Public Integrity Auditing for Shared Dynamic Cloud Data with Group User Revocation

0018-9340 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TC.2015.2389955, IEEE Transactions on Computers
1
Public Integrity Auditing for Shared Dynamic
Cloud Data with Group User Revocation
Tao Jiang, Xiaofeng Chen, and Jianfeng Ma
3
Abstract—The advent of the cloud computing makes storage out-
sourcing become a rising trend, which promotes the secure remote
data auditing a hot topic that appeared in the research literature.
Recently some research consider the problem of secure and efficient
public data integrity auditing for shared dynamic data. However,
these schemes are still not secure against the collusion of cloud
storage server and revoked group users during user revocation in
practical cloud storage system. In this paper, we figure out the
collusion attack in the exiting scheme and provide an efficient public
integrity auditing scheme with secure group user revocation based
on vector commitment and verifier-local revocation group signature.
We design a concrete scheme based on the our scheme definition.
Our scheme supports the public checking and efficient user revoca-
tion and also some nice properties, such as confidently, efficiency,
countability and traceability of secure group user revocation. Finally,
the security and experimental analysis show that, compared with its
relevant schemes our scheme is also secure and efficient.
Index Terms—Public integrity auditing, dynamic data, victor commit-
ment, group signature, cloud computing.
1 INTRODUCTION
The development of cloud computing motivates en-
terprises and organizations to outsource their data
to third-party cloud service providers (CSPs), which
will improve the storage limitation of resource con-
strain local devices. Recently, some commercial cloud
storage services, such as the simple storage service
(S3) [1] on-line data backup services of Amazon and
some practical cloud based software Google Drive
[2], Dropbox [3], Mozy [4], Bitcasa [5], and Memopal
[6], have been built for cloud application. Since the
cloud servers may return an invalid result in some
cases, such as server hardware/software failure, hu-
man maintenance and malicious attack [7], new forms
of assurance of data integrity and accessibility are
required to protect the security and privacy of cloud
user’s data.
To overcome the above critical security challenge
of today’s cloud storage services, simple replication
and protocols like Rabin’s data dispersion scheme [8]
are far from practical application. The formers are
• Tao Jiang, Xiaofeng Chen and Jianfeng Ma are with the State Key
Laboratory of Integrated Service Networks(ISN), Xidian University,
P.R. China, e-mail: jiangt2009@gmail.com, xfchen@xidian.edu.cn, and
jfma@mail.xidian.edu.cn.
not practical because a recent IDC report suggests
that data-generation is outpacing storage availability
[9]. The later protocols ensure the availability of data
when a quorum of repositories, such as k-out-of-n of
shared data, is given. However, they do not provide
assurances about the availability of each repositories,
which will limit the assurance that the protocols can
provide to relying parties.
For providing the integrity and availability of re-
mote cloud store, some solutions [10], [11] and their
variants [12], [13], [14], [15], [16], [17], [18] have
been proposed. In these solutions, when a scheme
supports data modification, we call it dynamic scheme,
otherwise static one (or limited dynamic scheme, if
a scheme could only efficiently support some speci-
fied operation, such as append). A scheme is publicly
verifiable means that the data integrity check can be
performed not only by data owners, but also by any
third-party auditor. However, the dynamic schemes
above focus on the cases where there is a data owner
and only the data owner could modify the data.
Recently, the development of cloud computing
boosted some applications [19], [20], [21], where the
cloud service is used as a collaboration platform. In
these software development environments, multiple
users in a group need to share the source code, and
they need to access, modify, compile and run the
shared source code at any time and place. The new
cooperation network model in cloud makes the re-
mote data auditing schemes become infeasible, where
only the data owner can update its data. Obviously,
trivially extending a scheme with an online data
owner to update the data for a group is inappro-
priate for the data owner. It will cause tremendous
communication and computation overhead to data
owner, which will result in the single point of data
owner. To support multiple user data operation, Wang
et al. [22] proposed a data integrity based on ring
signature. In the scheme, the user revocation problem
is not considered and the auditing cost is linear to
the group size and data size. To further enhance the
previous scheme and support group user revocation,
Wang et al. [23] designed a scheme based on proxy
re-signatures. However, the scheme assumed that the
private and authenticated channels exist between each

2
pare of entities and there is no collusion among them.
Also, the auditing cost of the scheme is linear to the
group size. Another attempt to improve the previous
scheme and make the scheme efficient, scalable and
collusion resistant is Yuan and Yu [24], who designed
a dynamic public integrity auditing scheme with
group user revocation. The authors designed polyno-
mial authentication tags and adopt proxy tag update
techniques in their scheme, which make their scheme
support public checking and efficient user revocation.
However, in their scheme, the authors do not consider
the data secrecy of group users. It means that, their
scheme could efficiently support plaintext data update
and integrity auditing, while not ciphertext data. In
their scheme, if the data owner trivially shares a
group key among the group users, the defection or
revocation any group user will force the group users
to update their shared key. Also, the data owner does
not take part in the user revocation phase, where
the cloud itself could conduct the user revocation
phase. In this case, the collusion of revoked user and
the cloud server will give chance to malicious cloud
server where the cloud server could update the data
as many time as designed and provide a legal data
finally. To the best of our knowledge, there is still
no solution for the above problem in public integrity
auditing with group user modification.
The deficiency of above schemes motivates us to ex-
plore how to design an efficient and reliable scheme,
while achieving secure group user revocation. To
the end, we propose a construction which not only
supports group data encryption and decryption dur-
ing the data modification processing, but also real-
izes efficient and secure user revocation. Our idea
is to apply vector commitment scheme [25] over the
database. Then we leverage the Asymmetric Group
Key Agreement (AGKA) [26] and group signatures
[27] to support ciphertext data base update among
group users and efficient group user revocation re-
spectively. Specifically, the group user use the AGKA
protocol to encrypt/decrypt the share database, which
will guarantee that a user in the group will be able
to encrypt/decrypt a message from any other group
users. The group signature will prevent the collusion
of cloud and revoked group users, where the data
owner will take part in the user revocation phase and
the cloud could not revoke the data that last modified
by the revoked user.
1.1 Our Contribution
In this paper, we further study the problem of con-
struing public integrity auditing for shared dynamic
data with group user revocation. Our contributions
are three folds:
1) We explore on the secure and efficient shared
data integrate auditing for multi-user operation
for ciphertext database.
Data
Owner
Group Users Third Party Auditor
Revoked
Group Users
Public Parameters
Data Integrity Auditing Results
Cloud Storage Server
Figure 1. The cloud storage model
2) By incorporating the primitives of victor com-
mitment, asymmetric group key agreement and
group signature, we propose an efficient data
auditing scheme while at the same time provid-
ing some new features, such as traceability and
countability.
3) We provide the security and efficiency analysis
of our scheme, and the analysis results show that
our scheme is secure and efficient.
1.2 Organization
The rest of this paper is organized as follows: In
section 2, we describe the problem formulation. In
section 3, we present the used preliminaries. Then,
we provide the detail of our scheme in section 4. We
conduct the security and efficiency analysis in section
5 and section 6 and leave the related works in section
7. Finally, we show our conclusion in section 8.
2 PROBLEM FORMULATION
In this section, we first describe the cloud storage
model of our system. Then, we provide the threat
model considered and security goals we want to
achieve.
2.1 Cloud Storage Model
In the cloud storage model as shown in Figure 1, there
are three entities, namely the cloud storage server,
group users and a Third Part Auditor (TPA).
Group users consist of a data owner and a number
of users who are authorized to access and modify the
data by the data owner. The cloud storage server is
semi-trusted, who provides data storage services for
the group users. TPA could be any entity in the cloud,
which will be able to conduct the data integrity of the
shared data stored in the cloud server. In our system,
the data owner could encrypt and upload its data to
the remote cloud storage server. Also, he/she shares
the privilege such as access and modify (compile and
execute if necessary) to a number of group users.
The TPA could efficiently verify the integrity of the

