Research On Data Collection And Privacy Preservation Mechanism In Distributed Systems

A distributed system is a computer system which consists of multiple functional independent computing resources connecting with network.These system resources can cooperate with each other to provide services.As a result of flexibility,scalability,openness,and high cost performance,distributed system has been widely studied and applied.Based on the model that data can be regard as services,distributed system first needs to perform data collection.However,in an open,self-organized communication model,the ways of data collection and the efficiency of the proposed ways affect the performance of distributed system.After data collection,the security and privacy issues have been concerned by people all the time.These problems have become one of the main obstacles to popularity and wide application of the distributed system.This thesis explores the data collection and privacy protection mechanism of the distributed system from three typical distributed systems(two-tiered sensor networks-a low-power distributed system,cloud computing–a highperformance distributed system,social networks-a high-interaction distributed system),and then proposes a series of schemes or protocols.The main contributions are summarized as follows:(1)This paper proposes a dynamic data reduction architecture in Two-tiered wireless sensor networks.In wireless sensor networks,the energy of sensor node is limited.In order to reduce the energy consumption of sensor node and extend the lifetime of sensor network with the collected data satisfies the accuracy requirement of the specific application,a dynamic three-level data reduction architecture,called DR3,is designed.In the first level,the sensor nodes are divided into several groups.For the sake of achieving internal group data reduction,a centroid node selection algorithm and a translation model are presented.This translation model can utilize the reading of the centroid sensor to approximate the whole group's reading to avoid the redundant real-time sampling of the other sensor nodes in this group.In the second level,a predication model is developed.Instead of real-time sampling,the centroid sensor node can use the past sensing data to predict the future outcome by this predication model.In the third level,to reduce unnecessary transmission between sensor and the amount of sampled sensor node,this paper designs a correlated group data reduction model.The group nodes,whose readings have high correlation with other groups,can utilize other group's readings to predict their own readings.Three different levels of data reduction program are collaborating with each other and interacting with each other.Within the scope of the predefined tolerance system error,three different levels of data reduction can be run in a parallelled way to finish the data collection with low communication and computation overhead.Furthermore,based on this architecture,we propose two implementations.Two different implementations are suitable for different monitoring applications.Finally,extensive experiments on real-world datasets demonstrate that DR3 can satisfy users' requirements and the energy consumption of sensor node is relatively low.(2)As for the digital devices have limited computation and storage ability,this paper proposes a video surveillance scheme in cloud computing,called dynamic Compressive Sensing(DCS)scheme.Since the digital cameras everlastingly collect the video data,the volume of sampled data is extensively large,which is hard to be stored and managed locally.Outsourcing surveillance video data to the cloud can achieve cost saving and flexibility.The video frames are divided into key frames and non-reference frames.To protect the security of video data,this scheme first utilizes the “downsampling” and “encryption” properties of Compressed Sensing(CS)technique to improve the efficiency of video coding and ensure the security of the video.Then,to resist the the known-plaintext attack,we design a dynamic measurement matrix to sample the key frames of video.Only authorized users can recover the original video data for further analysis.Furthermore,in order to improve the performance of this scheme,authorized users are allowed to take full advantage of the powerful computation ability of the cloud to decode the non-reference frames without leaking any information.Finally,security analysis and extensive experiments on real-world datasets confirm that the proposed scheme can ensure the security of the video and the communication overhead of the proposed scheme is relatively low.(3)This paper explores how to protect user's popularity privacy in online social networks.We consider an “honest-but-curious” cloud server in our model.Different from previous data privacy protection schemes,based on this model,we propose a novel utilitybased popularity anonymization(UPA)scheme to protect user's authentic popularity privacy in online social networks without utility loss.There are two type of users in our system,followers and followees.If a follower A wants to access or retrieve the private data files of a followee B,A must follow the followee B.First,to protect user's authentic popularity privacy,we design a k-anonymous popularity-based following(KPF)protocol based on k-anonymity model to achieve an approximately k-in-degree anonymous social network graph.Then,we design a hierarchical authorization and capability delegation(HACD)model to implement fine-grained access control for encrypted data files.Based on HACD model and cryptographic techniques,we develop a fully utility-based interaction(FUI)protocol to achieve normal services of online social network,such as retrieving/sharing/reading the sensitive data files,without compromising data files privacy.Finally,we conduct security analysis and extensive experiments on the real-world social network.The experimental results verify that,compared with existing work,the proposed scheme can keep the utility of the social network and the computation overhead is relatively low.(4)This paper proposes a privacy-preserving proximity based location query(PPLQ)protocol to protect the location privacy of users in location based social networks.First,we divide user's location coordinate into two parts.The first part is origin location coordinate,the second part is offset location coordinate.To protect the privacy of the user's location coordinate,this paper designs a ciphertext-policy attribute based encryption and a symmetric encryption algorithm to encrypt the user's origin coordinate,offset coordinate,respectively.Then,we utilize the multi-scale technique to represent users' location coordinates and coordinates ranges in the index and query conditions,and then utilize prefix membership verification and predicate encryption technique to achieve the secure multi-dimensional keyword search and private proximity testing simultaneously.This query protocol enable a querier to query the publisher's information from cloud service provider effectively.Furthermore,the proposed protocol enable each user to maintain his own location policy,and even assign different location policies for different queries.Finally,we conduct rigorous security analysis and extensive experiments for the proposed protocol.The simulation results confirm the efficacy and efficiency of our scheme.