Research On Incentive Mechanism Of Crowd Sensing For Localized Privacy Preservation

In recent years,with the popularity of various mobile intelligent terminal devices,Mobile Crowd Sensing(MCS)as a data collection method has gradually penetrated in people's daily life.It uses mobile devices to carry out various sensing tasks,with the advantages including lower costs,convenient maintenance,and strong scalability.However,mobile users participating in sensing tasks could expose their privacy when uploading data to the platform.Although the platform promises rewards to participants,the risk of privacy leakage will still discourage the mobile users to participate in the sensing tasks.Therefore,this thesis focuses on how to combine the incentive mechanism with the privacy preserve mechanism effectively in order to encourage users to participate in the sensing tasks.The main contents of this thesis are following three aspects:(1)This thesis proposes an Incentive Mechanism for MCS based on Local Differential Privacy(IM-LDP),a novel incentive framework with privacy preservation whose functions include incentive,reputation update,data perturbation,and data aggregation.Specifically,the incentive mechanism selects reliable users and provides reasonable rewards.The reputation update mechanism quantifies and dynamically updates the user's reputation.The data perturbation mechanism provides personalized differential privacy for user's sensing data,and the data aggregation mechanism outputs accurate results.(2)For continuous sensing tasks,the differential privacy-based data perturbation mechanism adds random noise,which follows Laplace distribution in the aggregation results to achieve privacy preservation.Users generate noisy data locally according to their privacy budget and add noise to the original sensing data before uploading to the platform.The platform aggregates sensing data and dynamically updates user's reputation based on the amount of noise added by users.The user's reputation will affect his payment and chances in the future auctions.(3)For categorical sensing tasks,users use random response mechanism to locally perturb their original sensing data and upload the perturbed sensing data to the platform.The platform evaluates the quality of the sensing data submitted by users and dynamically updates their reputation,which will also affects their payment and winning chances.This thesis validates the effectiveness of IM-LDP through theoretical analysis and simulation experiments.The experimental results show that the incentive mechanism is closed to the optimal payment cost of the platform while satisfying the task accuracy requirements.The data perturbation mechanism achieves a good tradeoff between data privacy and availability.The data aggregation mechanism incorporates users' reliability to generate highly accurate aggregated results.Meanwhile,payment and auction based on reputation encourage users to provide long-term and reliable services.