3
m+Tag m' +Tag' m' +Tag"
Data last modified
by user Eve
Data last modified
by cloud server
Legitimate data authorized
by the data owner
Cloud server
and Eve
Cloud server
Modify Revoke
Figure 2. Security problem of server proxy group user
revocation
data stored in the cloud storage server, even the data
is frequently updated by the group users. The data
owner is different from the other group users, he/she
could securely revoke a group user when a group
user is found malicious or the contract of the user
is expired.
2.2 Threat Model and Security Goals
Our threat model considers two types of attack:
1) An attacker out side the group (include the
revoked group user cloud storage server) may
obtain some knowledge of the plaintext of the
data. Actually, this kind of attacker has to at
lease break the security of the adopted group
data encryption scheme.
2) The cloud storage server colludes with the re-
voked group users, and they want to provide a
illegal data without being detected.
Actually, in cloud environment, we assume that
the cloud storage server is semi-trusted. Thus, it is
reasonable that a revoked user will collude with the
cloud server and share its secret group key to the
cloud storage server. In this case, although the server
proxy group user revocation way [24] brings much
communication and computation cost saving, it will
make the scheme insecure against a malicious cloud
storage server who can get the secret key of revoked
users during the user revocation phase. Thus, a ma-
licious cloud server will be able to make data m, last
modified by a user that needed to be revoked, into a
malicious data m′
. In the user revocation process, the
cloud could make the malicious data m′
become valid.
To overcome the problems above, we aim to achieve
the following security goals in our paper:
1) Security. A scheme is secure if for any database
and any probabilistic polynomial time adver-
sary, the adversary can not convince a verifier
to accept an invalid output.
2) Correctness. A scheme is correct if for any
database and for any updated data m by a valid
group user, the output of the verification by an
honest cloud storage server is always the value
m. Here, m is a ciphertext if the scheme could
efficiently support encrypted database.
3) Efficiency. A scheme is efficient if for any data,
the computation and storage overhead invested
by any client user must be independent of the
size of the shared data.
4) Countability. A scheme is countable, if for any
data the TPA can provide a proof for this misbe-
havior, when the dishonest cloud storage server
has tampered with the database.
5) Traceability. We require that the data owner
is able to trace the last user who update the
data (data item), when the data is generated
by the generation algorithm and every signature
generated by the user is valid.
3 PRELIMINARIES
Our scheme makes use of bilinear groups. The se-
curity of the scheme depends on the Strong Diffie-
Hellman assumption and the Decision Linear assump-
tion. In this section, we review the definitions of
bilinear groups and the complexity assumption.
3.1 Bilinear Groups
We review a few concepts related to bilinear maps,
which follow the notation of [28]. Let G1 and G2 be
two multiplicative cyclic groups of prime order p, g1
is a generator of G1 and g2 is a generator of G2. ψ is
an efficiently computable isomorphism from G2 to G1
with ψ(g2) = g1, and e : G1 × G2 → GT is a bilinear
map with the following properties:
1) Computability: there exits an efficiently com-
putable algorithm for computing map e;
2) Bilinearity: for all u ∈ G1, v ∈ G2 and a, b ∈ Zp,
e(ua
, vb
) = e(u, v)ab
;
3) Non-degeneracy: e(g1, g2) = 1.
3.2 Complexity Assumption
The security of our scheme relies on the difficulty of
some problems: the Strong Diffie-Hellman problem,
the Decision Linear problem, and the Computational
Diffie-Hellman problem. We describe these problems
as follows.
Definition 1. q-Strong Diffie-Hellman problem. Let
G1, G2 be cyclic group of prime order p, where
possibly G1 = G2. Let g1 be a generator of G1
and g2 be a generator of G2. Given a (q + 2) −
tuple(g1, g2, gγ
2 , g
(γ2
)
2 , ..., g
(γq
)
2 ) as input, output a pair
(g
1/(γ+x)
1 , x) where x ∈ Z∗
p.
The assumption could be used to construct short
signature scheme without random oracles [29]. The
assumption has properties similar to the Strong-RSA
assumption [30] and the properties are adopted for
building short group signature in our scheme.
Definition 2. Decision Linear problem. Let g1 be a
generator of G1, and G1 be a cyclic group of prime
order p. Given u, v, h, ua
, ub
,uc
∈ G1 as input, output
yes if a + b = c and no otherwise.
Boneh et al. [31] introduced the Decision Linear
assumption and they proved that the problem is
intractable in generic bilinear groups.

