A comparative analysis of zebra optimization algorithm and chaotic sinusoidal zebra optimization algorithm for video forgery detection system

Abstract

This study presents a comparative evaluation of two metaheuristic optimization strategies: Zebra Optimization Algorithm (ZOA) and Chaotic Sinusoidal Zebra Optimization Algorithm (CSZOA) for enhancing the performance of Convolutional Neural Networks (CNNs) in video forgery detection. A dataset comprising 270 videos with deletion, duplication, and insertion forgeries was used to train and evaluate CNN models optimized with ZOA and CSZOA. The experimental results indicate that the CSZOA-CNN model consistently outperforms both the baseline CNN and ZOA-CNN models across all evaluation metrics, achieving an accuracy of 99.51%, a false positive rate of 0.32%, and a detection time of 39.86 seconds. These findings highlight the effectiveness of integrating chaotic sinusoidal dynamics into optimization processes to enhance CNN training efficiency and detection robustness in video forgery applications.