Next-gen security for medical data: optical encryption empowered by generative adversarial networks

Next-gen security for medical data: optical encryption empowered by generative adversarial networks

연구 분야: Artificial Intelligence

학회: Multimedia Tools and Applications

Medical records play a vital role in the field of healthcare as they enable precise diagnosis and treatment planning of a wide range of disorders using diagnostic modalities like ultrasound, CT scans, MRIs, and X-rays.In the healthcare domain, maintaining the confidentiality of medical images stands as a paramount concern. This research introduces an innovative strategy to address this critical issue, leveraging Generative Adversarial Networks (GANs) for optical image encryption and decryption. The methodology presented offers a unique approach to securing medical images while ensuring accessibility for authorized users. The crux of this technique lies in harnessing the capabilities of GANs. The generator network takes an encrypted medical image as input and produces the corresponding decrypted image.To gauge the effectiveness of the proposed method, experiments were conducted using medical images sourced from reputable platforms. Additionally, test images from the open-source repository were employed to further validate the feasibility and robustness of the encryption method. Comprehensive analyses, including histogram and correlation analysis, NPCR, UACI, PSNR, and SSIM, were employed to evaluate the results. The implementation of the devised approach was executed using Python, showcasing its practical applicability. Furthermore, benchmarking against existing methods was performed, with a specific focus on assessing its performance using the widely recognized "Lena" image. Notably, the results consistently surpassed those obtained by existing techniques. Additionally, scrutiny of the method's resilience against cropping and noise attacks underscored its robust nature in preserving image confidentiality, even when confronted with potential threats.