4
Definition 3. Square Computational Diffie-Hellman
(Square-CDH) problem. With g ∈ G1 as above, given
(g, gx
) for x ∈R Zp as input, output gx2
.
It has been proved that the Square-CDH assump-
tion is equivalent to the classical CDH assumption
[32], [33].
3.3 Vector Commitment
Commitment is a fundamental primitive in cryptog-
raphy and it plays an important role in security pro-
tocols such as voting, identification, zero-knowledge
proof, etc. The hiding property of commitment re-
quires that it should not reveal information of the
committed message, and the binding property re-
quires that the committing mechanism should not
allow a sender to change his/her mind about the
committed message.
Recently, Catalano and Fiore [25] put forward a new
primitive called Vector Commitment. Vector Com-
mitment satisfies position binding that an adversary
should not be able to open a commitment to two
different values at the same position, and the Vector
Commitment is concise, which means that the size
of the commitment string and its openings have to
be independent of the vector length. We provide
the formal definition of Vector Commitment [25] as
follows.
Definition 4. (Vector Commitment) A vector commit-
ment scheme is a collection of six polynomial-time
algorithms (VC.KeyGen, VC.Com, VC.Open, VC.Ver,
VC.Update, VC.ProofUpdate) such that:
VC.KeyGen(1k
, q). Given the security parameter k
and the size q of the committed vector (with q =
poly(k)), the key generation outputs some public pa-
rameters pp.
VC.Compp(m1, ..., mq). On input a sequence of q
messages m1, ..., mq ∈ M (M is the message space
) and the public parameters pp, the committing algo-
rithm outputs a commitment string C and an auxiliary
information aux.
VC.Openpp(m, i, aux). This algorithm is run by the
committer to produce a proof i that m is the i-th
committed message. In particular, notice that in the
case when some updates have occurred the auxiliary
information aux can include the update information
produced by these updates.
VC.Verpp(C, m, i, Λi). The verification algorithm ac-
cepts (i.e., it outputs 1) only if Λi is a valid proof
that C was created to a sequence m1, ..., mq such that
m = mi.
VC.Updatepp(C, m, m′
, i). This algorithm is run by
the committer who produces C and wants to update
it by changing the i-th message to m′
. The algorithm
takes as input the old message m, the new message
m′
and the position i. It outputs a new commitment
C’ together with an update information U.
VC.ProofUpdatepp(C, Λj, m′
, i, U). This algorithm
can be run by any user who holds a proof Λj for
some message at position j w.r.t. C, and it allows
the user to compute an updated proof Λ′
j (and the
updated commitment C′
) such that Λ′
j will be valid
with regard to C′
which contains m′
as the new
message at position i. Basically, the value U contains
the update information which is needed to compute
such values.
The primitive of verifiable database with efficient
update based on vector commitment is useful to solve
the problem of verifiable data outsourcing. Recently,
Chen et al. [34], [35] figured out that the basic vector
commitment scheme suffers from forward automatic
update attack and backward substitution update at-
tack. They also proposed a new framework for ver-
ifiable database with efficient update from vector
commitment, which is not only public verifiable for
dynamic outsourced data but also secure against the
two attacks. The solution in their scheme is easy to
apply in our scheme, which will overcome the attacks
they figured out in our scheme.
3.4 Group Signature with User Revocation
We present the formal definition of group signatures
with verifier-local revocation [27] as follows.
Definition 5. A verifier-local group signature scheme
is a collection of three polynomial-time algorithms
(VLR.KeyGen, VLR.Sign, VLR.Verify), which behaves as
follows:
VLR.KeyGen(n). This randomized algorithm takes
as input a parameter n, the number of mem-
bers of the group. It outputs a group public key
gpk, an n-element vector of user keys gsk =
(gsk[1], gsk[2], ..., gsk[n]), and an n-element vector of
user revocation tokens grt, similarly indexed.
VLR.Sign(gpk, gsk[i], M). This randomized algo-
rithm takes as input the group public key gpk, a
private key gsk[i], and a message M ∈ {0, 1}
∗
, and
returns a signature σ.
VLR.Verify(gpk, RL, σ, M). The verification algo-
rithm takes as input the group public key gpk, a set of
revocation tokens RL (whose elements form a subset
of the elements of grt), and a purported signature σ
on a message M. It returns either valid or invalid. The
latter response can mean either that σ is not a valid
signature, or that the user who generated it has been
revoked.
4 SCHEME CONSTRUCTION
In this section, we provide the formal definition of
our scheme according to the definition in [23], [24].
Then, we design the concrete scheme based on our
definition.

5
4.1 New Framework
We consider the database DB as a set of tuple (x, mx),
where x is an index and mx is the corresponding
value. Informally, a public integrity auditing scheme
with updates allows a resource-constrained client to
outsource the storage of a very large database to a
remote server. Later, the client can retrieve and update
the database records stored in the server and publicly
audit the integrity of the updated data.
According to previous researches, the proposed
framework of our public integrity auditing for shared
dynamic cloud data with secure group user revocation
is given as follows:
Setup(1k
, DB):
Let the database be DB = (i, mi) for 1 ≤ i ≤ q and
the database is shared by a group of n users with only
one data owner.
1) The data owner run the key generation algo-
rithm of vector commitment to obtain the public
parameters pp ← VC.KeyGen(1k
, q).
2) Run the key generation of verifier-local revo-
cation to obtain the user keys and revocations
(gsk, grt) ← VLR.KeyGen(1k
, n), where gsk =
(gsk[1], gsk[2]...gsk[n]) and an n-element vector
of user revocation tokens grt.
3) Run the computing algorithm to compute com-
mitment and auxiliary information (C, aux) ←
VC.Compp(c1, ..., cq). Let the current database
modifier be group user s(0 ≤ s ≤ n − 1), and
(gsk[s], gpk) be the secret/public key pair of the
group user. Let Ct
= V C.ComPP(ct
1, ..., ct
q) be the
commitment on the latest database vector, where
t is a counter with 0 as its initial value.
4) Run the signing algorithm over the commit-
ment C. Specially, for the t-th time the group
user s(0 ≤ s ≤ n − 1), whose secret
key is gsk[s], compute and output a signa-
ture σt
← VLR.Sign(gpk, gsk[s], {C(t − 1), Ct
, t}).
Then, sends the signature σt
to the cloud storage
server. If σt
is valid, then the server computes
C(t) = σt
·Ct
. Also, the cloud storage server adds
the information of Σ(t) = (C(t − 1), Ct
, t, σt
) to
aux.
5) Finally, set public key parameter PK =
(pp, gpk, C(t − 1), C(t)).
Query(PK, PP, aux, DB, i):
1) A group user run the opening algorithm to com-
pute a proof Λi ← VC.Openpp(ci, i, aux), where Λi
is the proof of the i-th committed message and
return τ = (ci, Λi, Σ(t)).
Verify(PK, RL, i, τ):
1) Parse τ = (ci, Λi, Σ(t)). If the signature
is valid after running the algorithm
VLR.Verify(gpk, RL, Σ(T )). Then, run the
verification algorithm of vector commitment
{0, 1} ← VC.Verpp(C(t), σt
, ci, i, Λi). The
algorithm accepts when it output 1, which
means that Λi is a valid proof that Ct
was
created by a sequence c1, ...cq, such that c = ci.
Otherwise, return an error ⊥.
Update(i, τ):
1) A group user first queries and verifies the
database to make sure the current database is
valid. More precisely, the group user obtain
τ ←Query(PK, PP, aux, DB, i) and check that
Verify(PK, i, τ) = mi.
2) Run the update algorithm over the new data and
output the updated commitment and the update
information (C′
, U) ← VC.Update(C, m, m′
, i).
ProofUpdate(C, Λj, c′
i, i, U):
1) A third part auditor can first verify that,
compared with the stored counter t, the lat-
est counter equals t + 1. Then, run the
proof of update algorithm of vector commit-
ment to compute an update proof Λj ←
VC.ProofUpdatepp(C, Λj, m′
i, i, U) for the mes-
sage at position j, such that Λj is valid with
respect to C′
which contains m′
as the new
message at position j. Here, U = (m, m′
, i) is
the update information.
2) Verify the commitment C′
, and its corresponding
proof Λi is also valid over message m′
i.
UserRevocation(PK, i, τ):
1) The third part auditor can run the veri-
fication algorithm of verifier-local revocation
and return either valid or invalid {0, 1} ←
VLR.Verify(gpk, RL, σ, M). Here, RL are a set of
revocation tokens.
4.2 A Concrete Scheme
In this section, we provide a concrete scheme from
vector commitment [25] and verifier-local revocation
group signature [27].
Setup(1k
, DB):
Let k be a security parameter and DB = (i, mi) for
1 ≤ i ≤ q be the database. The database DB = (i, mi)
is shared by a group of n users with only one data
owner. The message space is M = Zp.
1) Let G, GT be two bilinear groups of prime
order p equipped with a bilinear map
e : G × G → GT , and g be a random generator
of G. Randomly choose z1, ..., zq ←R Zp.
For all i = 1, ..., q, set hi = gzi
. For all
i, j = 1, ..., q, i = j, set hi,j = gzizj
. The data
owner runs the key generation algorithm
of vector commitment VC.KeyGen(1k
, q)
to obtain the public parameters PP =
(p, q, G, GT , H, g, ({hi})i∈[q], {hi,j}i,j∈[q],i=j) and
the message space M= Zp. By using a collision-
resistant hash function H : {0, 1}∗
→ Zp, our
scheme can be easily extended to support
arbitrary messages in {0, 1}∗
.
2) Run the key generation of verifier-local revo-
cation VLR.KeyGen(1k
, n). Let G1, G2 be cyclic

