Hypervisor, the core software in cloud computing, is susceptible to malicious intrusions, potentially jeopardizing overall platform security. Physical hosts within the cloud computing environment constantly face persistent attacks, and not all hosts are equipped with essential security hardware. The prolonged latency of previous software-based detection methods could not comprehensively address these cloud threats. Furthermore, the widespread deployment of security hardware incurs substantial costs for hardware-based detection. To address these challenges, we introduce Morpheus, an efficient framework ensuring hypervisor integrity in the current cloud. This framework rapidly employs software-based methods to detect malicious hosts, utilizing a subset of hosts equipped with security hardware as the Root of Trust. Efficiency is augmented through a Neural Network scheduling module, and an embedded exponential aging mechanism fortifies time-aging trust against consistent cloud threats. Evaluation results demonstrate that it can promptly identify threatened hosts with acceptable system overhead loss, solidifying its position as a robust cloud security solution.