Paper Details

Name: Pragya Rajput, Ujjwal Kumar, Parit Rajput, Gautam Das, Raja Siddharth A R & Shubham
Country: India

Year: 2026
Volume: Volume No: 13, January, Year: 2026 (Special Issue)
Page Number: 97-108
DOI: https://doi.org/10.5281/zenodo.18951814

ABSTRACT
Deep fake technology brings significant concerns regardless of the domain in which it is applied from misinformation to cyber criminals and privacy violation. This new technology is a real danger to several fields as it can disseminate fake news, contribute to the increase of the number of cyberthreats and compromise the protection of personal data. The techniques previously used in detecting deep fake basically do not follow the rather high evolutionary rates of these generation techniques hence yielding a very high level of false positives and false negatives. This work seeks to investigate the viability of FA as an innovative method that strengthens the signal and the spatial resolution of deepfake detection techniques. This research aims to create multiple and complex datasets by combining the changes in facial features
comprising expressions, lighting and occlusion to assist the training of detection models. To assess the proposed approach in depth, the current and one of the most developed machine learning models including CNNs and high-level models are used. Last but not the least, we observed that when the proposed method includes dynamically augmented data, it added
even more value to the detection and reduces error rates substantially; thus it offers more effective ways to counter deep fake threats. These findings outline how knowledge of new strategies to counter the contamination of digital media or the protection against improper use of the deepfake technology is important.

Keywords: Deepfake Detection, Dynamic Face Augmentation, Generative Adversarial Networks (GANs), Machine Learning, Convolutional Neural Networks (CNNs), Data Augmentation, Misinformation, Cybersecurity, Image Analysis, Model Performance.