6
group of prime order p, and g1 be a generator
of G1 and g2 be a generator of G2. Consider
bilinear groups (G1, G2) with isomorphism ψ,
where g1 ← ψ(g2). Select γ ←R Z∗
p and set
w = gγ
2 . For each user, generate an SDH tuple
(Ai, xi) by selecting xi
R
← Z∗
p such that γ+xi = 0,
and setting Ai ← g
1/(γ+xi)
1 . Then, set the group
public key gpk = (g1, g2, w). The private key is a
tuple gsk[i] = (Ai, xi). The revocation token cor-
responding to a user’s secret key is grt[i] = Ai.
Finally, the algorithm outputs (gpk, gsk, grk). γ
is only known to the private-key issuer (the data
owner).
3) Run the computing algorithm
VC.Compp(m1, ..., mq) to compute commitment
C = hm1
1 , hm2
2 , ..., h
mq
q and auxiliary information
aux = (m1, ..., mq).
4) Employ hash functions H0 and H as random
oracles, with respective ranges G2
2 and Zp. For
the t-th time data updating, run the signing
algorithm VLR.Sign(gpk, gsk[i], {C(t − 1), Ct
, t})
over the commitment. Assume that the input
message is {C(t − 1), Ct
, t} ∈ {0, 1}∗
. Then, pick
a random nonce r ←R Zp and obtain generators
(û, ˆv) ← H0(gpk, {C(t − 1), Ct
, t} , r) ∈ G2
2 and
compute their images in G1 with u ← ψ(û)
and v ← ψ(ˆv) . Select an exponent α ←R Zp
and compute T1 ← uα
and T2 ← Aivα
. Set
δ ← xiα ∈ Zp. Pick blinding values rα, rx, and
rδ ←R Zp. Compute helper values R1 ← urα
,
R2 ← e(T2, g2)rx
· e(v, w)−rα
· e(v, g2)−rδ
and
R3 ← T rx
1 ·u−rδ
. Compute a challenge value c ←
H(gpk, (C(t − 1), Ct
, t), r, T1, T2, R1, R2, R3) ∈ Zp
using H. Compute sα = rα + cα, sx = rx + cxi,
and sδ = rδ + cδ ∈ Zp. Finally, output a signa-
ture σt
← (r, T1, T2, c, sα, sx, sδ). Then, sends the
signature σt
to the cloud storage server. If σt
is
valid, then the server computes C(t) = σt
· Ct
.
Also, the cloud storage server adds the informa-
tion of Σ(t) = (C(t − 1), Ct
, t, σt
) to aux.
5) Set public key parameter PK = (pp, gpk, C(t −
1), C(t)).
Query(PK, pp, aux, DB, i):
1) We assume that the current public key is PK =
(PP, gpk, C(t − 1), C(t)). A user runs the open-
ing algorithm VC.Openpp(ct
i, i, aux) to compute
a proof Λt
i =
q
j−1,j=i h
mt
j
i,j = (
q
j=1,j=i h
mt
j
j )zi
of the i-th committed message and return τ =
(mt
i, Λt
i, Σ(t)).
Verify(PK, i, τ):
1) On input a group public key gpk, a purported
signature σt
, and the message {C(t − 1), Ct
, t},
the auditor first verify whether the signature is
valid.
2) If τ = (mt
i, Λi, Σ(t)), run the verification
algorithm of vector commitment
VC.Verpp(Ct
i , ct
i, i, Λt
i) to verify that the equation
e(Ct
/h
mt
i
i , hi)
?
= e(Λt
i, g) holds. The algorithm
accepts when it outputs 1, which means that
Λt
i is a valid proof that Ct
was created to a
sequence m1, ...mq, such that m = mi.
Update(i, τ):
1) A group user first queries and verifies the
database to make sure the current database is
valid.
2) If the user wanted to update mi to
m′
i, the user runs the update algorithm
VC.Update(C, m, m′
, i) and outputs the updated
commitment C′
= C · hm′
−m
i and the updated
information U = (m, m′
, i).
ProofUpdate(C, Λj, m′
i, i, U):
1) The third part auditor can run the
proof of update algorithm of vector
commitment to compute an update proof
Λj ← VC.ProofUpdatepp(C, Λj, m′
, i, U) for the
message at position j, such that Λj is valid with
respect to C′
which contains m′
as the new
message at position j.
2) For the auditor who owns a proof Λj, the auditor
uses the update information U = (m, m′
, i) to
generate the proof of update. If i = j, compute
the updated commitment C′
= C · hm′
−m
i and
the updated proof is Λ′
j = Λj · (hm′
−m
i )zj
=
Λj · hm′
−m
j,i ; If i = j, compute the updated
commitment C′
= C·hm′
−m
i while do not change
the proof Λi. Verify the commitment C′
and
its corresponding proof Λi is also valid over
message m′
i.
UserRevocation(PK, i, τ):
1) To verify the validity of the signature, the
auditor need to conduct the signature
check. The third part auditor runs the
verification algorithm of verifier-local revocation
VLR.Verify(gpk, RL, σ, M), M = C(t − 1), Ct
, t.
More precisely, compute û and ˆv and their
image u ← ψ(û) and v← ψ(ˆv) in G1.
Derive ˜R1 ← usα
/T c
1 , ˜R2 ← e(T2, g2)sx
·
e(v, w)−sα
· e(v, g2)−sδ
· (e(T2, w)/e(g1, g2))c
and
˜R3 ← T sx
1 · u−sδ
. Check the challenge that
c
?
= H(gpk, (C(t − 1), Ct
, t), r, T1, T2, ˜R1, ˜R2, ˜R3)
and return either valid or invalid. Then, conduct
the revocation check.
2) For each element A ∈ RL, check whether A is
encoded in (T1, T2) by checking if e(T2/A, û)
?
=
e(T1, ˆv). If no element of RL is encoded in
(T1, T2), the signer of σ has not been revoked.
Here, RL is a set of revocation tokens.
4.3 Supporting Ciphertext Database
In cloud storage outsourcing environment, the out-
sourced data is usually encrypted database, which is
usually implicitly assumed in the exiting academic

