Research On Privacy Preserving Data Collection And Publishing Based On Compressed Sensing

The Internet of Things(Io T)is widely deployed in daily life,which generates a large number of images and videos every minute.It is difficult for normal users or smart devices in the Io T to support the unrealistic large storage and high-performance computing requirements for massive Io T data.Therefore,effective data collection is inevitable.Data collection technology reduces communication consumption,saves energy,improves accuracy,and extends network life.Sensor devices in Io T can sense,process,and disseminate real-time information,but they also have limitations of limited energy,small memory capacity and slow processing speed.The application of Io T further expands the advantages of Internet,such as data sharing,facilitating data collection and achieving more precise management.However,due to the fact that sensor devices are usually deployed in harsh,unattended or even hostile environments,Io T system is facing more and more attacks,and data security is threatened.Therefore,it is essential to consider the low-complexity processing and confidentiality of sensing nodes in data collection.Secure data collection technology combined with secure data publishing can fully protect the personal privacy of users.At the same time,the privacy preserving data publishing is not suitable for a single user type,and the diversity of user permissions should be considered.Users with different permissions obtain different released data in accordance with a variety of application scenarios.In addition,the privacy of the unpublished data is as important as the published data.Due to the strong correlation of data,a small amount of the published data may also reveal the privacy of the unpublished data.Therefore,considering the user's controllable authority and the data privacy protection in data collection meets the actual requirements.This thesis focuses on the problems of a large amount of data,rich data types,data security threats in data collection,and controllable access and privacy protection in data publishing.The work includes the following aspects:(1)In view of the neglect of privacy protection and high consumption in the existing compressed data collection,a privacy preserving data collection scheme based on compressed sensing(CS)and embedding is proposed.The scheme embeds different factors into the compressed measurement values for different types of nodes,so as to destroy the inherent structure of data and achieve privacy protection.Any intermediate node does not need to decrypt the received message before encoding,but can directly encode the received message and its own reading.The scheme realizes the compressed data collection with low computation consumption at the node under the security guarantee.(2)After analyzing the powerful attack capability of sparse estimation technology on the unpublished data,a privacy preserving controllable compressed data publishing scheme is proposed.The scheme considers not only the availability of the published data,but also the privacy protection of the unpublished data.The error of CS reconstruction naturally provides protection for the unpublished data to resist attacks.At the same time,for the same released measurement,different users can get different publishing results according to the different keys held by them,so as to achieve controllable access.In addition,considering the utility of the released data and the privacy requirement of the unreleased data,the scheme discusses the upper bound of the number of the released data.The data owner can freely select the released data within the maximum release number.Extending the scheme to the Internet of Multimedia Things(Io MT)can protect the sensitive information in the image,which reflects the potential of the scheme.(3)Aiming at the challenges faced by the massive image data acquisition in Io MT,a secure multi-image compressed acquisition and separate reconstruction scheme is proposed.The scheme analyzes the signal energy leakage of the CS measurement,designs the measurement protection,breaks the original distribution of the measurement,and maintains the preliminary uniform distribution of the measurement.In addition,the separate reconstruction of a single image is provided for the fused measurement of a data set,which can reconstruct the corresponding image according to the actual needs,reducing unnecessary resource consumption.The corresponding images are reconstructed according to actual needs,reducing unnecessary resource consumption.(4)Aiming at the multiple types of data and the diversity of user permissions in Io MT,a fog-assisted nested compressed collection scheme for multiclass data is proposed.This scheme breaks the traditional data collection framework and introduces fog computing to improve the quality of service,and ensures the security of input and output in each layer.Different types of data correspond to different user rights,and the authentication of authorized users is realized by embedding identity watermark in fog.In addition,the design of reconstruction task transformation for the sensing data ensures that the cloud reconstructs data in encryption mode.