Research On Privacy-preserving Technology Based On Blockchain Application

The era of big data has arrived,data is being disseminated and traded more and more frequently,which also poses challenges to privacy protection and verifiability in the data sharing process.Blockchain,as a peer-to-peer trusted data network technology for building decentralized distributed storage,provides a technical foundation for building trusted,peer-to-peer data security sharing.The emergence of blockchain provides new ideas for today’s network architecture,enabling users to establish peer-to-peer trusted value delivery between unfamiliar nodes without relying on third-party trusted institutions.Since the data transmitted by users on blockchain networks are open and transparent,in order to protect users’ sensitive data,the data should be encrypted before transmission.The encrypted data will lose some characteristics of plaintext data,such as operations cannot be executed directly on ciphertext,and integrity and correctness verification of ciphertext data cannot be done.Meanwhile,in order to protect the identity of users in the blockchain network,users’ signatures on messages should be anonymous,but this can also lead to the spread of some illegal and criminal acts on the network.In order to solve these problems,this dissertation is concerned with the privacy protection problem in blockchain-based cryptocurrency and the privacy protection problem in blockchain-based federated learning in depth,and proposes a series of blockchain-based privacy protection schemes using the features of blockchain traceability,public verifiability and avoidance of single point of failure,combined with key negotiation,homomorphic encryption,group signature and other key privacy protection techniques,and the performance of the algorithms and schemes is analyzed.The main research results of the dissertation are as follows.(1)To address the problem that digital currency transactions expose the privacy of transaction contents,a blockchain-based privacy protection model for currency transaction contents is proposed,and a quasi-homomomorphic symmetric encryption scheme QHSE(quasi-homomorphoc symmetric encryption)is designed to achieve the hiding of transaction amount data on the blockchain and the validity verification of posted transactions.For each transaction,the transaction participants use Diffie-Hellman key exchange protocol to generate a specific transaction key as a symmetric key.That is,each transaction has a separate transaction key.Even if the key of one of the transactions is leaked,the other transactions will not disclose any private information,which ensures the atomicity and stability of the transactions.In order to achieve interoperability of different symmetric keys,QHSE includes a key update algorithm that enables operations between different ciphertexts.Combined with the key update algorithm,QHSE also includes various types of arithmetic algorithms(multiplication,addition,and comparison),which enable transactions to be executed and verified normally in the ciphertext state.Theoretical analysis and experimental results show that QHSE is secure and efficient.(2)To address the problem that cryptocurrency transactions expose the identity privacy of the counterparty,a blockchain-based identity privacy protection model for cryptocurrency counterparties is proposed,and a group signature algorithm based on the Chinese remainder theorem is designed to hide the identity information of the sender of the transaction.According to the nature of the Chinese residual theorem,the group can dynamically add or revoke group members without changing the key information of the remaining group members,but only dynamically update the public parameters,which greatly improves the flexibility and efficiency of the blockchain cryptocurrency transaction system.It also combines with Schnorr signature algorithm to improve the efficiency of signature as well as verification process.By introducing a fake address of the transaction sender,multiple fake transactions are initiated to hide the identity of the real transaction recipient when initiating a transaction.Fake transactions omit the overhead of real transaction key negotiation and proof tuple generation,so initiating multiple fake transactions has less impact on computational performance.(3)To address the problem of model parameter information exposing the original data privacy in federated learning,a blockchain-based privacy-preserving model for federated learning is proposed,and a blockchain-based no trusted third party federated learning system Nttp FL(no trusted third party federated learning)is designed to achieve the hiding of local model parameters from participating users and the verification of the ciphertext of the aggregated model parameters.and verification of the ciphertext of the aggregated model parameters.The initiator and participants of the federated learning task negotiate the key through group key negotiation,and do not need to distribute the key through a trusted third party.A two-layer encryption mechanism is designed to protect gradient privacy so that the participants cannot obtain any private information other than their own.The blockchain system enables the participants to actively participate in the federal learning process,making the whole process transparent and traceable and avoiding the single-node failure problem.Theoretical analysis and experimental results show that Nttp FL is able to have lower computational and communication overheads without compromising performance and security than existing privacy-preserving oriented federated learning schemes.