7
research. Actually, our scheme could support the au-
diting of database of both plaintext and ciphertext
database. However, it is not straightforward to extend
a scheme to support encrypted database.
In order to achieve the confidentiality of the data
record mx, the client can use his/her secret key to
encrypt each mx using a encryption scheme. When
there is only one user (data owner) in the group, the
user only needs to choose a random secret key and
encrypt the data using a secure symmetric encryption
scheme. However, when the scheme needs to support
multi-user data modification, while at the same time
keeping the shared data encrypted, a shared secret key
among group users will result in single point failure
problem. It means that any group user (revoked or
leave) leak the shared secret key will break the confi-
dentiality guarantee of the data.
To overcome the above problem, we need to adopt
a scheme, which could support group users data
modification. Luckily, Wu et al. [26] designed an
Asymmetric Group Key Agreement scheme (ASGKA).
The scheme has a nice property that, instead of a
common secret key, only a shared encryption key
is negotiated in an ASGKA protocol. Also, in the
scheme, the public key can be simultaneously used
to verify signatures and encrypt messages while any
signature can be used to decrypt ciphertext under this
public key. Using the bilinear pairings, the authors
instantiate a one-round ASGKA protocol tightly re-
duced to the decision Bilinear Diffie-Hellman Expo-
nentiation (BDHE) assumption in the standard model.
Thus, according to the ASGKA protocol, we consider
the case of encrypted database (x, cx), where x is an
index and cx is the corresponding cipher value.
We provide the detailed changes upon our scheme
to support encrypted database.
1) In the Setup phase, the scheme has to run the
key agreement of ASGKA for the group users.
Then, the database DB = (i, mi) is encrypted
by the group key gpk of data owner. Finally, the
stored database is a ciphertext database DB =
(i, ci).
2) In the second step of the Update phase, a group
user firstly decrypts the record ci using the AS-
GKA secret key gsk[∗] to get plaintext database
DB = (i, mi). Then, update the data to m′
i, and
later encrypt the data with the public key gpk
of ASGKA scheme to get the new encrypted
database DB = (i, c′
i).
4.4 Probabilistic Detection
Actually, the position binding property of vector com-
mitment of the scheme allows the cloud storage server
to prove the data item correctness of certain position.
Ateniese et. al. [10] figured out that the sampling
ability greatly reduces the overhead on the server
and provides high detection probability of server
misbehavior. Then, among the q data items, we as-
sume that the third part auditor randomly select x
items out of the q-block item database as the target
item. In the database, only y items of the database
are incorrect. Then, if x, y and q satisfy the specific
relationship, the third part auditor could provide a
high possession detection ability over the database.
The result is interesting that when y is a fraction of
the total item number q, the detection probability of
server misbehavior is a constant amount of item. For
example, if y = 1% of q, then the third part auditor
asks for 460 blocks and 300 blocks in order to achieve
the detection probability of at least 99% and 95%,
respectively.
5 ANALYSIS OF OUR SCHEME
Our scheme is designed to solve the security and
efficiency problems of public data integrity auditing
with multi-user modification, where the data has to
be encrypted among a dynamic group and any group
user can conduct secure and verifiable data update
when necessary.
Some basic tools have been used to construct our
scheme. Thus we assume that the underlying building
blocks are secure, which include the vector commit-
ment, group signature, and asymmetric group key
agreement scheme. Based on this assumption, we
show that our scheme is secure with respect to the
following security analysis.
5.1 Security of Our Scheme
• Security of Our Scheme. Actually, Wu et al.
[26] instantiated a one-round ASGKA scheme
tightly reduced to the decision Bilinear Diffie-
Hellman Exponentiation assumption in the stan-
dard model. Also, the security of adopted group
signature scheme has been proved to be secure
in the random oracle model. The security of the
scheme is based on the strong Diffie-Hellman
assumption and the Decision Linear assumption
in bilinear groups as defined in Definition 1 and
Definition 2. Thus, if we assume the two building
blocks of our scheme is secure, then our scheme
can be proven to be secure similar to [25].
If we assume there exists a polynomial-time ad-
versary A that has a non-negligible advantage ǫ
in the experiment for some initial database. Then
we can use the adversary to build an efficient
algorithm to break the Squ-CDH assumption in
Definition 3. It means that the algorithm takes
a tuple g, ga
as input and output ga2
. Triv-
ially, suppose that (î, ˆτ) in a experiment where
τ = (ct
i, Λt
i, Σ(t)), the verify query output a value
ˆc =⊥, ˆc = ct
i and e(Ct
, hî) = e(h
ct
î
î
, hî)e(Λt
î
, g) =
e(hˆc
î
, hî)e(ˆΛ, g). If the simulation does not fail, we
have hî = ga
. Then, the adversary can compute

8
Table 1
Performance evaluation and comparison
Scheme Scheme [23] Scheme [24] Our Scheme
Query - - (q − 1)(Mul + Exp)
Verify 2Exp + Mul + 2Pair + Hash Exp + 2Pair 7Pair + Mul + 9Exp + 5Hash
Update - (s + 2)Exp + (s + 1)Mul s(Mul + Exp)*
ProofUpdate - sExp + (s + 1)Mul + cPair 2s(Mul + Exp)ors(Mul + Exp)*
UserRevocation (c+d)Exp+(c+3d)Mul+(d+1)Pair+
cHash
c(Pair + Exp) z(Mul + 2Pair)
* In our scheme, we do not need to verify up to s elements each time. For comparison, we assume that our scheme conducts a
verification of s data items here.
ga2
= (
ˆΛ
Λt
î
)(ct
î
−ˆc)−1
and the success probability of
the algorithm built by the adversary is ǫ/q.
• Correctness of Our Scheme. If the server is
assumed to be honest, then the proof τ =
(ct
i, Λt
i, Σ(t)), where Λt
i = i=j,1≤j≤q h
ct
j
i,j. Since
C(t)/Hth
ct
i
i = Ct
/h
ct
i
i = i=j,1≤j≤q h
ct
j
i,j, we have
e(Ct/Hth
ct
i
i , hi) = e(Λt
i, g). Thus, the verification
algorithm always output ct
i.
• Efficiency of Our Scheme. It is trivial that, except
for the one time setup, the computational and
storage overhead in our scheme invested by the
group users are independent of the size of the
data. More precisely, to verify the validity of the
scheme, the verify algorithm run by the client
requires only pairings and exponentiation in G.
Also, in the update algorithm, the computation
overhead of the client is independent of the size
of data. The storage overhead of a group user is
also independent of the size of data. We will pro-
vide the detail efficiency and experiment analysis
in the full version of this paper.
• Countability of Our Scheme. Sine the update
counter t is a public parameter, given the proof
with the counter t′
, the client will firstly compare
it with the public latest counter. if t′
= t, then the
auditor verifies the corresponding signature σt′
over t′
. Otherwise, if t′
= t the group auditor will
reject the current result and report the malicious
activity of the cloud storage server.
• Traceability of Our Scheme. The traceability of
our scheme is based on the traceability of the
adopted group signature. In the theorem 2 of
reference [25], the authors provide the formal
proof of the traceability of the group signature
adopted. It means that if SDH is hard on (G1,G2),
the group signature scheme is traceable.
6 PERFORMANCE EVALUATION
In this section, we provide both the numerical and the
experimental analysis of our scheme and conduct the
computation time cost comparison with [23] and [24].
6.1 Numerical Analysis
In this section, we conduct the numerical analysis of
our scheme and compare the scheme with references
[23] and [24].
First of all, all of the three schemes require one-time
expensive computational effort in the Setup phase.
Then, our scheme is secure against the collusion attack
of the cloud storage server and the revoked users in
the efficient scheme [24], and it is also efficient, since
the computational resources invested by the client is
independent on the size of the database. The reason
is that, most of the expensive computation overhead
is outsourced to the cloud storage server. Finally,
the cloud storage server store all the database and
its relevant materials. Thus, except some private key
materials, the group users do not require to store any
data locally.
We provide the time cost simulation for our scheme
in different phases and the Table 1 presents the nu-
merical analysis of computation of our scheme and
two other schemes related. For the convenience of
analysis, we denote by Mul a multiplication in G
(G1, G2 and GT ), Exp an exponentiation in G, Pair
a computation of the pairing, and Hash a regular
hashing operation. We omit other operations such as
addition in G for all the schemes.
As shown in Table 1, in the Query algorithm of our
scheme, the computation overhead increases with the
database item q. However, we need to remark that
the server does not need to compute the proof each
time. The reason is that the proof is identical for the
same data item and the server only need to compute
once for the first query on each index. Thus, the server
could adopt some storage overhead to reduce the
computational cost in the Query algorithm. Compare
with scheme [24], the Verify algorithm of our scheme
bring much more computation overhead. The reason
is that scheme [24] adopt the delegation technology
for data updating. In our scheme, to prevent the
attack against the collusion of the malicious and
revoked group users, we adopt the group signature
scheme with secure group user revocation. Although
the Verify algorithm bring much more computational
overhead than scheme [24], it is important that it is

