The Research On Data Security And Data Selective Transmission In Crowdsensing

Sensing is essential for understanding physical world in practice by providing the basic facts required for analyzing phenomena and mining the causes.Efficient information acquisition is among the most important steps during the process of promoting information construction in cities and industries.It generally bridges the gap between facts and data that containing patterns and knowledge.Meanwhile,the emergence and development of Internet of Things(IoT)continuously deepen the demand on thorough and large-scale sensing.However,the high installation and maintenance costs of traditional wireless sensor networks,together with the nonnegligible energy consumption and limited sensing coverage capacity impede the sensing target to be fulfilled.On the other hand,the amount and variety of smart devices on the market have increased rapidly with the development of wireless communication and sensor technology.These devices are usually equipped with a large number of sensors and have the potential to replace the traditional sensors for sensing the world.In this context,the paradigm of crowdsensing,which leverages users and their devices to observe the around circumstance and events for complex and large-scale sensing tasks,is brought to us.Such a paradigm has promoted applications in many areas,including environment monitoring,city management,public safety,and etc.The participation of individuals facilitates crowdsensing the wisdom of crowd for pervasive sensing.However,user involvement is a double-edged sword and the problem mainly lays in the collection of sensory data.First,unlike traditional centralized deployed sensors,individuals and their devices are uncontrolled,which means they may upload false or even falsify measurements to save their costs.Second,individuals may set their sensitive information at risk when sharing contextual data in crowdsensing,thus degrading their incentives to participate.Finally,using crowd devices as sensors would result in a large amount of reports with redundant data,which poses challenges for efficient data collection and publication.These issues arise from the life cycle of the sensory data in terms of the data generation,data transmission,and data analysis stages.As a result,the issues limit the development of crowdsensing and hinder the emergence of new crowdsensing applications.Thus,this thesis focuses on the data credibility,data privacy,and data redundancy issues in crowdsensing applications.We attempt to guarantee data credibility and protect individual privacy for the consideration of security,and provide selective data transmission from the aspect of quantity.A serious of novel models and schemes are accordingly proposed with the main contributions as follows:(1)We design a rogue access point(RAP)detection application based on crowdsensing.We consider to introduce the idea of crowdsensing into existed RAP problem to analyze the data related issues with specific application.In such an application,we leverage the mobile devices to measure the spatial signal strength around and profile the nearby legitimate access point.Then spatial correlation of measurements is used to identify a rogue by dynamically matching new samples with their nearby recordings in the profile.The RAP detection application helps to clarify the general process and limitations of crowdsensing tasks.(2)We propose a data credibility improving scheme for crowdsensing.We define the independent and collusive data falsification threats for crowdsensing data collection and propose a scheme to guarantee the credibility of sensory data against these threats.First,spatial correlation of sensory data is leveraged to group normal data and false data into different clusters.Then we introduce the participant provenance information and contextual provenance information to estimate the credibility of clusters.The participant provenance refers to the reputation of contributors for the data in one cluster to evaluate the overall reliability.The contextual provenance includes events related to a certain crowdsensing task in time or space domain and enables to build trust on the clusters based on logical supports from these events.Using provenance for credibility estimation,we can finally filter the clusters consist of corrupted data.Experimental results evaluate the effectiveness of our proposal on data credibility assurance with respect to both independent and collusive data falsification.(3)We propose a participant location privacy preservation scheme for sparse crowdsensing.We note that the existed location privacy protection techniques would impact the data recovery process of sparse crowdsensing.In view of such problems,this thesis proposes a scheme to preserve location privacy and maintain the accuracy of data recovery at the same time.We first propose a conjecture about the data correlation required by sensory data recovery with theoretical proof.Based on this conjecture,we design a correlationpreserving,distributed location obfuscation approach with the help of the blind signature technique to protect participants privacy with data recovery process intact.Meanwhile,in order to avoid location disclosure through using sensory data as quasi-identifier,which we denote as the inference attacks,we further design an encrypted data recovery approach based on homomorphic encryption.In this approach,data recovery is decomposed into 7atomic vector operations and we use additive homomorphic encryption and integer vector homomorphic encryption to perform these operations on ciphertexts.We theoretically prove the security of our proposal and evaluate its effectiveness and performance based on real-world environmental datasets.(4)We propose a certain coverage-oriented crowdsensing photo selective transmission scheme.This thesis attempts to relieve the data redundancy during crowdsensing photo collection.First,we identify the uncertain properties of crowdsensing photos and model the certain coverage expectation of data requester as the goal of photo selection by integrating photo coverage and view quality.We then design a novel utility model based on photo distribution diversity and content influence and propose two schemes to maximize the photo selection utility with multiple levels of granularity.In this way,photo coverage can be enlarged by finding uniformly distributed photos,while favoring photos with good content representativeness leads to well quality on the views.Experimental results evaluate the effectiveness of our photo selection schemes and show their performance advantages in terms of the existed techniques.In summary,this thesis starts with a specific crowdsensing applications and brings forth a series of models and schemes for data security and data selective transmission issues in crowdsensing.We have evaluated the effectiveness and the performance of the proposals with theoretical analysis and extensive experiments.The models and schemes can provide both theoretical and technical support for the development of technology and applications in the crowdsensing paradigm.