2 Jinbo Xiong et al.
1 Introduction
The explosive growth of global data and the rapidly development of heteroge-
neous multimedia and social network promote us to enter the era of heteroge-
neous multimedia data [33]. More and more enterprises and individuals utilize
the cloud platform for multimedia data storage and management to achieve the
resources sharing. And the gradually evolution of multimedia data referring
various media data types like text, audio, video [31], and animation, brings a
challenge for these cloud platforms, where a traditional media file with single
type develops into the multimedia big data [33]. In addition, the development
of the heterogeneous network and the heterogeneous cloud, which is referred
that a network connects computers and other devices with different networks,
operating systems and clouds, [19], such as D2D network of 5G [32], e-health
network [11] [12], social network, heterogeneous multimedia network and wire-
less heterogeneous network [34], facilitates multimedia data distribution and
dissemination [32]. Statistics show that the global data volume reached 8.61
ZB by the end of 2015, and expected to 2020, it will unexpectedly reach 44
ZB [35] [37]. Furthermore, the proportion of multimedia data, such as, image
and video data has exceeded 90% [27]. According to the official statistics of
each application, more than 51000 apple applications are downloaded from the
apple store per mimute, the user uploads 300 hours new videos to YouTube
per minute, in one mimute, more than 24.30 million photos are posted to
the Instagram. Therefore, facing the huge amount of multimedia data, how
to improve the storage efficiency and reduce the management expenditure in
heterogeneous networks is a critical challenge for the cloud service providers
[13].
Data compression is the direct technology to save storage space and band-
width overheads [22]. However, the different individuals may use the different
data compression technologies, the same file will generate multiple replicas in
the cloud server. To solve the above issue, the data deduplication [30] [35] has
been proposed, which is aimed at duplicating the replicas for the data sets. For
a plaintext, the cloud server adopts random sampling or extracting hash value
to check with the source data, if exists, the plaintext will not be uploaded [16].
However, multimedia data deduplication may arise serious privacy concerns
and pose new security challenges, such as privacy leakage, the side-channel
attacks and unauthorized access [14] [39].
Typically, the cloud service platforms are honest-but-curious [36], that is,
they may leak the privacy information because of the curious property or col-
lusion with the adversary. In order to prevent privacy leakage, the common
solution is to encrypt the privacy information [5] [6]. Meanwhile, for achiev-
ing data deduplication in cloud, the convergence encryption (CE) [29] and
message-locked encryption (MLE) [1] have been considered, where the encryp-
tion key is a hash value of the original file. Therefore, the identical files will
generate the same key and the same ciphertext. Based on the MLE and the
idea of key updating, Li et al. [15] and Qin et al. [25] proposed the rekeying
scheme to implement encrypted deduplication storage system. The another