9
0 100 200 300 400 500 600 700 800 900
0
0.5
1
1.5
2
2.5
3
3.5
4
q: the number of data items
TimeCost(s)
Figure 3. Query Time Cost
0 100 200 300 400 500 600 700 800 900
10
20
30
40
50
60
70
The number of data items or blocks
TimeCost(ms)
Scheme [23]
Scheme [24]
Our Scheme
Figure 4. Verify Time Cost
0 10 20 30 40 50 60 70 80 90
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
s: the elements of each block
TimeCost(s)
Scheme [24]
Our Scheme
Figure 5. Update Time Cost
a constant part of our scheme. In the Update and
ProofUpdate algorithms, the computation cost of our
scheme and the scheme [24] grow with the increase
of data elements (data items in our scheme). For
the UserRevocation algorithm, all the computation
computation time cost grow with the increase of
the challenge blocks (items) number. The different
is that, scheme [24] is efficient because the compu-
tation time cost will not grow with the increase of
the group users. Thus, their scheme provide constant
computation overhead with different group size. The
computation time cost of our scheme grows with the
revoked users number z, which is different from the
scheme [23] growing with the increase of the group
users number. Since it is reasonable that the revoked
users number z is small than the group users d, we
use 2z = d in our simulation.
6.2 Experimental evaluation
In this section, we evaluate the thorough experimental
evaluation of our scheme. Our experiments are sim-
ulated with the pairing-based cryptography library
(PBC)[36] on Linux Machine with Intel CoreTM
2 Duo
Processor T9500 running at 2.60GHz and 3G memory.
To precisely evaluate the computation complexity at
different entities, we simulate all the entity on this
machine.
As shown in Figure 3, the Query time cost of our
scheme is linear with the data items number q, which
will take approximately 4 seconds to query about
1000 data items. However, we need to emphasize that
the computation cost is at the cloud storage server
side, which is very powerful compare with the Linux
system running on our laptop. More over, the server
does not need to run the whole Query algorithm
every time as analyzed in the previous section.
Actually, in the Verify algorithm, the computation
overhead mostly comes from the group signature
scheme. More precisely, to verify the validity of this
phase, we need firstly to verify the integrity of the
signature, which means that our scheme need to gen-
erate the time costed parameters such as R1, R2, and
R3. Actually, the computation time cost of our scheme
a constant number. Also, it is around 5 times that of
the most efficient scheme [24]. In the Figure 5, we
show the data update computation comparison with
scheme [24], and both their computation overheads
grow with the increase of the element number in each
block.
In the PoofUpdate algorithm, as shown in Figure 6,
the computation time cost grows with the increase of
the elements number of each block and the number
of selected challenging data blocks. Actually, the data
blocks contain data elements in scheme [24] while not
in our scheme. It is interesting that, if we do not
consider the data elements in the scheme [24], the
computation overhead of the two schemes in with the
same challenging number are almost the same.
The UserRevocation algorithm simulation in Fig-
ure 7 shows that scheme [24] is the most efficient
one. Compare with scheme [23] whose computation
overhead grows rapidly with the increase of group
users number and the selected challenging blocks
number, scheme [24] and our scheme have a flat-
ting growth. The reason is that, scheme [23] has to
consider the whole group users number, while the
computation time cost in our scheme is related to the
revoked group users. It means that we need to verify
e(T2/A, û)
?
= e(T1, ˆv) for each user in the revocation
list. The best scheme is [24], whose computation
overhead is irrelevant to the group users number.
They achieve this by allowing the cloud storage server
to recompute the authentication tag of blocks last
modified by a revoked group user. We analyze this
tag update delegation way in the previous section and
point out that it is not secure against the cloud storage
server and revoked group users collusion attack.
7 RELATED WORK
Plenty of researchers have devoted considerable at-
tention to the problems on how to securely outsource
local store to remote cloud server. Among which,
the problem of remote data integrity and availability
auditing attacks the attestation of many researchers.
The concepts and solution Provable Data Possession
(PDP) and Proofs of Retrievability (PoR) were first
proposed by Ateniese et al. [10] and Juels et al. [11]. In

