ROBUST OBJECT-BASED VIDEO FORGERY DETECTION USING DEEP CONVOLUTIONAL NEURAL NETWORKS

Abstract

The rapid advancement of video editing technologies has made it increasingly easy to manipulate video content, leading to the rise of sophisticated object-based forgeries that can mislead viewers and compromise the authenticity of digital media. Detecting such forgeries has become a critical challenge in fields such as digital forensics, media verification, and security. This paper presents a robust approach for object-based video forgery detection using Deep Convolutional Neural Networks (CNNs), designed to identify subtle spatial and temporal inconsistencies introduced during manipulation. The proposed method leverages a deep CNN architecture to automatically extract discriminative features from video frames, capturing both visual artifacts and motion irregularities associated with tampered regions. The system incorporates frame-level analysis combined with temporal consistency checks to improve detection accuracy. Advanced preprocessing techniques, including frame extraction, noise filtering, and normalization, are applied to enhance model performance. The model is trained on diverse datasets containing both authentic and forged videos, enabling it to generalize effectively across different types of manipulations. Experimental results demonstrate that the proposed approach achieves high detection accuracy and robustness against complex forgery techniques such as object insertion, removal, and replacement. Compared to traditional methods, the deep learning-based model shows superior performance in terms of precision, recall, and processing efficiency. Overall, this work provides an effective and scalable solution for detecting object-based video forgeries, contributing to improved digital content authenticity and trust in multimedia systems.