AN INTELLIGENT SYSTEM TO IDENTIFY FAKE VIDEOS ON ONLINE SOCIAL NETWORKS USING MACHINE LEARNING

Abstract

Advances in emerging technologies have led to the wide dissemination of fake videos on online social networks. This research employs the hybrid machine learning and deep learning algorithms to recognise video forgery on social media. This research presents a hybrid deep learning and machine learning approach for detecting manipulated videos. The proposed model employs the OpenCV deep neural network (DNN) face detector to locate facial regions in video frames, after which a pretrained ResNet50 convolutional neural network is applied to extract deep features, and a Support Vector Machine (SVM) is used to classify authentic and fake content in a binary fashion. The model was trained and evaluated on a dataset of publicly available 128 short-form videos gathered from TikTok and Facebook platforms. The five-fold stratified cross-validation results generated an average accuracy of 89.1%, precision of 87.6%, recall of 96.1%, F1-score of 91.5%, and an AUC of 0.95 for the SVM model. Furthermore, the comparative analyses showed that SVM outperformed Logistic Regression, Random Forest, Gradient Boosting, and K-Nearest Neighbour classifiers. The findings demonstrate that combining automated face detection with deep feature extraction and classical machine learning significantly improves fake-video detection and contributes to preserving authenticity in digital communication.