a 6.6% average increase in accuracy and a 5.1% improvement in the area under the curve (AUC) metric compared to six benchmark models. The results demonstrate the effectiveness of the proposed solution in countering the proliferation of deepfake content. The method is designed to address the critical challenge of identifying and mitigating deepfake content by collecting diverse datasets, extensive preprocessing, and leveraging deep learning models tailored for multimedia forensics. The system integrates blockchain-based FL for data anonymity, SegCaps, and CNN fusion for robust feature extraction, and addresses data heterogeneity using a proposed normalization technique. The approach ensures adaptability to varied video qualities while preserving data privacy. The BFLDL system is designed to pool data from multiple clients and train a deep learning model cooperatively. The method uses a blockchain-based FL framework to train and distribute a collaborative model, ensuring secure data sharing and efficient feature extraction. The system is evaluated on various deepfake datasets, demonstrating its effectiveness in detecting deepfake videos. The proposed method is a promising solution for advancing deepfake detection, leveraging existing data resources and the power of FL and blockchain technology to address the critical need for media authenticity and security.This paper proposes a novel blockchain-based deepfake detection method using federated learning (FL) and deep learning models. The approach aims to preserve data source anonymity while improving the accuracy of deepfake detection. The method combines the strengths of SegCaps and convolutional neural networks (CNNs) for enhanced image feature extraction, followed by capsule network (CN) training to improve generalization. A novel data normalization technique is introduced to address data heterogeneity from diverse global sources. Transfer learning (TL) and preprocessing methods are also employed to enhance deep learning performance. The system leverages blockchain and FL to enable collaborative global model training while maintaining data confidentiality. Extensive experiments show that the proposed method achieves a 6.6% average increase in accuracy and a 5.1% improvement in the area under the curve (AUC) metric compared to six benchmark models. The results demonstrate the effectiveness of the proposed solution in countering the proliferation of deepfake content. The method is designed to address the critical challenge of identifying and mitigating deepfake content by collecting diverse datasets, extensive preprocessing, and leveraging deep learning models tailored for multimedia forensics. The system integrates blockchain-based FL for data anonymity, SegCaps, and CNN fusion for robust feature extraction, and addresses data heterogeneity using a proposed normalization technique. The approach ensures adaptability to varied video qualities while preserving data privacy. The BFLDL system is designed to pool data from multiple clients and train a deep learning model cooperatively. The method uses a blockchain-based FL framework to train and distribute a collaborative model, ensuring secure data sharing and efficient feature extraction. The system is evaluated on various deepfake datasets, demonstrating its effectiveness in detecting deepfake videos. The proposed method is a promising solution for advancing deepfake detection, leveraging existing data resources and the power of FL and blockchain technology to address the critical need for media authenticity and security.