Recommendation-based trust computation and rating prediction model for security enhancement in cloud computing systems

The cloud service providers need to deliver cloud services based on the service level agreement to their users. The services must be secure and privacy should be maintained for the data uploaded on the cloud domain. The cloud offers multi-tenancy; hence, the risk of unauthorized access, intrusion of malicious attackers, cloud attacks, and data redundancy increases and thus raises questions about the security of cloud storage systems. The identification of malicious users and cloud attacks like collusion attacks with the help of the computed trust based on recommendation rating has been in the present work. Rating prediction in the recommender systems helps both the service users (SU) as well as the service providers (SP) to take correct decisions for appropriate service selection and ensure secure access control for the service users. The currently existing recommendation models suffer from challenges like data sparsity and cold-start issues. Moreover, security analyses based on rating values haven't been done in past. The soft security mechanism, trust has been computed here based on recommendation rating. In this work, a novel recommendation-based trust computation and rating prediction model (RBTCRP) is proposed to identify malicious users with the help of trust computation and eventually suggest schemes that help overcome various attacks. The model also suggests mitigation techniques to overcome the issue of cold start and data sparsity. RBTCRP model generates ratings for both SUs and SPs using an amalgamated scheme of collaborative filtering model and graph-based recommendation systems. The former is used to compute similarity index computation between service users, the latter is used to construct a trusted user network set for a service user. The comparison of this work has been done with its peers under various test conditions in handling malicious and collusion attacks.