Research On Trustworthy Mobile Crowdsensing Data Trading Mechanisms In Open System Circumstance

With the widespread proliferation of mobile smart devices and increasingly powerful computing power available,mobile crowdsensing is a newly-emerging mobile computing paradigm derived from crowdsourcing,which can coordinate the mobile crowd users in the network to process large-scale data collection and complex task computing through smart devices.On the other hand,data trading can break the isolated data islands and improve the negotiability and reusability of data,but it also faces the problems of lack of data sources and diversity.Therefore,we leverage the advantages of mobile crowdsensing in data collection,such as intelligence,ubiquity,and low cost,to build a novel mobile crowdsensing data trading system.This data trading system can not only collect and analyze data on demand,but also realize efficient sharing and utilization of data,which is of great significance to promoting data monetization and commercialization.In an open system circumstance,there are many challenges in the design of a mobile crowdsensing data trading system.First,each mobile user can freely enter and exit the system,and it is difficult for the system to learn the user’s personal information and capabilities,so user recruitment is an online decision-making process with unknown information.Second,there is usually a game among multiple participants in the system,and it is necessary to balance the benefits of these participants to motivate their participation.Third,there is a large amount of private information and data interactions among multiple participants,and the untrustworthy behaviors of any participant(such as eavesdropping on others’ private information,violating pre-designed mechanisms,fictitious information,etc.)will affect the fairness of the trading,thus leading to the decrease of participated willingness.In response to the above problems,we conduct in-depth research from the following three aspects,and the main contributions are as follows:(1)Online Recruitment Mechanism of Unknown Users Based on Combinatorial Multi-Armed Bandit.Aiming at the problem of unknown user quality information in an open system circumstance,we propose an online recruitment mechanism for unknown users based on combinatorial multi-armed bandit.This mechanism formalizes the learning process of the unknown quality information of mobile users as the arm-pulling process of the combinatorial multi-armed bandit,and designs an extended Upper Confidence Bound(UCB)index for the quality learning process to balance the uncertainty caused by the quality estimation error.Compared with existing works,we also consider the limited budget of data requesters,heterogeneous data collection tasks,task preferences of mobile users,and cost overstatement issues.To this end,the mechanism establishes an index model for user recruitment by comprehensively considering budget,UCB index,cost and task preference under the combinatorial multi-armed bandit model.On this basis,a greedy strategy for user recruitment is proposed,and the reverse auction and critical theory are used to calculate the payments for mobile users,so as to prevent mobile users with socialized behaviors from untruthfully reporting costs to obtain extra benefits and affect the interests of other mobile users.Through rigorous theoretical proofs and extensive experimental simulations on real-world vehicle trajectory datasets,it is verified that the proposed mechanism can achieve approximate truthfulness,individual rationality,computational efficiency,better quality learning effect,and user recruitment performance under a limited budget.(2)Incentive Mechanism of Heterogeneous Data Trading Based on Stackelberg Game.Aiming at the problem of multi-participant game in an open system circumstance,we propose a data trading incentive mechanism based on the three-stage hierarchical Stackelberg game.Unlike most existing three-party data trading systems which focus on unilateral or bilateral two-party games,our proposed incentive mechanism considers a more complex and simultaneous three-party game process,as well as the situation of unknown quality.By combining valuation information of data requesters(called data consumers),data service cost information of platform,data collection cost information of mobile users(called data sellers),and quality information learned by the combinatorial multi-armed bandit,we design a corresponding profit function for each participant,and deduce the optimal incentive strategy under the two trading scenarios of homogeneous and heterogeneous data prices based on the backward method.The optimal incentive strategy satisfies the special Stackelberg equilibrium,which ensures that no participant can improve his profit by deviating from the optimal incentive strategy.Therefore,the incentive mechanism can simultaneously maximize the profits of the three-party participants(i.e.,equilibrium)and realize the fairness of the trading,thereby encouraging parties with low participated willingness to actively participate in the mobile crowdsensing data trading and reside in the system.Through rigorous theoretical analysis,we prove that the Stackelberg equilibrium composed of the optimal incentive strategy in the two data trading scenarios exists and is unique.Extensive experimental simulations also verify the superior performance of the incentive mechanism.(3)Secure and Trustworthy Data Trading System Based on Blockchain.Any participant with socialized behaviors in an open system circumstance may violate trading rules,including withdrawing from the system at any time,submitting unreal cost,data,evaluation information,eavesdropping on other’s costs and data privacy issues,and most existing works using a centralized third-party system platform as a data trading broker bring about the hidden trust risks and reduced willingness to participate in the system with opaque process.In order to tackle the above security challenges,we design a blockchain-based crowdsensing data trading system,and embed the following two mechanisms in the smart contract running on the blockchain.1)The blockchainbased reverse auction mechanism uses the blockchain as the auction executor and adopts a two-stage bidding strategy to select data sellers and determine payments,which can ensure that all data sellers follow the auction process and report their costs truthfully.Meanwhile,no participant can manipulate and benefit from the reverse auction by eavesdropping on others’ cost information during the trading process.2)The secure truth discovery and reliability rating mechanism is based on homomorphic encryption and data hiding technology,which can motivate data sellers to submit truthful sensing data,urge data consumers to evaluate data truthfully,and protect data privacy from being leaked during trading.By attaching some blockchain-specific design tricks to smart contract and combining the above two mechanisms,we can ensure that the system satisfies the system-level security,i.e.,the trustworthiness of the trading process,truthfulness of cost,data,and evaluation,as well as the privacy-preserving of data and cost information.We verify the practicality and feasibility of the proposed system through extensive experimental simulations in real blockchain environments.In general,the proposed three research schemes can solve the different needs of mobile crowdsensing data trading in an open system circumstance and different problems faced in data collection,data trading,data evaluation and other stages in various scenarios.Moreover,they can be integrated to solve complex data trading scenarios involving multiple stages at the same time,with high practicability and scalability.