A Data Mining Method For IoV Privacy Protection Based On Homomorphic Encryption

In recent years,the Io V(Internet of Vehicles)has undergone gradual development,whose commercial potential has garnered widespread attention.The Io V not only excels in enhancing road safety and driving comfort but also plays a crucial role in constructing novel transportation services.Nevertheless,as the number of vehicles connected to Io V continues to increase,the data resources are expanding exponentially.Since the data are typified by multi-source isomorphism,low-value density,and fast processing requirements,traditional analytical tools and technical processing methodologies are inadequate for large-scale,real-time Io V environments.Thereafter,the integration of machine learning and data mining techniques is imperative for constructing highperformance Io V systems.Nevertheless,given the limited computing and storage resources available for onboard units,it is viable to employ AI algorithms such as deep learning and clustering to implement these systems and transmit the resulting data to cloud servers.The deep integration of the cloud and Io V enables not only efficient information exchange and processing but also cheap data storage and high-performance computing services for Io V.However,ensuring safe transmission,storage,and data processing for privacy is an undressed issue in these processes.Traditionally,data can be encrypted before being transferred and stored but requires decryption before processing.In an open and untrustworthy environment such as Io V,there is a risk of data breaches if they are decrypted in the cloud.To eliminate this conflict,homomorphic cryptography is hence introduced,which possesses the merit of "computable invisibility",enabling the calculation of encrypted data without decryption where the decryption result is the same as plaintext.Based on homomorphic cryptography,this thesis conducted a detailed study on data mining for Io V privacy protection.(1)To ensure the confidentiality of Io V data during transmission and in the cloud,data can be encrypted before transfer and storage.However,a large amount of redundant cipher data may lead to excessive storage space due to the high volume of network data.On the other hand,ciphers satisfy semantic security and cannot be directly compared to determine the identity of their corresponding plaintexts.To deduplicate identical encrypted messages,this thesis constructed a scheme based on the NTRU(Number Theory Research Unit)homomorphic encryption to realize private data deduplication in Io V.This approach overcomes the disadvantage that traditional data encryption schemes cannot resist dictionary attacks and requires the intervention of a trusted third party.Additionally,it solves the contradiction between indistinguishability and the comparison of cryptography restricted to semantic security.(2)With the increasing saturation of Io V data,exploiting the potential value of these resources and building a more structured high-performance Io V have proven to be an urgent problem.Meanwhile,since Io V data contains a lot of private information,such as identities and assets,data should be generally represented in the form of ciphertexts.Traditional usage of encrypted data is limited to decrypting before processing,where the confidentiality of decrypted data cannot be guaranteed within open and untrustworthy Io V environments.Thereafter,this thesis constructed a machine learning scheme based on light-weighted CSP(Conjugate Search Problem)homomorphic encryption to simultaneously protect the privacy of Io V data and realize the classification tasks using a neural network.(3)With further amassing of Io V data,data mining becomes the key technology to break through the barrier of Io V development.Data mining can support better data analysis for service providers and better-personalized service for Io V users.However,it requires a large collection of users’ personal information,which may lead to privacy breaches.To address this issue,this thesis uses the homomorphic encryption scheme to carry out data mining.Since data mining usually involves multi-party joint computation,single-key homomorphic computation is limited.Therefore,this thesis improved the CSP encryption scheme to a public key encryption paradigm named P-CSP and constructed an Io V privacy protection data mining method based on it.