Blockchain-Assisted Video Integrity Verification UsingResNeXt and LSTM-Based Deepfake Detection

With the rapid advancement of deep learning algorithms, synthetic media commonly known as deepfakes have reached a level of realism that makes them nearly indistinguishable from authentic human appearances. Such content poses serious threats, including misinformation, impersonation, and cybercrimes. In this paper, we propose a novel blockchain-assisted framework for the detection and validation of deep-fake videos. Our approach integrates deep learning and decentralized verification mechanisms to enhance multimedia integrity. Frame-level features are extracted using a ResNeXt Convolutional Neural Network, and these are further processed using a Recurrent Neural Network (RNN) equipped with Long Short-Term Memory (LSTM) units to analyze temporal patterns across video sequences. To ensure tamper-proof logging and traceability, detection results and video metadata are immutably stored and verified on a Blockchain ledger. This combination not only improves classification accuracy but also provides a transparent and secure method for verifying video authenticity. Experimental comparisons demonstrate that our Blockchain-integrated model outperforms existing methods in both accuracy and reliability, contributing to the state-of-the-art in secure multimedia authentication.