Innovative Approaches to Secure Image Processing in Decentralized Environment

Abstract

Ensuring robust image security in cloud environments is a critical challenge due to risks such as unauthorized access, data tampering, and privacy breaches. This study introduces a Blockchain-based Secure Image Encryption (BC-SIE) method using Chebyshev Polynomial Fostered Hierarchical Auto-Associative Polynomial Convolutional Neural Network (CPHAPCNN) to enhance security, integrity, and high-fidelity image reconstruction. During encryption, the input image is divided into two unpredictable cryptographic shares, represented by black dot patterns, rendering them meaningless individually and preventing unauthorized access. These shares are then secured on a blockchain using an optimized BLAKE2b hashing algorithm, providing efficient and collision-resistant storage. Furthermore, the Chebyshev polynomial-based encryption strengthens security by introducing pixel scrambling, which makes the method resistant to cryptographic attacks. For decryption, the shares are recombined to reconstruct the image, but this introduces noise, impacting image quality. To mitigate this, a Hierarchical Auto-Associative Polynomial Convolutional Neural Network (HAPCNN) is utilized to reduce noise and preserve image details, ensuring near-lossless recovery. The performance of the BC-SIE-CPHAPCNN framework is evaluated using various metrics, including processing time, correlation coefficient, entropy, peak signal-to-noise ratio (PSNR: 28.44 dB), and mean square error (MSE). The results demonstrate superior encryption security and image reconstruction accuracy, with an updated computed SSIM accuracy of 91.75%. Additionally, the Delegated Proof of Stake (DT-DPoS) blockchain consensus mechanism enhances both security and scalability. Experimental evaluations confirm that this approach outperforms existing methods, making it ideal for cloud storage, medical imaging, and secure surveillance systems.