AUTOMATIC DETECTION OF OBJECT-BASED VIDEO FORGERY: A MOTION RESIDUALS ANALYSIS APPROACH

Abstract

The proliferation of surveillance cameras has led to an exponential increase in video data, necessitating the development of efficient forgery detection techniques to ensure the integrity of surveillance footage. In this manuscript, we propose an object-based approach
for forgery detection in surveillance videos, leveraging optimized Convolutional Neural Networks (CNNs). Our method focuses on identifying forged frames by analyzing objectlevel features extracted from each frame, thus enhancing detection accuracy and robustness. We describe the methodology, including data preprocessing, CNN architecture design, training, and evaluation. Experimental results demonstrate the effectiveness of the proposed approach in accurately detecting forged frames in surveillance videos, even amidst complex scenes and varied manipulation techniques. The integration of object-based analysis with CNNs showcases promising prospects for enhancing the security and reliability of surveillance systems.