With the cloud-driven growth in big data, issues of data privacy and secure analytics have become more significant. With the movement of more computations to outsourced settings, mechanisms to maintain confidentiality with minimal performance costs are becoming dominant. Homomorphic Encryption (HE) is unique in that the computations can be performed over encrypted data, yet it cannot be used practically due to computational overhead, latency, and incompatibility of security and efficiency. In the given paper, the performance analysis of the advanced HE methods including Partially Homomorphic Encryption (PHE), Somewhat Homomorphic Encryption (SHE), Fully Homomorphic Encryption (FHE), and SIMD-FHE is thoroughly explored with a focus on big-data analytics on the public clouds. The suggested approach exploits such factors as optimized parameterization, non-streamed processing, and noise-resistant encoding to improve its viability computationally. The simulation reveals that SIMD-FHE reduces the encryption time by 42.3 on that of conventional FHE and also offers a throughput rate of more than 55 operations/s. The size of ciphertext expansions is reduced to 0.62x-4.1x with SIMD techniques, massively decreasing the communication expenses. The emulated reaches the minimum possible average delay of 18.6 ms, and batched modular arithmetic restrains the propagation of noise between any two multiplications to less than 0.9 bits. As well, there is a 15–30% improvement in memory and depth usage in varying workloads. The research gives measurable tradeoffs and practical recommendations on how to implement HE in massive cloud-based systems. These results establish a foundation of secure privacy-preserving analytics in areas including finance, healthcare, and public governance, beyond the scope of the practicality and trustworthiness of mapped out correspondence in untrusted cloud circumstances.