A hybrid model for multimedia data compression using generative adversarial networks and chaotic encryption
P. T. Sivagurunathan, A. Sridevi · Systems and Soft Computing · 2025
With the onset of web 2.0 and booming communication technologies like 5G, the amount of internet data being generated on a day-to-day basis has become massive and unimaginable. The fact to be noted is that, much of this generated data comes under the multimedia sector which needs to be efficiently stored and transmitted over public networks. Although several advancements have happened in Internet and telecommunication technologies, there is not much development as far as storage and security are concerned. Hence there arises a dire need for data compression and encryption for an effective and secure transmission of data. This paper proposes a multimedia data compression technique based on generative adversarial network and encryption scheme based on chaos theory. The input images are acquired from the image compression dataset of Kaggle website containing 1, 02, 000 images. The images are encrypted using chaotic encryption scheme involving Arnold’s cat map and Chen’s chaotic system. The encrypted images are then subjected for compression using generative adversarial network model. The encryption performance is evaluated using traditional histogram analysis, chi square test and metrics like variance, maximum deviation, irregular deviation and correlation coefficient. Compression performance of the proposed technique has been measured in terms of mean square error, peak signal to noise ratio, structural similarity index, compression ratio and space saving factors. The proposed system is also compared with existing techniques like Huffman encoding, JPEG, JPEG2000, run length encoding, discrete cosine transform etcetera and has been concluded that the proposed system performs in a better manner than the existing algorithms in terms of both encryption and compression.