Privacy-preserving blockchain storage using hybrid ECIES-AEAD and CNN-RNN classifier with attention mechanism

Blockchain technology offers a secure and decentralized data storage solution, yet scalability and performance degradation remain key challenges as data volumes increase. This study introduces a secure and scalable blockchain-based framework for managing sensitive financial data. Users register with fingerprint identification to ensure maximum security, while pre-processed data is encrypted using the Exp-HFECC algorithm and hashed to generate unique block IDs. Data is stored through a key-sharing protocol, improving scalability via sharded storage mechanisms. For intelligent access control, a hybrid deep learning model, DARC-Net, combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) with a Dual Attention Block (DAB), classifies nodes as malicious or non-malicious. The system integrates dimensionality reduction and node feature extraction for efficient classification. Experimental results demonstrate high accuracy across multiple datasets, achieving 99.5% accuracy on financial data, 97% on cryptojacking and 98% on fingerprint datasets, with strong precision and recall values. The proposed approach enhances blockchain scalability, security and interpretability, offering robust and adaptive performance across diverse applications.