Cloud computing has set a trend on a worldwide stage along with the rapid growth to enhance global technology standard and market scale in recent years. For the cloud users, load balancing in data center networks initiates the necessity of reducing the downtime for migrating overloaded virtual machines. To achieve better during-task deployment, optimal physical host must be selected efficiently. Nowadays, cloud customers are facing security risks in the context of load balancing of Virtual Machines (VM) which is infrequently addressed. This research addresses this pertinent issue and provides a different perspective of studying ways to develop VM deployment strategy by reducing the probability of VM co-tenancy with their targets. This will in turn make it difficult for attackers to evaluate the strategy. A model called Optimal Physical Host for Load Balancing (OPH-LB) is proposed to find the probability with probabilistic estimation in the form of its computing capability and performance in secure multi-tenant cloud. The proposed solution is evaluated via Cloudsim 3.0.3 and compared with two existing well-known algorithms. The reported results indicate that OPH-LB outperforms in improving the makespan, throughput, performance and reduces the failure number of task deployment. The results show that OPH-LB can effectively reduce the risks and security score and upgrades the utilisation of resources, with an improvement of 42.13% in all types of analyses for the experimental data.