Abstract
Cloud computing storage security remains a critical challenge as users must often trust third-party providers with sensitive data. This paper proposes an optimized Homomorphic Encryption (HE) framework designed to enable secure data processing and retrieval without decryption. The approach includes a 9-step security framework, an optimized HE algorithm based on the Learning With Errors (LWE) problem with a novel "refresh" operation, and an efficient data retrieval strategy using an inverted index. Experimental results demonstrate that the proposed method significantly outperforms the standard BGV algorithm, achieving a 2x speedup in encryption and a 3x speedup in retrieval efficiency.
Introduction
The primary hurdle in cloud adoption is the "dark data" problem: while data is protected during transit and at rest, it must typically be decrypted to be searched or processed, exposing it to potential breaches. Homomorphic Encryption (HE) offers a solution by allowing mathematical operations on ciphertext, but its practical application has been limited by high computational overhead and "noise" accumulation. This research presents a high-performance HE framework tailored for cloud storage.
The 9-Step Security Framework
The authors propose a comprehensive "client encryption-cloud storage" model. This architecture manages the entire data lifecycle, ensuring that the cloud provider never accesses plaintext.
The framework integrates key distribution, client-side encryption, and proxy re-encryption. This allows authorized users to retrieve and decrypt results while the cloud only handles encrypted "express" tokens and ciphertexts.
Optimized Homomorphic Encryption
The core of the system is an HE algorithm based on the Learning With Errors (LWE) problem. To address the performance bottlenecks of traditional HE, the authors implemented a "refresh" operation. In standard HE, multiplying ciphertexts causes their mathematical complexity (dimension) and noise to grow exponentially. The refresh step utilizes key exchange and mode switching to reset the ciphertext to a manageable size, preventing the "noise explosion" that typically slows down deep computations.
Efficient Data Retrieval
To avoid the inefficiency of linear scans over encrypted databases, the system employs a specialized retrieval algorithm.
As shown in Figure 2, the authors designed an inverted index structure for ciphertexts. By using a similarity formula based on keyword frequency, the cloud can rank and return the most relevant encrypted files to the user. This allows for sub-linear search times, ensuring the system remains responsive as the dataset grows.
Performance Results
The proposed method was benchmarked against the BGV (Brakerski-Gentry-Vaikuntanathan) algorithm using CloudSim.
- Encryption Efficiency: At a 700MB key capacity, the proposed method completed encryption in under 0.7 seconds, nearly twice as fast as BGV's 1.38 seconds.
- Decryption Efficiency: Decryption was achieved in less than 0.15 seconds, a significant improvement over BGV's 0.38 seconds.
- Retrieval Speed: The most notable gain was in data retrieval. For 600 search requests, the proposed method took 4.4 seconds, whereas BGV required 13.5 seconds—a 3x performance increase.
Conclusion
By combining noise-managed homomorphic encryption with an efficient inverted index retrieval strategy, this framework provides a viable path for secure, high-performance cloud storage. The significant speedups over the BGV algorithm make it a strong candidate for real-world privacy-preserving cloud applications.
Summary updated