10
0
20
40
60
80
100 0
20
40
60
80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
c: selected challenging blocks
s: the elements of each block
TimeCost(s)
Scheme [24]
Our Scheme
Figure 6. ProofUpdate Time Cost
0
100
200
300
400
500 0
10
20
30
0
0.5
1
1.5
2
2.5
d: the group users number
c: selected challenging blocks number
TimeCost(s)
Scheme [23]
Scheme [24]
Our Scheme
Figure 7. UserRevocation Time Cost
their scheme, the homomorphic authentication tech-
nique was adopted to reduce both the communication
and computation cost. Later, a number of variants of
PDP and PoR schemes are designed to improve the
efficiency and enhance the function of basic schemes,
such as allowing public auditing [16], [22], [17] and
supporting data update [14], [15].
To enhance the previous works, Wang et al. [22]
designed a scheme to support share data integrity
auditing, whose scheme adopted ring signature to
protect the privacy of users. The limitation of the
scheme is that it does no support dynamic group
and also suffers from a computational overhead linear
to the group size and the number of data auditing.
To further support user revocation, Wang et al. [23]
designed another scheme based on the assumption
that no collusion occurs between cloud servers and
revoked user. As a matter of fact, they assumed that
the private and authenticated channels exit between
each pair of entities and collusion between invalid
users and cloud servers will lead to the disclosure
of secrets of all other valid users. Recently, Yuan
and Yu [24] designed a dynamic public integrity
auditing scheme with secure group user revocation.
The scheme is based on polynomial authentication
tags and adopts proxy tag update techniques, which
makes their scheme support public checking and
efficient user revocation. However, the authors do
not consider the ciphertext store. Also, to make the
scheme efficient, the data owner (the data owner’s
private key is not necessary) does not take part in the
user revocation phase, where the cloud could conduct
some malicious operation of user’s data when it col-
ludes with the revoked users.
Gennaro et al. [37] formalized the notion of verifi-
able computation which allows a client to outsource
the computation of an arbitrary function. However, it
is inefficient for practical applications due to the com-
plicated fully homomorphic encryption techniques
[38], [39]. Also, another disadvantage of the schemes
based on fully homomorphic encryption is that, the
client must repeat the expensive pre-processing stage
if the malicious server tries to cheat and learn a
bit of information. Benabbas et al. [40] proposed the
first practical verifiable database scheme based on
the hardness of the subgroup membership problem
in bilinear groups with composite order. However,
the scheme does not support the public verifiability
property. Catalano and Fiore [25] proposed a practical
solution to build verifiable database (VDB) from vec-
tor commitment that supports the public verifiability.
Both of the schemes assume that the size of the out-
sourced database should be fixed and the client can
know the outsourcing function in advance. Recently,
Backes et al. [41] presented a flexible VDB scheme
with two additional properties that eliminates the
assumption.
Group signature is introduced by Chaum and Heyst
[42]. It provides anonymity for signers, where each
group member has a private key that enables the
user to sign messages. However, the resulting signa-
ture keeps the identity of the signer secret. Usually,
there is a third party that can conduct the signa-
ture anonymity using a special trapdoor. Some sys-
tems support revocation [43], [44], [45], [27], [46],
[47], where group membership can be disabled with-
out affecting the signing ability of unrevoked users.
Boneh and Shacham [27] proposed an efficient group
signature with verifier-local revocation. The scheme
provides the properties of group signature such as
selfless-anonymity and traceability. Also, the scheme
is a short signature scheme where user revocation
only requires sending revocation information to sig-
nature verifiers. Libert et al. [46] proposed a new
scalable revocation method for group signature based
on the broadcast encryption framework. However,
the scheme introduces important storage overhead at
group user side. Later, Libert et al. [47] designed a
scheme to enhance the former scheme which could
obtain private key of constant size. In their scheme,
the unrevoked members still do not need to update
their keys at each revocation.

11
8 CONCLUSION
The primitive of verifiable database with efficient
updates is an important way to solve the problem
of verifiable outsourcing of storage. We propose a
scheme to realize efficient and secure data integrity
auditing for share dynamic data with multi-user mod-
ification. The scheme vector commitment, Asymmet-
ric Group Key Agreement (AGKA) and group sig-
natures with user revocation are adopt to achieve
the data integrity auditing of remote data. Beside
the public data auditing, the combining of the three
primitive enable our scheme to outsource ciphertext
database to remote cloud and support secure group
users revocation to shared dynamic data. We provide
security analysis of our scheme, and it shows that our
scheme provide data confidentiality for group users,
and it is also secure against the collusion attack from
the cloud storage server and revoked group users.
Also, the performance analysis shows that, compared
with its relevant schemes, our scheme is also efficient
in different phases.
9 ACKNOWLEDGMENT
This work is supported by the National Natural Sci-
ence Foundation of China (No. 61272455), China 111
Project (No. B08038), Doctoral Fund of Ministry of
Education of China (No. 20130203110004), Program
for New Century Excellent Talents in University (No.
NCET-13-0946), and the Fundamental Research Funds
for the Central Universities (Nos. BDY151402 and
JB142001-14).
REFERENCES
[1] Amazon. (2007) Amazon simple storage service (amazon s3).
Amazon. [Online]. Available: http://aws.amazon.com/s3/
[2] Google. (2005) Google drive. Google. [Online]. Available:
http://drive.google.com/
[3] Dropbox. (2007) A file-storage and sharing service. Dropbox.
[Online]. Available: http://www.dropbox.com/
[4] Mozy. (2007) An online, data, and computer backup software.
EMC. [Online]. Available: http://www.dropbox.com/
[5] Bitcasa. (2011) Inifinite storage. Bitcasa. [Online]. Available:
http://www.bitcasa.com/
[6] Memopal. (2007) Online backup. Memopal. [Online].
Available: http://www.memopal.com/
[7] M. A. et al., “Above the clouds: A berkeley view of cloud
computing,” Tech. Rep. UCBEECS, vol. 28, pp. 1–23, Feb. 2009.
[8] M. Rabin, “Efficient dispersal of information for security,”
Journal of the ACM (JACM), vol. 36(2), pp. 335–348, Apr. 1989.
[9] J. G. et al. (2006) The expanding digital universe: A forecast of
worldwide information growth through 2010. IDC. [Online].
Available: Whitepaper
[10] G. Ateniese, R. Burns, R. Curtmola, J. Herring, L. Kissner,
Z. Peterson, and D. Song, “Provable data possession at un-
trusted stores,” in Proc. of ACM CCS, Virginia, USA, Oct. 2007,
pp. 598–609.
[11] A. Juels and B. S. Kaliski, “Pors: Proofs of retrievability for
large files,” in Proc. of ACM CCS, Virginia, USA, Oct. 2007,
pp. 584–597.
[12] K. D. Bowers, A. Juels, and A. Oprea, “Proofs of retrievability:
theory and implementation,” in Proc. of CCSW 2009, llinois,
USA, Nov. 2009, pp. 43–54.
[13] Y. Dodis, S. Vadhan, and D. Wichs, “Proofs of retrievability
via hardness amplification,” in Proc. of TCC 2009, CA, USA,
Mar. 2009, pp. 109–127.
[14] Q. Wang, C. Wang, J. Li, K. Ren, and W. Lou, “Proofs of
retrievability via hardness amplification,” in Proc. of ESORICS
2009, Saint-Malo, France, Sep. 2009, pp. 355–370.
[15] C. Erway, A. Kupcu, C. Papamanthou, and R. Tamassia,
“Dynamic provable data possession,” in Proc. of ACM CCS,
Illinois, USA, Nov. 2009, pp. 213–222.
[16] C. Wang, Q. Wang, K. Ren, and W. Lou, “Privacy-preserving
public auditing for data storage security in cloud computing,”
in Proc. of IEEE INFOCOM 2010, CA, USA, Mar. 2010, pp. 525–
533.
[17] J. Yuan and S. Yu, “Proofs of retrievability with public
verifiability and constant communication cost in cloud,” in
Proc. of International Workshop on Security in Cloud Computing,
Hangzhou, China, May 2013, pp. 19–26.
[18] E. Shi, E. Stefanov, and C. Papamanthou, “Practical dynamic
proofs of retrievability,” in Proc. of ACM CCS 2013, Berlin,
Germany, Nov. 2013, pp. 325–336.
[19] Cloud9. (2011) Your development environment, in the cloud.
Cloud9. [Online]. Available: https://c9.io/
[20] Codeanywhere. (2011) Online code editor. Codeanywhere.
[Online]. Available: https://codeanywhere.net/
[21] eXo Cloud IDE. (2002) Online code editor. Cloud IDE.
[Online]. Available: https://codenvy.com/
[22] B. Wang, B. Li, and H. Li, “Oruta: Privacy-preserving public
auditing for shared data in the cloud,” in Proc. of IEEE CLOUD
2012, Hawaii, USA, Jun. 2012, pp. 295–302.
[23] B. Wang, L. Baochun, and L. Hui, “Public auditing for shared
data with efficient user revocation in the cloud,” in Proc. of
IEEE INFOCOM 2013, Turin, Italy, Apr. 2013, pp. 2904–2912.
[24] J. Yuan and S. Yu, “Efficient public integrity checking for cloud
data sharing with multi-user modification,” in Proc. of IEEE
INFOCOM 2014, Toronto, Canada, Apr. 2014, pp. 2121–2129.
[25] D. Catalano and D. Fiore, “Vector commitments and their
applications,” in Public-Key Cryptography - PKC 2013, Nara,
Japan, Mar. 2013, pp. 55–72.
[26] Q. Wu, Y. Mu, W. Susilo, B. Qin, and J. Domingo-Ferrer,
“Asymmetric group key agreement,” in Proc. of EUROCRYPT
2009, Cologne, Germany, Apr. 2009, pp. 153–170.
[27] D. Boneh and H. Shacham, “Group signatures with verifier-
local revocation,” in Proc. of ACM CCS, DC, USA, Oct. 2004,
pp. 168–177.
[28] D. Boneh, B. Lynn, and H. Shacham, “Short signatures from
the weil pairing,” in Proc. of Asiacrypt 2001, Gold Coast,
Australia, Dec. 2001, pp. 514–532.
[29] D. Boneh and X. Boyen, “Collision-free accumulators and fail-
stop signature schemes without trees,” in Proc. of EUROCRYPT
2004, Interlaken, Switzerland, May 2004, pp. 56–73.
[30] N. Baric and B. Pfitzman, “Collision-free accumulators and
fail-stop signature schemes without trees,” in Proc. of EURO-
CRYPT 1997, Konstanz, Germany, May 1997, pp. 480–494.
[31] D. Boneh, X. Boyen, and H. Shacham, “Short group signa-
tures,” in Proc. of CRYPTO 2004, CA, USA, Aug. 2004, pp.
41–55.
[32] U. M. Maurer and S. Wolf, “Diffie-hellman oracles,” in Proc.
of CRYPTO 1996, CA, USA, Aug. 1996, pp. 268–282.
[33] F. Bao, R. Deng, and H. Zhu, “Variations of diffie-hellman
proble,” in Information and Communications Security, Huhe-
haote, China, Oct. 2003, pp. 301–312.
[34] X. Chen, J. Li, J. Weng, J. Ma, and W. Lou, “Verifiable compu-
tation over large database with incremental updates,” in Proc.
of ESORICS 2014, Wroclaw, Poland, Sep. 2014, pp. 148–162.
[35] X. Chen, J. Li, X. Huang, J. Ma, and W. Lou, “New pub-
licly verifiable databases with efficient updates,” to appear
in IEEE Transactions on Dependable and Secure Computing,
Accepted.
[36] B. Lynn. (2006) The pairing-based cryptography library.
[Online]. Available: http://crypto.stanford.edu/pbc/
[37] R. Gennaro, C. Gentry, and B. Parno, “Non-interactive verifi-
able computing: Outsourcing computation to untrusted work-
ers,” in Proc. of CRYPTO 2010, CA, USA, Sep. 2010, pp. 465–
482.
[38] C. Gentry, “Fully homomorphic encryption using ideal lat-
tices,” in Proc. of ACM STOC 2009, Washington DC, USA, May
2009, pp. 169–178.

