In memory processing has gained importance as a key enabler for security sensitive cloud applications, breaking the obstacles of computation efficiency and data privacy. This thesis explores the use of modern technologies that enable in-memory processing, with the aim of improving performance and security simultaneously within cloud computing systems. We then discuss cryptographic secure processors, such as memory encryption algorithms, secure enclaves, and homomorphic computation all of which allow us to securely process sensitive data without leaking it to possible attackers. Our evaluation shows that Intel SGX-based implementations achieve 2.3× faster throughput compared to disk-based systems and can enforce cryptographic security guarantees. We analyze PIM designs that cut data movement by 67%, drastically reducing the attack vector. Performance comparisons show that optimised in-memory caching strategies, paired with hardware-accelerated isolation yield 78% of latency reduction (for transactional workoads) and 84% (for analytics queries). It also faces other challenges, such as memory overhead (around 15-23% in the case of encryption), side-channel threats, and limitations in scalability that occur because of a multi-tenant environment. We suggest a hybrid architecture that combines TEE and memory-centric computing, which can utilize 89% resource. We show that correct usage of these techniques is sufficient to enable real-time processing of classified workloads with end-to-end encryption at sub-millisecond 95th percentile query latency