Research On Privacy Preserving Support Vector Machine Algorithm Based On Secure Multi-party Computation

Data mining refers to the process of searching for hidden information from massive data.In order to improve the accuracy of data mining,on the one hand,the algorithm needs to be improved.On the other hand,data mining needs to be done on a large amount of data,which is generally derived from different units or users.Due to the limitations of local storage and computing,with the development of cloud computing,more and more users choose to upload data to the cloud to achieve storage outsourcing and computing outsourcing.The cloud is a third party that is not completely trusted.It will cause the user to separate the ownership and control of their own data,which in turn leads to the risk of data privacy information leakage.In addition,data mining is a "double-edged sword".Data mining directly on data with private information can also lead to the risk of data privacy information leakage.Based on secure multi-party computation,this paper proposes a data mining scheme with privacy preserving function on multi-user encrypted data by using support vector machine algorithm.Aiming at the computational problem of multi-user encrypted data in integer domain,this paper proposes a dual cloud framework model with storage outsourcing and computing outsourcing.A homomorphic addition and homomorphic multiplication protocol supporting addition and multiplication on the integer domain of multi-key encryption is designed.The protocol first blinds the ciphertext data based on the "blind" technology,and then converts the multi-key encrypted data into the same single-key encrypted data through the interaction calculation between the two clouds,and finally,the calculation of adding and multiplying on the ciphertext is completed by using the properties of the single-key homomorphic protocol.Aiming at the computational problem of multi-user encrypted data on the rational number domain,this paper follows the dual-cloud framework designed on the integer domain.Rational numbers contain integers and decimals,and the encryption and decryption calculation and storage of decimals are different from integers.This paper first calculates the numerator and denominator by using the homomorphic addition and homomorphic multiplication protocol designed on the integer domain.Finally,the addition and multiplication calculations are performed on the rational number field of multi-key encryption.Based on the homomorphic protocol designed to support addition and multiplication on ciphertext,the support vector machine algorithm can be used for data mining on the rational number field and integer domain of multi-key encryption.In the semi-honest security model,it can be proved that under the premise of ensuring the accuracy of data mining,the algorithm designed in this paper can protect the user's data privacy,the privacy of intermediate calculation results,the privacy of the classification model and the privacy of the final classification prediction results.Based on the algorithm designed in this paper,we built a data mining system with privacy preserving function using the support vector machine algorithm,and demonstrated the application in the medical environment.