Blockchain based privacy preserving statistical calculations

In the present study, a blockchain-based application architecture is proposed that uses smart contracts and homomorphic cryptography to allow statistical calculations on confidential data provided by third parties. Blockchain systems provide security features such as integrity and fault tolerance, while homomorphic encryption preserves data confidentiality. In addition to these basic security principles of blockchain, approaches that can improve security and safety during data transfer and storage were also designed. Although similar topics have been encountered in the literature, this thesis differs from other studies since it performs linear regression analysis in an Ethereum smart contract. In the scope of the study, one smart contract was developed for the Ethereum blockchain, and user scripts were written to send data to the smart contract and read data from it. The sample scenario of this study was developed through linear regression analysis. Encrypted data are collected from third parties and sent to the smart contract through linear regression analysis. Encrypted data received by the smart contract was recorded by homomorphically adding them to the relevant variables. Later, a user who carries authority to read encrypted data from the smart contract reads these values and performs their calculations.In the later sections of the study, the design, implementation, and results based on different homomorphic encryption key-lengths of the proposed system are presented. The numerical results of the application show how much time and gas fees are consumed by transactions with different key-lengths sent to the smart contract. Similarities and differences in the gas fee and time consumption of data with varied key lengths and their reasons are examined. In addition, in the evaluation section, light is shed on the security improvements brought about by the proposed system. The proposed architecture was developed within a general perspective which is applicable to all other statistical calculations. If desired, the application steps can be repeated to be applied to different statistical calculations. The study demonstrates that a blockchain-based data-sharing mechanism with homomorphic calculations can be implemented through a smart contract, and that improvements can be made in protecting data from unauthorized users.