<p>Fast-forwarding with respect to surface advancements in deepfake generation technologies, the digital media credibility is now under threat; there are serious threats posed to privacy and security upon deepfake video generation. It is hypothesized in the present study that employing advanced algorithms along with blockchain technology will greatly enhance the accuracy and security of deepfake detection systems. To this end, a novel framework for detecting deepfake combinations was designed by conceiving DDO-AGNN with blockchain technologies for federated learning. The present model includes the use of a comprehensive dataset, detailing diverse deepfake videos; these were normalized via pre-processing by min-max normalization. The DDO-AGNN algorithm was implemented in Python and optimized by DDO for improved feature extraction and classification. Blockchain technology was employed for federated learning, ensuring privacy-preserving collaborative model training across multiple nodes. The tool interchangeably used in this framework includes Python for the algorithm implementation, blockchain for federated learning, and several machine learning libraries for model training and evaluation. The results indicate that the proposed DDO-AGNN outperforms all existing methods for deepfake detection with regard to accuracy, precision, recall, and F1-score. More precisely, the model with 99.38% accuracy and 99.26% precision was able to surpass state-of-the-art methods including ResNet-SwishDense54, BlazeFace+DFN + XGBoost, and YIX by 98.67% recall and 98.89% F1-score. Blockchain-enabled federated learning successfully retained data privacy, but due to the high computational cost (10–100 GPU hours), scalability posed a challenge. These findings indicate that the integration of DDO-AGNN with blockchain technology provides an effective and secure solution for deepfake detection, representing a significant enhancement over current techniques.</p>