A Simulation Framework for Evaluating the Performance of Blockchain-based IoT Ecosystems

Abstract

Recently, it has been appealing to integrate Blockchain with IoT in several domains, such as healthcare and smart cities. This integration facilitates the decentralized processing of IoT data, enhancing cybersecurity by ensuring data integrity, preventing tampering, and strengthening privacy through decentralized trust mechanisms and resilient security measures. These features create a secure and reliable environment, mitigating potential cyber threats while ensuring non-repudiation and higher availability. However, Blockchain performance is questionable when handling massive data sets generated by complex and heterogeneous IoT applications. Thus, whether the Blockchain performance meets expectations will significantly influence the overall viability of integration. Therefore, it is crucial to evaluate the feasibility of integrating IoT and Blockchain and examine the technology readiness level before the production stage. This thesis addresses this matter by extensively investigating approaches to the performance evaluation of Blockchain-based IoT solutions. Firstly, it systematically reviews existing Blockchain simulators and identifies their strengths and limitations. Secondly, due to the lack of existing blockchain simulators specifically tailored for IoT, this thesis contributes a novel blockchain-based IoT simulator which enables investigation of blockchain performance based on adaptable design configuration choices of IoT infrastructure. The simulator benefits from lessons learnt about the strengths and limitations of existing works and considers various design requirements and views collected through questioners and focus groups of domain experts. Third, the thesis recognises the shortcomings of blockchain simulators, such as support for smart contracts. Therefore, it contributes a middleware that leverages IoT simulators to benchmark real blockchain platforms' performance, namely Hyperledger Fabric. It resolves challenges related to integrating distinctive environments: simulated IoT models with real Blockchain ecosystems. Lastly, this thesis employs Machine Learning (ML) techniques for predicting blockchain performance based on predetermined configurations. Contrariwise, it also utilises ML techniques to recommend the optimal configurations for achieving the desired level of blockchain performance.