Introduction
As sensitive data increasingly migrates to cloud environments, encryption has become a prerequisite for privacy. However, standard encryption renders data "opaque," disabling traditional search functionalities. While Searchable Encryption (SE) was developed to address this, existing Multi-keyword Ranked Search (MRSE) schemes often struggle with two critical limitations: the language barrier and a lack of semantic depth. Most systems require exact keyword matches within a single language, failing to recognize synonyms or cross-lingual equivalents (e.g., matching "doctor" in English with the Chinese "医生").
This paper introduces CLRSE, a novel framework designed to enable secure, semantically-aware, and language-agnostic multi-keyword searches. By integrating the Open Multilingual WordNet (OMW) with advanced cryptographic techniques, CLRSE allows users to perform personalized searches that transcend linguistic boundaries without compromising data confidentiality.
System Architecture
The CLRSE architecture is a multi-entity framework designed to balance security with computational efficiency. It involves Data Owners (DOs), a Cloud Platform (CP), a Computing Service Provider (CSP), and a specialized Query Transform Server (QTS).
In this model, the QTS acts as a bridge between the user and the encrypted cloud environment. It handles complex linguistic processing on the user side, allowing the cloud to focus on high-performance encrypted computations. The Key Generation Center (KGC) manages the distribution of public and secret keys across the system.
Technical Implementation
CLRSE relies on several key technical innovations to achieve its goals:
- Paillier Cryptosystem with Threshold Decryption (PCTD): Moving beyond traditional k-nearest neighbor (kNN) algorithms, CLRSE utilizes PCTD. This enables secure mathematical operations, such as addition and multiplication, directly on encrypted data. The threshold mechanism ensures that the decryption key is split between the CP and CSP; neither entity can access the underlying data independently.
- OMW Integration and Semantic Extension: The QTS leverages the Open Multilingual WordNet to transform simple user queries into comprehensive "target queries." This process includes translating keywords into multiple target languages, adding synonyms (semantic extension), and automatically calculating relevance scores based on how closely related an extended term is to the original query.
- Keyword to Unique BigInteger (K2U) Protocol: To handle arbitrary languages without requiring a massive, predefined global dictionary, the authors developed the K2U protocol. This algorithm maps multi-language keywords into unique large integers, facilitating efficient indexing and matching.
Performance Evaluation
The effectiveness of CLRSE was evaluated using a custom multilingual database of 2,048 encrypted documents and the OMW corpus via the Python Natural Language Tool-Kit (NLTK). A user feedback survey involving 100 participants confirmed high precision, with 80% of users reporting satisfaction with the semantically extended search results.
From an efficiency standpoint, while CLRSE exhibits a slight overhead for single-result queries ($k=1$), it becomes significantly more efficient than standard MRSE as the number of requested results ($k$) increases. This performance gain is attributed to the use of "language tags," which allow the system to quickly prune the search space by filtering out documents that do not match the target language before the final ranking phase.
Conclusion
CLRSE represents a significant step forward in making cloud search services both secure and user-friendly. By moving linguistic intelligence to the edge (via the QTS) and employing robust threshold cryptography, the system provides a scalable solution for multilingual data environments. Future work will focus on further optimizing the "Top-k" search protocol to reduce latency in even larger, more complex datasets.
Summary updated