Privacy-Preserving Remote Sensing Image Crypto-Watermarking using Discrete Wavelet Transform and Improved Perfumer Optimization Algorithm based on Slow Learner Enhancement Scheme for Band Selection

Rapid growth in multimedia, communication, and deep learning technologies increases privacy, integrity, authenticity, and security issues for Remote Sensing Hyperspectral Images (RSHIs). This paper presents a crypto-watermarking framework based on the Block-based Discrete Wavelet Transform (BBDWT) to ensure the privacy, integrity, authenticity, and security of RSHIs. It uses an improved Slow Learner Improvement Scheme-based Performer Optimization Algorithm (SLIS-POA) for spectral band selection, leveraging higher entropy and higher variance across bands. Furthermore, it uses the Arnold Tent Map (ATM) to encrypt the message image, providing privacy and security for the authentication message. The effectiveness of the system is evaluated on the public Indian Pines (IP) and Salinas datasets based on Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Metrics (SSIM) and Normalized Correlation (CC). The SLIS-POA-DWT provides improved PSNR in the range of 52.26 dB to 62.20 dB, SSIM in the range of 0.9947 to 0.9999 and NC in the range of 0.9835 to 0.9999 for Salinas and IndianPines dataset for recovered images and message.