Research On Service Recommendation Privacy Protection Model In Big Data

Nowadays,with the rapid development of big data and explosive growth of information,personalized service recommendation system is more and more popular with users’ needs and preferences,and the application level is also more and more extensive.Existing recommendation system mainly adopts differential privacy protection method,k-anonymous,perturbation method,but the service quality is poor,low factor of safety problem is very serious,can’t meet the needs of contemporary users,so how to provide users with a large number of recommended data at the same time,speed up the recommendation system efficiency,improve service quality and to recommend the user data security is particularly important.In view of the user privacy leakage and other problems existing in the recommendation system,this paper mainly carries out the following technical research:(1)For content-based service recommendation,the ALS recommendation algorithm in the collaborative filtering algorithm of machine learning is used to make personalized service recommendation of the recommendation system through data analysis,training model,storage model and the main steps of recommendation.According to the content of the item attributes,similar to the user’s preferences and the history of the user rating,etc,to extract the user requirements and the interests of the different content attributes,the recommend server generated in the suggestion list returned to the user,but an attacker could be deduced from the recommended list of information users of sensitive information,thus put forward a kind of personalized services recommendation algorithm based on differential privacy DPk-median.The algorithm generated in the list of recommended for recommendation system server,first of all,using the k-median clustering algorithm will recommend in the data with the same attribute data clustering,according to the risk level of different clusters,adding corresponding Laplace noise mechanism,the privacy budget parameters in the same cluster is the same,the reasonable control of noise added to the experimental results show that the algorithm not only enhance the security of the system,and also improve the quality of the service is recommended and the execution efficiency of the algorithm.(2)Recommended for location-based services,the existing interests service location limitation existing in the recommended method,similar to the user and concealment of data sparseness and the problem of poor quality of service and low safety factor,this paper proposes a trajectory privacy protection service recommendation model,this model is based on preference patterns of perception,first of all,the mix-zone using the method of clustering,when service request,through the clustering quantitative Mix-user similarity in the zone,to protect the user’s personal information;Then based on the difference of privacy preference perception algorithm(PPBP)is recommended as a result,according to user privacy preference of risk assessment,through adding different size of noise evaluation results,the reasonable distribution difference privacy budget,implementation service quality has increased,the experimental results show that the algorithm not only improve the degree of users’ privacy protection,and improve the service quality of the system.