A machine learning architecture for quality of service-aware side-chain management in tamper-proof healthcare internet of things and industrial internet of things networks

Blockchain-powered networks have paved their way into industrial and healthcare internet of things (IoT) due to their immutability, improved security, and high trust levels. The distributed nature of these blockchains further adds to their security performance, thereby making them suitable for such high-sensitivity data applications. But in order to manage these blockchains, a large amount of computational power and storage capacity is required. A large portion of this computational power is spent performing repetitive hash computations, which can be conserved. Moreover, a large portion of memory used to store these chains also goes unused, because data stored in the chain by one entity is rarely accessed by other entities. In order to tackle these issues, this paper proposes a machine learning-based side-chaining model for improved blockchain performance. This model divides the main blockchain into multiple smaller chains called side chains, and sandboxes each side-chain from others. Due to this sandboxing approach, these side chains have an optimal memory footprint and require optimal computational power, resulting in improved performance. The proposed algorithm achieves up to 58% to 69% reduction in block addition delay and significant improvements in communication delay and throughput compared with conventional Ethereum and static side-chain architectures.