Blockchain Based Deep Fake Detection and Verification

Abstract

The rapid advancement of deep learning has led to the proliferation of deepfake content, posing significant threats to privacy, media integrity, and digital trust. To counteract this growing concern, this paper presents a novel Blockchain-Based Deepfake Detection and Verification System that integrates machine learning with decentralized ledger technology. The proposed system utilizes advanced convolutional neural networks (CNNs), including the Xception architecture, to accurately identify manipulated visual media by extracting spatial features from video frames. A federated learning approach ensures privacy-preserving model training across distributed devices, eliminating the need for central data aggregation. Once deepfake content is detected, the results—along with metadata such as content hash, timestamp, and classification label—are immutably stored on the Ethereum blockchain using smart contracts and IPFS for transparency, traceability, and verification. Experimental results demonstrate high detection accuracy and robustness against various forgery techniques. The integration of blockchain not only secures the integrity of detection results but also promotes trust in automated verification systems. This hybrid framework paves the way for scalable, secure, and privacy-preserving solutions in combating deepfake threats across digital ecosystems.