Trust Assessment-Based Multiple Linear Regression for Processing Big Data Over Diverse Clouds Hadeel El-Kassabi 1,2(&) , Mohamed Adel Serhani 2 , Chafik Bouhaddioui 2 , and Rachida Dssouli 1 1 Concordia Institute for Information Systems Engineering, Concordia University, Montreal, Canada h_elkass@encs.concordia.ca, rachida.dssouli@concordia.ca 2 College of Information Technology, UAE University, Al Ain, United Arab Emirates {serhanim,chafikb,htalaat}@uaeu.ac.ae Abstract. Assessing trust of cloud providers is considered to be a key factor to discriminate between them, especially once dealing with Big Data. In this paper, we apply Multiple Linear Regression (MLR) to develop a trust model for processing Big Data over diverse Clouds. The model relies on MLR to predict trust score of different cloud service providers. Therefore, support selection of the trustworthiness provider. Trust is evaluated not only on evidenced infor- mation collected about cloud resources availability, but also on past experiences with the cloud provider, and the reputation collected from other users experi- enced with the same cloud services. We use cross validation to test the con- sistency of the estimated regression equation, and we found that the model can perfectly be used to predict the response variable trust. We also, use bootstrap scheme to evaluate the confidence intervals for each pair of variables used in building our trust model. Keywords: Trust Á Multiple Linear Regression Á Cloud Á Big Data Á Community management 1 Introduction With the abundance of cloud services sharing the market space, it becomes challenging to select the appropriate, and trustworthy cloud providers that guarantee user ’s quality preferences and ensure continuity of service provisioning especially when dealing with Big Data. Big Data processing requires trustworthy cloud provider who ensures service delivery with high QoS guarantee. The dynamic nature of cloud makes it hard to evaluate the trust of cloud providers to process Big Data as it is dynamic in nature and can be subject of continuous resource availability, high dependability, and fault tol- erance. Previous trust models are non-dynamic and lack of real-time adaptability, which makes them unsuitable in the context of Cloud and Big Data. Building trust only based reputation can be irrelevant if the users are untrustworthy or subjective. Also, © ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2018 F. Belqasmi et al. (Eds.): AFRICATEK 2017, LNICST 206, pp. 99–109, 2018. https://doi.org/10.1007/978-3-319-67837-5_10