The increasing dependence on cloud-based data processing for industrial applications, smart devices, and Cyber-Physical Systems (CPS) emphasizes the necessity to address the inherent vulnerabilities of multi-tenant cloud environments.The existing research have limited focus at the simultaneous handling of security and reliability in cloud workload management. This article proposes a novel Quantum Toffoli Learning-based Service Management (QTL-SM) model to concurrently enhance both security and reliability during cloud applications processing. The model comprises two main components: (1) a reliability management unit composed of a novel Toffoli gate-embedded Quantum Adam neural network and (2) a security management unit for detecting and mitigating malicious virtual nodes. The former unit proactively estimates resource contention-based failures of physical nodes and manages them by analyzing reliability scores. It then allocates physical nodes to maximize these scores before executing client requests. The latter unit calculates vulnerability scores for each physical node by assessing multiple risk factors to identify potential malicious activities, mitigating their impact by preemptively terminating compromised nodes and connections. This integral approach ensures client requests are allocated to the most reliable and secure computation nodes, optimizing performance and service management. The QTL-SM model was implemented and evaluated using two real-world workloads. The comparative analysis with different model versions and state-of-the-art methods demonstrated its effectiveness in failure analysis and management, resulting in a 59.7% improvement in reliability and a 51.4% reduction in malicious activities compared to models without QTL-SM.