Research On Correlated Data Trading Technology Based On Fast Iterative Combinatorial Auctions

In the era of big data,data is an extremely important and valuable asset.Data trading is a simple but efficient way of data circulation.Participants usually include data consumers,data owners and data trading platforms.Data consumers decide which data to buy and how much to buy,and data owners decide whether to sell data and how to sell data.This paper mainly considers the Internet of Things data collected by end devices.These data are often correlated in terms of content(such as spatio-temporal correlation),and the correlated data should not be sold separately.Pricing data by auction is a common data trading mechanism.Data owners submit data descriptions,data consumers submit bids according to the data descriptions,and the data trading platform determines the data allocation results and data trading prices.However,existing work ignores some important characteristics,such as the correlation between data items.It is very difficult to estimate the value of data;Data valuation is privacy information for data consumers.In addition,the centralized data trading platform may suffer problems such as single point of failure,distributed denial of service(DDo S)attack or privacy-leakage.Firstly,this paper considers the problem of correlated data trading to maximize the sum of data value without disclosing data valuation.We propose a data trading algorithm based on an iterative combinatorial auction.The algorithm finds the final data allocation results and data trading prices after finite rounds of iteration by automatic price increasing mechanism.Data consumers submit bids in the current round to bid for the data combination they want to buy.The data trading platform selects the allocation result of the current round according to the submitted bids.When the trading prices and temporary allocation results in two consecutive rounds are exactly the same,the temporary allocation results becomes the final allocation results and the operation of the algorithm ends.Secondly,because the convergence speed of the data trading algorithm based on iterative combinatorial auction will slow down sharply with the size of the problem,this paper considers the fast iterative combinatorial auction data trading algorithm based on Bayesian learning.This method introduces the prior knowledge of data consumers on data valuation,constructs the data valuation model,calculates the maximum a posteriori probability of the price function through the expectation maximization algorithm based on Monte Carlo sampling,and updates the data trading prices of each round.Experiments show that the fast iterative combinatorial auction data trading algorithm based on Bayesian learning can greatly improve the convergence speed.Finally,this paper proposes to construct a safe and reliable data trading model for participants and data trading platform.This paper uses blockchain technology to build a decentralized data trading platform which can effectively prevent single point of failure,DDo S attack,and privacy leak.Since data owners may refuse to send data after receiving rewards,or data consumers refuse to pay,we have designed a series of smart contracts to prevent fraud.The data auction smart contract realizes the first data trading method.The whole data trading process can be traced and tampered with to ensure the security and credibility.