Research On Privacy Protection-oriented IoT Service Recommendation Method

Accompanying with the rapidly development of 5G,mobile computing and smart device,a vast amount of Internet of Things(Io T)services emerge from various areas.It turns out to be a challenge for users when selecting high quality Io T services to meet their demand.Service recommendation techniques can apply high quality services for users based on historical quality of service(Qo S)data.Io T service has shortcomings like low stability,user mobility and repeated invocation,and so on.Traditional service recommendation techniques can hardly apply to Io T service recommendation directly.Moreover,a user may choose services from different platforms while the historical quality of service data of users is stored across different platforms.However,because of data privacy concerns,service providers are worried about preserving data privacy when they share their data with each other.In view of the above-mentioned problems,this thesis conducts research from two aspects.Firstly,combined with the characteristics of the Io T services,the historical Qo S data of users at different historical moments are considered as a time series,and achieve the goal of improving the accuracy of the Io T service recommendation by improving the traditional time series forecasting method;Secondly,under the premise of ensuring the accuracy of recommendations,privacy protection methods are introduced into the recommendation of Io T services to realize the privacy protection of historical Io T service quality data.Specifically,the main research of this thesis is listed as follows:(1)Based on the fact that Io T services are usually invoked by the same user repeatedly,an Io T service recommendation technique based on metricized Autoregressive Integrated Moving Average(ARIMA)model is proposed to improve the accuracy of Io T service quality prediction.The approach treats historical Qo S data as a time series,then predict the Qo S data at next time slot with ARIMA model.In order to improve the learning efficiency of the model,by incorporating the core matrices extracted from Qo S matrices through Tucker decomposition into the classical ARIMA model,this thesis extends the ARIMA model to metricized form,and use it to predict Qo S values simultaneously.Experimental results show that the algorithm can effectively improves the accuracy of Io T service recommendation in a short time.(2)An improved cross-platform Io T service recommendation method is proposed based on Locality Sensitive Hashing with privacy preservation,making use of the historical Qo S data of the last t times in Io T.Firstly,using LSH indices of the historical Qo S data at platform from each service provider.Then predict the service quality data on each platform according to the aggregated index,and recommend the best quality service for users based on the predicted Qo S data.Experiment results demonstrate that the proposed method still has done well in recommendation accuracy,while preserving data privacy.