Privacy-preserving auction using multi-party computation

Abstract

When it comes to trading and auctions, a party would not want to reveal their intention to buy or sell, as other parties may use the information to influence the market (such as sellers driving market prices up when they know an incoming buyer intends to purchase a large volume of items). Traditional approaches have relied on a trusted third party, which are susceptible to illegally performing trades with their insider information for their personal gain. To eliminate the reliance on a human intermediary (whether that is a trusted broker or a system operator for electronic trading means), one possible way is to conduct an auction using Multi-Party Computation (MPC). In short, the MPC technique shares inputs across multiple parties – where no individual parties have the full input, but all the parties will jointly compute with functions to get the output. Knowing more about MPC and their current works on auction, this project sees the use of MPC technique to perform an auction with privacy-preserving properties. As this was a pair project, my role is to create an algorithm to trade multiple volumes of items between buyers and sellers. The MPC algorithm will be compared with the same algorithm in cleartext form (that is, without regards for maintaining privacy of inputs or variables inside the algorithm), to reflect the trade-offs between efficiency favouring the cleartext algorithm, and security which should be achieved in the MPC algorithm. The MPC algorithm was created with 2 buyers and 2 sellers in mind, and we will analyse the impact on algorithmic complexity with more parties involved. Lastly, the algorithm will be integrated with the system architecture proposed by my teammate. This section will be brief to reflect the end-product of the project.