Secure News Video Transmission Using Bidirectional GAN Encoding and Honey Encryption: A Hybrid Deep Learning Approach

Abstract

This study addresses the challenges of data compression and security protection during the dissemination of news videos and proposes an encryption protection method combining bidirectional generative adversarial networks and honey encryption algorithms. This method first uses a bidirectional generative adversarial network encoder to convert video frames into latent codes, and then applies the honey encryption algorithm for encryption processing. At the receiving end, the latent code is decrypted with the correct key and then restored to a video frame by the generative adversarial network decoder, and the complete video is reconstituted. The experimental results on the 50GB dataset show that after 100 iterations, this method achieves an accuracy rate of 0.9, a recall rate of over 0.8, a loss function value of below 10-4, and the mean absolute error and root mean square error stabilize below 0.1. In addition, the encryption speed of this algorithm is 0.45 seconds, the decryption speed is 0.32 seconds, the compression ratio is 30.13%, and the reconstruction quality is 35.89dB, all of which are superior to the existing technologies. This research provides an efficient and reliable solution for the secure dissemination of news videos, effectively preventing the tampering and leakage of video content, and ensuring the authenticity and authority of news content.