SAIF-Cnet: self-attention improved faster convolutional neural network for decentralized blockchain-based key management protocol

• Published in: Wireless networks 2024-07, Vol.30 (5), p.3211-3228
• Main Authors: Rejin Paul, N. R., Purnendu Shekhar, P., Singh, Charanjeet, Rajesh Kumar, P.
• Format: Article
• Language: English
• Subjects: Communications Engineering; Computer Communication Networks; Electrical Engineering; Engineering; IT in Business; Networks; Original Paper
• ISSN: 1022-0038; 1572-8196
• DOI: 10.1007/s11276-024-03728-y

Internet of Things (IoT) devices are an essential part of several aspects of daily life for people. They are utilized in a variety of contexts, including industrial monitoring, environmental sensing, and so on. But, secure communication is the major challenge in the IoT environment. Therefore, a decentralized Blockchain-based Key Management protocol using Levy Flight-Equilibrium Optimization and Self-Attention-based Improved Faster Region-based Convolutional Neural Network (BlkKM) method is proposed to determine stable security in tamper-resistant hardware machine that can protect sensitive secret data in the healthcare field i.e., stored cryptographic keys. The keys are categorized as Key Encryption Keys (KEKs) and Data Encryption Keys (DEKs). The number of the keys is decreased by using Levy Flight- Equilibrium Optimization (LF-EO) as organizing nodes with logical sets. Also, Self-Attention-based Improved Faster Region-based Convolutional Neural Network (SA-based IFRCNN) is used for reordering a set of logical nodes to minimize the number of sets after a node exits the network. Additionally, the system makes use of smart contracts for access control as well as proxy encryption to data encryption. The proposed method is compared with existing techniques to validate the security enhancement performance. The evaluation is performed based on throughput, end-to-end delay, storage overheads, and energy consumption. The experimentation results revealed that the proposed method improved the throughput to 220.52bps and diminished the utilization of energy. A greater degree of memory usage is also decreased by using this technique.