Research On Network Resource Auction Mechanisms With Differential Privacy

With the rapid development of the Internet,the demands of users for network resources have gradually increased.The traditional static resource allocation methods have low utilization rate and cannot meet the supply and demand requirements of resources.Therefore,the effective utilization and reasonable allocation of network resources have become one of the research hotspots in the Internet era.Due to their good efficiency and fairness,auctions are widely used in network resource allocation,including spectrum allocation,cloud resource allocation and task allocation of mobile crowd sensing.For the scarce and non-renewable spectrum resources,auction can be used to encourage authorized users to lease their idle spectrum to cognitive users to alleviate the shortage of spectrum.In addition,with the popularity of cloud computing,cloud service providers sell virtual machine resources to cloud users in the way of auction,so that cloud users can complete computing tasks.At the same time,with the enhancement of computing power and perception ability of mobile devices,mobile crowd sensing that collects large amounts of perception data at low cost has emerged.An auction-based incentive mechanism of mobile crowd sensing can encourage more workers to participate in the perceptual task and submit their collected perception data to the platform.There are a large number of auction-based network resource allocation schemes.These schemes can ensure the efficiency of resource allocation and the benefit of auctions,but they rarely consider the issue of user privacy protection.If the privacy of user is not effectively guaranteed,it will reduce the enthusiasm of users to participation.Aiming at the privacy protection requirements in the spectrum auction,cloud auction and the incentive mechanism of mobile crowd sensing,this dissertation designs the reasonable auction mechanisms to solve the above problems by introducing differential privacy technology.The main research of this dissertation is as follows:(1)A differentially private double spectrum auction with approximate social welfare maximization is proposed.The existing differentially private spectrum auctions only consider one seller and its revenue,but ignore the situation that multiple sellers participate in the auction.This dissertation integrates the exponential mechanism with double spectrum auctions,and establishes an appropriate matching relationship between the buyers and selles through the method of bidding order,so as to realize the two-side privacy protection between buyers and sellers,and guarantee the approximate optimization of social welfare.Then,the auction mechanism is improved from three aspects: mechanism design,utility function design and grouping algorithm design.Theoretical analysis proves that the proposed mechanisms have the characteristics of differential privacy,approximate truthfulness and approximate maximization of social welfare.Extensive experimental results show that the proposed mechanisms achieve good performances in terms of social welfare,and the improved mechanisms ensure a better utility than the basic mechanism.(2)A differentially private combinatorial cloud auction is proposed.The existing secure combinatorial cloud auctions have the risk of disclosing the bid privacy of users from the auction results.To solve the above problems,this dissertation designs a differentially private combinatorial cloud auction mechanism,which adopts two-level pricing method and employs the exponential mechanism to select the pricing vector,and takes the revenue of the seller as the utility function of the exponential mechanism,so as to realize the bid privacy protection of users and ensure the revenue of the cloud service provider.Then,this dissertation adjusts the operation cost and auction utility by increasing the number of calls to the exponential mechanism and decreasing the space of pricing choice for each call.Theoretical analysis proves that the proposed mechanisms achieve differential privacy,approximate truthfulness and high revenue.Experimental results show that the proposed mechanisms can achieve a tunable tradeoff between operation cost and auction utility.(3)A differentially private double auction with reliability-aware in mobile crowd sensing is proposed.Considering that the existing privacy protection schemes do not ensure the quality of sensory data,this dissertation first heuristically selects the most reliable worker as the candidate winner of each clearing price tuple to collect accurate sensory data.Secondly,the proposed mechanism uses the exponential mechanism to determine the transation prices,so as to protect the privacy of the bidders.Finally,the proposed mechanism designs a more proper utility function or a more practical pricing strategy to improve the social welfare.Theoretical analysis proves that the proposed mechanism can not only guarantee differential privacy,but also ensure the economic properties of the double auction.The experimental results demonstrate that the improved mechanisms outperform the basic mechanism in term of social welfare,and all proposed mechanisms can generate high-quality sensory data.