Essay: Cloud computing

Cloud computing is the long dreamed vision of computing as a utility, where cloud customers can remotely store their data into the cloud so as to enjoy the on-demand high quality applications and services from a shared pool of configurable computing resources. Its great flexibility and economic savings are motivating both individuals and enterprises to outsource their local complex data management system into the cloud, especially when the data produced by them that need to be stored and utilized is rapidly increasing. To protect data privacy and combat unsolicited accesses in cloud and beyond, sensitive data, e.g., emails, personal health records, photo albums, tax documents, financial transactions, etc., may have to be encrypted by data owners before outsourcing to commercial public cloud; this, however, obsoletes the traditional data utilization service based on plaintext keyword search. The trivial solution of downloading all the data and decrypting locally is clearly impractical, due to the huge amount of bandwidth cost in cloud scale systems. Moreover, aside from eliminating the local storage management, storing data into the cloud serves no purpose unless they can be easily searched and utilized. Thus, exploring privacy-preserving and effective search service over encrypted cloud data is of paramount importance. Considering the potentially large number of on demand data users and huge amount of outsourced data documents in cloud, this problem is particularly challenging as it is extremely difficult to meet also the requirements of performance, system usability and scalability.
On the one hand, to meet the effective data retrieval need, large amount of documents demand cloud server to perform result relevance ranking, instead of returning undifferentiated result. Such ranked search system enables data users to find the most relevant information quickly, rather than burdensomely sorting through every match in the content collection. Ranked search can also elegantly eliminate unnecessary network traffic by sending back only the most relevant data, which is highly desirable in the “pay-as-youuse” cloud paradigm. For privacy protection, such ranking operation, however, should not leak any keyword related information. On the other hand, to improve search result accuracy as well as enhance user searching experience, it is also crucial for such ranking system to support multiple keywords search, as single keyword search often yields far too coarse result. As a common practice indicated by today’s web search engines (e.g., Google search), data users may tend to provide a set of keywords instead of only one as the indicator of their search interest to retrieve the most relevant data. And each keyword in the search request is able to help narrow down the search result further. “Coordinate matching”, i.e., as many matches as possible, is an efficient principle among such multi-keyword semantics to refine the result relevance, and has been widely used in the plaintext information retrieval (IR) community. However, how to apply it in the encrypted cloud data search system remains a very challenging task because of inherent security and privacy obstacles, including various strict requirements like data privacy, index privacy, keyword privacy, and many others (see section III-B).
In the literature, searchable encryption is a helpful technique that treats encrypted data as documents and allows a user to securely search over it through single keyword and retrieve documents of interest. However, direct application of these approaches to deploy secure large scale cloud data utilization system would not be necessarily suitable, as they are developed as crypto primitives and cannot accommodate such high service-level requirements like system usability, user searching experience, and easy information discovery in mind. Although some recent designs have been proposed to support Boolean keyword search as an attempt to enrich the search flexibility, they are still not adequate to provide users with acceptable result ranking functionality. Our early work has been aware of this problem,
and solves the secure ranked search over encrypted data with support of only single keyword query. But how to design an efficient encrypted data search mechanism that supports multikey word semantics without privacy breaches still remains a challenging open problem.
In this paper, for the first time, we define and solve the problem of multi-keyword ranked search over encrypted cloud data (MRSE) while preserving strict system-wise privacy in cloud computing paradigm. Among various multi-keyword semantics, we choose the efficient principle of “coordinate matching”, i.e., as many matches as possible, to capture the similarity between search query and data documents. Specifically, we use “inner product similarity”, i.e., the number of query keywords appearing in a document, to quantitatively evaluate the similarity of that document to the search query in “coordinate matching” principle. During index construction, each document is associated with a binary vector as a subindex where each bit represents whether corresponding keyword is contained in the document. The search query is also described as a binary vector where each bit means whether corresponding keyword appears in this search request, so the similarity could be exactly measured by inner product of query vector with data vector. However, directly outsourcing data vector or query vector will violate index privacy or search privacy. To meet the challenge of supporting such multikey word semantic without privacy breaches, we propose a basic MRSE scheme using secure inner product computation, which is adapted from a secure k-nearest neighbor (kNN) technique, and then improve it step by step to achieve various privacy requirements in two levels of threat models.
Our contributions are summarized as follows,
1) For the first time, we explore the problem of multikey word ranked search over encrypted cloud data, and establish a set of strict privacy requirements for such a secure cloud data utilization system to become a reality.
2) We propose two MRSE schemes following the principle of “coordinate matching” while meeting different privacy requirements in two levels of threat models.
3) Thorough analysis investigating privacy and efficiency guarantees of proposed schemes is given, and experiments on the real-world dataset further show proposed schemes indeed introduce low overhead on computation and communication.