Machine learning based deep job exploration and secure transactions in virtual private cloud systems

摘要

Virtual Private Cloud (VPC) is an emerging cloud environment used to provide more secure data communication. VPC provides authentic communication channel for secure communication between the cloud participants. The cloud jobs and the description of the runtime cloud events must be evaluated to provide flawless VPC service. Although VPC provides security in network services, it has to be enriched with internal and external platform level security features. In this regard, secure job service schemes ensure elimination of attacks, unauthorized jobs, improper accesses and intrusions in VPC. These irrelevant tasks (activities) can be isolated before initiating job scheduling process. Particularly, providing security for zero-trust cloud environment is more challenging task. Zero-trust cloud environment has completely vulnerable trust model on both internal and external circumstances. The proposed Machine Learning Based Secure Cloud Job Services (MLSCS) is implemented to provide multi-level security in this zero-trust cloud job servicing system. The proposed MLSCS develops Multi-Server Queue Management techniques, Reinforcement Learning based deep Q Matrix (RL-Q Matrix) techniques, Authentic VPC configuration and VPC Genetic Algorithm Network (VGAN) for establishing security practices in complex job handling system. MLSCS applies effective techniques for eliminating irrelevant cloud jobs to reduce the scheduler complexity and processor utilization. In this work, irrelevant jobs are considered as the jobs that are not appropriate for particular VPC scheduling policies and security principles (attacks). These jobs are identified through various key validation procedures and VPC policy determination procedures. These jobs are eliminated and prohibited in to job scheduler. Consequently, the legitimate jobs are securely forwarded in to job scheduler through multi-server queues. In the experimental setup, the proposed MLSCS is compared with existing schemes such as Reinforcement Learning Based Distributed Heterogeneous Servicing technique (RLDH), Cat Swarm Optimization Based Job Servicing technique (CSOS) and Fuzzy Based Security-Driven Servicing technique (FSDS). The results show the MLSCS delivers 5% to 8% optimal results than existing schemes.