Efficient Video Deepfake Detection Using Boundary Hashing and Recurrent Neural Networks
Book Chapter
Hariprasad, Y, Govindarajan, G, Iyengar, SS et al. (2026). Efficient Video Deepfake Detection Using Boundary Hashing and Recurrent Neural Networks
. 1456 LNEE 261-272. 10.1007/978-981-95-2196-8_17
Hariprasad, Y, Govindarajan, G, Iyengar, SS et al. (2026). Efficient Video Deepfake Detection Using Boundary Hashing and Recurrent Neural Networks
. 1456 LNEE 261-272. 10.1007/978-981-95-2196-8_17
The rise of deepfake videos has posed significant challenges in digital forensics. Traditional deepfake detection techniques, such as Convolutional Neural Networks (CNNs) and optical flow methods, often struggle with detecting subtle manipulations in smaller facial regions, as they tend to focus on large-scale alterations. These limitations highlight the need for more refined detection strategies. In this paper, we introduce a novel boundary-based hashing technique combined with Recurrent Neural Networks (RNNs) to effectively detect small-scale facial anomalies in deepfake videos. By leveraging boundary-specific hashing, our approach addresses the limitations of CNN and optical flow methods, offering a more precise detection mechanism. We provide an in-depth discussion of the technical details, implementation, and evaluation of our method on a dataset of 200 deepfake videos. Our proposed approach significantly improves detection performance, achieving an accuracy of 98.26%, surpassing existing CNN-based methods.
