Cloud computing provides a variety of services to the various users according to their necessity in an on-demand manner with the deliberation of minimizing the computation and cost management. Nevertheless, security problems may occur at the time of resource allocation and in its usages. Most frequently occurred security issue is co-residence attacks, it will embezzle the details from the virtual machines by enduring with them. So, in the current work various Virtual Machine Allocation Policies (VMAP) were brought-in for allotting the VMs in a better way in the servers which can eliminate the co-residence attacks. Nevertheless, in the current system, only the existing system and the VM request were deliberated, i.e., the rest of things were not considered. This issue is overcome in the proposed research work by bringing-in the new methodology called Co-Residence Attacks Prevention and Detection Mechanism (CRA-PDM). In the proposed work, optimal VM allocation using Cuckoo Search algorithm (CSA) is performed in beginning, by deliberating various QoS parameter consideration. The parameters conceived here are, reliability, cost, stability along with parameters that are conceived in this current work such as security, workload balance and the consumption of power. In the proposed work, dynamic VM allocation is performed for many requests; this is done to favor the real world environment by utilizing Improved K Means clustering algorithm. The optimal task scheduling is assured with avoidance from the co-residence attacks, so the overall performance evaluation of the proposed research work is performed in the cloud sim environment. © 2018, Institute of Advanced Scientific Research, Inc.. All rights reserved.