12
[39] C. Gentry and S. Halevi, “Implementing gentrys fully-
homomorphic encryption scheme,” in Proc. of EUROCRYPT
2011, Tallinn, Estonia, May 2011, pp. 129–148.
[40] S. Benabbas, R. Gennaro, and Y. Vahlis, “Verifiable delegation
of computation over large datasets,” in Proc. of CRYPTO 2011,
CA, USA, Aug. 2011, pp. 111–131.
[41] M. Backes, D. Fiore, and R. M. Reischuk, “Verifiable delegation
of computation on outsourced data,” in Proc. of ACM CCS
2013, Berlin, Germany, Nov. 2013, pp. 863–874.
[42] D. Chaum and E. van Heyst, “Group signatures,” in Proc. of
EUROCRYPT 1991, Brighton, UK, Apr. 1991, pp. 257–265.
[43] E. Bresson and J. Stern, “Efficient revocation in group signa-
tures,” in Public-Key Cryptography - PKC 2001, Cheju Island,
Korea, Feb. 2001, pp. 190–206.
[44] J. Camenisch and A. Lysyanskaya, “Dynamic accumulators
and application to efficient revocation of anonymous creden-
tials,” in Proc. of CRYPTO 2002, CA, USA, Aug. 2002, pp. 61–76.
[45] G. Ateniese, D. Song, and G. Tsudik, “Quasi-efficient revoca-
tion in group signatures,” in Proc. of FC 2002, Soughamption,
Bermuda, Mar. 2002, pp. 183–197.
[46] B. Libert, T. Peters, and M. Yung, “Scalable group signatures
with revocation,” in Proc. of EUROCRYPT 2012, CA, USA, Aug.
2002, pp. 61–76.
[47] ——, “Group signatures with almost-for-free revocation,” in
Proc. of CRYPTO 2012, CA, USA, Aug. 2012, pp. 571–589.
Tao Jiang received his B.S. (2009) in Net-
work Engineering from Shandong Jianzhu
University, China. He got his M.S. (2012)
in Computer Application Technology from
Jiangsu University, China. Currently, he is a
Ph.D. student of Xidian University in Cryptog-
raphy. His research interests include cryptog-
raphy and cloud computing security.
Xiaofeng Chen received his B.S. and M.S.
on Mathematics from Northwest University,
China in 1998 and 2000, respectively. He got
his Ph.D degree in Cryptography from Xidian
University in 2003. Currently, he works at Xi-
dian University as a professor. His research
interests include applied cryptography and
cloud computing security. He has published
over 100 research papers in refereed inter-
national conferences and journals. His work
has been cited more than 1800 times at
Google Scholar. He is in the Editorial Board of Computing and
Informatics (CAI), International Journal of Grid and Utility Computing
(IJGUC), and International Journal of Embedded Systems (IJES)
etc. He has served as the program/general chair or program com-
mittee member in over 30 international conferences.
Jianfeng Ma received his B.S. degree in
mathematics from Shaanxi Normal Univer-
sity, China in 1985, and obtained his M.E.
and Ph.D. degrees in computer software
and communications engineering from Xid-
ian University, China in 1988 and 1995, re-
spectively. From 1999 to 2001, he was with
Nanyang Technological University of Singa-
pore as a research fellow. Now he is a Pro-
fessor in School of Computer Science at Xi-
dian University, China. His current research
interests include distributed systems, computer networks, and infor-
mation and network security.

Public Integrity Auditing for Shared Dynamic Cloud Data with Group User Revocation

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Public Integrity Auditing for Shared Dynamic Cloud Data with Group User Revocation (20)

Recently uploaded (20)

Public Integrity Auditing for Shared Dynamic Cloud Data with Group User Revocation