Research On Differential Privacy Based On ISVM And ILVQ

Data analysis technology based on machine learning continuously collects all kinds of information,which contains a lot of private information.How to use this information to bring benefits while ensuring that the private information in the data is not leaked is a major challenge.Among the many privacy protection technologies,differential privacy is in an extremely important position and has received extensive attention and research in recent years.So far,most of the existing differential privacy algorithms belong to batch learning mode or distributed learning model,and there are few research on differential privacy in big data environment.This paper mainly studies the problem of differential privacy protection based on incremental learning in the streaming data environment,and proposes two new algorithms to achieve the purpose of privacy protection.In this paper,we propose a differential privacy algorithm based on incremental support vector machine,and prove that the algorithm satisfies the definition of differential privacy,and discuss the range of noise introduced by the algorithm.Finally,we conduct simulation experiments on real data sets,and the experimental results show that the new algorithm achieves privacy protection in streaming data environments,and the performance of the algorithm is also maintained at a high level.Compared with the previous algorithms,the new algorithm has more advantages in classification accuracy and time consumption.Then,in order to better deal with the problem of concept drift caused by unstable data distribution in the streaming data environments,this paper proposes a differential privacy algorithm based on incremental learning vector quantization,and uses real data sets to analyze the performance of the new algorithm in the streaming data environments by comparing with previous algorithms.Finally,the shortcomings of this research are discussed,and the future work is prospected.