Tampering Detection and Encryption: A Deep Learning Approach for Video Security

With the proliferation of digital video content across various applications, ensuring the integrity and security of video data has become paramount. This research proposes a comprehensive methodology for enhancing video data security and detecting tampering attempts. By preprocessing video data into frames and utilizing pretrained models for object detection, sensitive regions are identified and encrypted using the CHACHA20 algorithm. Various tampering techniques are simulated to evaluate the robustness of the approach, with performance assessed using metrics like SSIM and PSNR. Furthermore, tamper detection is executed through classification using recurrent neural networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) architectures, enhanced with attention mechanisms. Performance evaluation is based on classification reports and loss curves. Overall, the outcomes highlight the effectiveness of the proposed framework in ensuring video data security and integrity, particularly in detecting tampering using deep learning methods.