A PRIVACY-PRESERVE METHOD OF CREATING FILE DIRECTORY OF
CLOUD STORAGE SYSTEM
J
UAN-PING LIAO, WANG-PING LUO, CHAO-SHENG FENG
*
, XIA LI
U, NA WANG
College of Computer Science, Sichuan Normal University, Chengdu Sichuan 61010, China
E-MAIL: mengke_hu@163.com, exiaowg@163.com, csfenggy@126.com
Abstract:
To preserve privacy has been a difficult problem in
designing secure cloud storage system .In existing cloud
storage systems, the privacy protection of user file directories
is ignored, while the privacy protection of files is focused on. A
privacy-preserving method of creation file directory in cloud
storage systems is proposed in this paper. This method makes
cloud service providers can't identify the file directory of any
user, which is stored on cloud servers. It also solves the
consistency of the directory tree file storage through client
memory caching file directory, and solves the integrity of the
storage through keeping up the atomicity of directory tree file
operations.
Keywords:
User file directory; Preserve privacy; File storage system;
Cloud computing
1
. Introduction
Cloud computing [1,2] provides customers with service
cluster through agglomerating resources and integrating
technology at low cost. It helps the enterprise to solve the
problems of insufficient storage resources and Big Data [3].
However, to achieve cloud storage, privacy preserving must
be the first problem to be solved.
In industry, there are many cloud file storage systems,
such as GFS, HDFS. However, they don’t preserve privacy
due to storing files in clear text. Although domestic cloud
disks, such as Baidu cloud disk, 360 cloud disk, claim to
protect users' privacy, but mainly protect the file itself
through password-based access control mode or encryption,
and they did not take the privacy of directory itself into
account.
In academia, the privacy-preserve secure cloud storage
system has been a hot research topic. The Institute of
Microsoft [6] proposed an encryption-based storage
framework of public cloud. Reference [7] proposes to
encrypt the data through the virtual machine monitor. All
these methods protect the privacy through encryption
[8,9,10]. Reference [11] proposed a data storage scheme
based on secret partitioning. In this scheme, the data of a
file is firstly divided into blocks on the client side. Next, all
these data blocks together with the reconstructed
information file of the file, which is used restoring the
origin file, are uploaded to the cloud servers for storage.
For the problem that the user file directory doesn’t get
privacy preserved, this paper presents a cloud
storage-oriented file directory construction method which
can ensure the security of privacy.
2. Security risks of user file directory storage
The file storage process of existing cloud storage system
is generally as below: user upload file to the could on its
client side, then its cloud side calls Distributed File Storage
System to divide the file into blocks and store , and
Distributed File Storage System call local file system to
complete storing the file blocks. If file privacy preserving is
needed, some file privacy-preserving measures should be
taken on client side before uploading. Privacy preserving
measures mainly include encryption and secret
segmentation. While, even though privacy-preserving
measures are taken, users’ privacy can still be leaked out
through file directory of cloud storage system. Given the
existing cloud disks or cloud storage systems are mostly
developed based on Hadoop, an open source platform,
directory structure of Hadoop file system-HDFS will be
taken as an example to analyze the security of cloud storage
system file directory. As Figure 1 shows, HDFS exposes the
following risks of leaking privacy.
(1) Files and folders are named in plain text. Naming files
or folders in plaintext facilitates user to manage files, also it
facilitates the cloud server to find out the general content
and importance of files or folders at the same time, which
lead to targeted attacks.
(2)Files are stored in folder named after user. This will
result in that cloud server knows what files are included and
which user the files belong to.
(3)User file directory structure is not preserved. Cloud
172
2016 IEEE978-1-5090-0390-7/16/$31.00 ©
Proceedings of the 2016 International Conference on Machine Learning and Cybernetics, Jeju, South Korea, 10-13 July