Research On Privacy Preserving Problem Of Several Classes Of Collective Intelligence Networks

With the development of computing technology,communication technology and information processing technology,human society is moving toward the era of intelligence,networking and informatization.A large number of smart devices are gradually integrated into people’s lives,which can collect massive data generated from the physical world.By analysis,fusion and modeling,they can realize collaborative sensing,intelligent decision and precise execution,thus providing the technical foundation for wise medical,smart grid,autonomous driving and other applications.In this background,collective intelligence networks have been greatly developed.Based on this theory,two typical data sensing paradigms have emerged,i.e.,wireless sensor networks and mobile crowd sensing.The deployment of wireless sensor networks needs a certain number of static sensors in a specific area for all-weather and all-day data collection.While the mobile crowd sensing is user-centric,where task requesters can use mobile devices possessed by ordinary users as the basic sensing units.With the existing communication infrastructure,the task distribution,data collection and analysis process of large-scale sensing tasks can be realized.Both the above sensing paradigms have their own advantages,but they face the same challenge.In wireless sensor networks,the communication channel is open and vulnerable to eavesdropping attacks.In mobile crowd sensing,the exposure of initial location may greatly threaten the users’ sensitive information.Therefore,privacy protection has become one of the most important research direction in this field.In this thesis,we conducted a detailed analysis of the existing researches at home and abroad and summarized the shortcomings of current researches.Then we focus on the privacy preserving problem of average consensus in wireless sensor networks and multi-objective task distribution in mobile crowd sensing,which are listed as follows.First,a multi-dimensional privacy preserving average consensus(MPPAC)scheme was designed for the multi-dimensional average consensus of wireless sensor networks.This algorithm designed a hierarchical model of sink nodes and ordinary nodes to prevent semi-honest nodes inside the network from collecting the intermediate state information in the consensus protocol.In addition,the RSA algorithm in public key cryptography was utilized to encrypt the sequential information of the message set.When there is an external eavesdropper in the channel,it cannot correctly verify the correct order of the data blocks.When the attacker fails to build an observer,the initial state privacy of the node can be protected.Mathematical proofs were also presented for correctness.Second,a multi-objective privacy preserving task allocation(MPPTA)scheme was proposed for the location privacy problem of multi-objective task allocation in mobile crowd sensing.Note that the current researches mostly used the minimization of the travel distance as the objective,which lacks generality.Therefore,the proposed scheme incorporated maximizing the sensing quality as another sub-objective based on Gray code encoding and bit flipping scheme.Then a multi-objective optimization problem was formulated.We solved the maximization of objective functions after normalization by an improved simulated annealing algorithm embedded with Lévy flight and thus can get the corresponding Pareto optimal solution.Privacy protection and task distribution can be realized simultaneously.