Research On Techniques Of Task Allocation And Data Security In Crowdsensing-based Internet Of Things

In recent years,with the development of Internet of Things(Io T),cloud computing and machine learning,crowdsensing has been proposed as a new data collection paradigm where mobile smart devices are recruited to sense data instead of the typical sensor networks.Crowdsensing based Io T has attracted significant attention from both industry and academia due to the fact that smart devices(e.g.,smartphones,laptops,cameras,etc.)have sufficient communication/computation capabilities,being able to collect data from surroundings,train machine learning models,and process data locally.Compared with sensor networks,crowdsensing doesn’t need to deploy and maintain Io T devices(e.g.,wireless sensors),having more benefits in financial costs and human labors.This dissertation focuses on crowdsensing based Io T applications,such as smart transportation and smart healthcare.It explores the challenges in task allocation and security in the above scenarios,proposes cryptography and machine learning based solutions.The proposed solutions achieve privacy preservation,effective data security,and efficient task allocation.The main contents and contributions of this dissertation is summarized as follows.1.To tackle the challenges in task allocation and location privacy preservation,we design a symmetric encryption based privacy preservation protocol,preserving both location privacy and data utility.In addition,based on the above privacy preservation protocol,a reinforcement learning based task allocation mechanism is proposed.In this mechanism,a reduced-dimensionality enabled Q-learning algorithm is designed.Compared with typical reinforcement learning,the proposed has better time efficiency and flexibility,being more suitable for large-scale crowdsensing.2.In smart transportation,vehicles have high mobility,triggering redundant task allocation.To solve this problem,an edge computing based cooperative task allocation scheme is proposed.The scheme allows edge nodes to share information and allocate tasks proactively,optimizing the task allocation steps,avoiding repeated task allocation,and improving system efficiency.3.To solve the challenges of balancing data security and data utility in federated learning-based crowdsensing,an anonymous authentication based privacy preservation mechanism is proposed.In this work,a pairing based certificateless signature system is first designed.Then,founded on the signature system,an anonymous authentication protocol is proposed.The protocol preserves the identity of data owners while maintaining sufficient data utility,being able to guarantee both strong data security and high data utility.Besides anonymity,the protocol satisfies confidentiality,mutual authentication,non-linkability,forward security,backward security,and non-repudiation.4.Considering the data aggregation challenges in federated learning based crowdsensing,we propose to implement the federated learning in a three-layer cloud-edge computing system,which can decompose a learning task to distributed edge nodes and participants hierarchically.In addition,we employ secure aggregation protocol in small groups by utilizing the social network of participants,improving the communication efficiency while preserving privacy.