Abstract: The blockchain technology allows participants to establish pseudonymous addresses, decoupling their real-world identities from their activities on the blockchain network. However, pseudonymity is not completely anonymous and several attacks pose a growing threat to transaction privacy between users. We propose a privacy protection scheme utilizing random mixing on Ethereum, which includes three strategies: RandomMix, TRandomMix, and VRandomMix. The three strategies can protect the sender accounts, the receiver accounts, and the transaction values for one transaction, two transactions, and some equal-value transactions, respectively. We conduct experiments to evaluate both security and performance. The security evaluation results indicate that RandomMix and TRandomMix can resist account clustering attacks, transaction fingerprinting analysis attacks, and Sybil attacks, while VRandomMix can safeguard against account clustering attacks, and is weak against the latter two attacks. In terms of performance, VRandomMix acquires fewer on-chain transactions than RandomMix and TRandomMix. Furthermore, all of them need no off-chain communication and low gas consumption.