Ternary Tree Based TGDH Protocol for Dynamic

Secure Group Data Sharing In Healthcare Cloud

Vaishali R. Thakare K. John Singh
School of Information Technology and Engineering School of Information Technology and Engineering
VIT University, Vellore-632014(TN), India VIT University, Vellore-632014(TN), India
Email: [email protected] Email: [email protected]

Abstract Healthcare requires continuous and systematic cloud clients with quality of service and resources are
innovation in order to provide high quality services. Cloud Infrastructure-as-a-Service (IaaS) and Software-as-a-Service
computing is a new model of computing that promises to provide (SaaS). The structure of the cloud computing moves the
more flexibility, less expense and more efficient IT services to
application software and databases to the large data centers,
end-users. Large amount of sensitive data is shared by the
where the management of the medical data and services are
healthcare cloud users. Here, privacy and security of group
sharing data have become two major issues. The aim of this
not trustworthy. Cloud nodes are inherently more vulnerable
paper is to propose a ternary tree based TGDH protocol for to cyber-attacks than traditional solutions, given their size and
dynamic group data sharing in healthcare cloud that could be underlying service-related complexity that brings an
used by a healthcare organization to share their data in unprecedented exposure to third parties of services and
dynamically secure groups containing other health organizations. interfaces. Nowadays healthcare organizations are sharing
Secure and reliable group communication in healthcare their data with each other electronically over cloud for their
organizations is an increasingly active research area by growing reliability, but cloud is a semi-trusted third party and hence,
popularity in group oriented and collaborative applications.
security and privacy of data have become two major issues
Ternary tree approach covers more other healthcare members
comes in. With the help of cryptographic encryption
(organizations) in a subgroup and height of ternary tree will get
increase when number of members in a group increase where
technologies one can store his/her data in the cloud. The users
height of tree is number of iterations required to compute group who want to download the files can download and decrypt
shared key. them with provided keys [9]. Here, the issue arises, how to
Keywords Healthcare Cloud, TGDH protocol, Group Data distribute and update session keys? For example, assume that
Sharing, Security, and Ternary Tree. Alice wants to send a file securely to recipient Bob, then first
Alice chooses a random session key and uses symmetric
I. INTRODUCTION
encryption algorithm [10] (such as AES, DES) to encrypt a
file:{FILE}. Then, Alice uses asymmetric encryption
According to the US National Institute of Standards and algorithm (such as RSA) to encrypt Session Key: E PuB .
Technology (NIST definition, cloud computing is a model Now, the Key: E PuB is called as digital envelope and this
for enabling convenient, on-demand network access to a can be distributed among the group members. However, if a
shared pool of configurable computing resources that can be file is shared to N number of users then N numbers of
rapidly provisioned and released with minimal management digital envelopes are to be generated, this creates
effort or service provider interaction [1]. Its applications communication overhead i.e., O(N) for a single file.
have been reported in business, industry, research, education, Similarly, the computational complexity is O (N). In addition,
transportation and even national security [25].Cloud to share M number of files to N specific recipients then, O
computing is significant in streamlining healthcare services (MN) number of digital envelopes to be formed. To reduce
whether it is for maintaining health records, monitoring of this overhead and to communicate securely over insecure
patients, collaboration with peers, and analysis of patient data. channels it is essential that secret keys are distributed securely.
To leverage cloud clients from hardware requirements while Even if the encryption algorithm used is computationally
reducing overall client side requirements and complexity, one infeasible to break, the entire system is vulnerable if the keys
the components used is the cloud platform service known as are not securely distributed. Research has focused on security
Platform-as-a-Service (PaaS). Other components offering and on efficiency. Many practical systems have been proposed
[6, 7]. The most familiar system is the TGDH key distribution III. SYSTEM MODEL
system. The system model (Fig. 1) consists of four entities: the cloud,
a hospital group MD, group admins, and large number of
II. RELATED WORK
group members. The data owner initially encrypts his/her data
Kaiping Xue and Peilin Hong [8] proposed a novel secure
locally to ensure privacy, outsources the data, and stores it in a
group sharing framework for public cloud, the framework can
group public cloud for easy user access. The data owner then
take the effective advantage of cloud help by taking care that
issues a decryption key for each authorized user according to
no sensitive data should be exposed to cloud provider and
his/her access rights.
attacker. It combines proxy signature, enhanced TGDH-based
Cloud is operated by CSPs and provides priced abundant
binary tree, proxy re-encryption as a protocol, with the help of
storage services. However, the cloud is not fully trusted by
this protocol the aim of authors proposed framework has been
user since the CSPs are very likely to be outside of the cloud
achieved. In this scheme authors used TGDH with binary tree
users trusted domain.
to negotiate and update the group key pairs with the help of
Hospital Group MD (HMD) is a group creator, only one group
cloud servers, but in binary tree based approach the group of
MD for a group exist, i.e., the topmost level of group
two members is formed, so per iteration the members covered
hierarchy is HMD. HMD can authorize specific group member
by a subgroup are less. Moreover height of binary tree will get
to manage the group and other group members and this
increase when number of members in a group increase where
privilege can be revoked by HMD when needed.
height of tree is number if iterations required to compute
group shared key. Similarly, cost will also increase when
number of members in a group will increase proportionally.
Hence, in proposed scheme ternary tree based approach is
used with GDH.2 and numbers of members covered by
subgroups are more, height is also decreased and it will take
less time to compute the shared group key and cost is also
decreased.
Priyanka Jaiswal et. al.[10] proposed an alternative approach
to group key agreement, i.e., a novel queue based group key
agreement protocol which uses the concepts of elliptic curve
cryptography to reduce unnecessary delays, considers member
diversity with filtering out low performance members in group
key generation processes. After analysing many prior group
key agreement protocols like TGDH, STR, BD, QBDH etc.,
they provide better security but they takes more computational
overheads. So, authors have used elliptic curve cryptographic
technique that removes exponentiation to reduce
computational overheads and provides better results than the Fig. 1 System Model od proposed system
other group key agreement protocols.
In addition, HMD generates initial security parameters and
Bharath K. Samanthula et. al. [9] proposed an efficient and
provides for all group members.
secure data sharing (SDS) framework using homomorphic
Hospital Group Admin (HGAd) takes a charge of user
encryption and proxy re-encryption schemes that prevent the
registration and user revocation. There can be 0,1 or more
leakage of unauthorized data when a revoked user rejoins the
group admins (HGAds)in a group, they acts like a member of
system. Also, authors present new solution based on the data
group but also having authority of group management. HGAds
distribution technique to prevent the leakage in case of
are guarantor to implement updating of group key.
collusion between revoked user and the cloud service
Hospitals Member (HM) have privileges to upload and
provider. To propose the new framework to support secure
download file in authorized group. Each HM can compute
query processing, authors have used existing work for
group key pair.
enhancement. Moreover, detailed security with experimental
Note: Here, HMD belongs to {HGAd} subset of {HM}.
analysis of the proposed framework on Amazon EC2 is
provided. IV. PROPOSED SCHEME
Proposed Protocol:
Following Table 1 shows the notations used in this paper. The .. (Where h is any integer and HMD=1 since HMD is
role of participants in this scheme is classified into following the topmost level)
way: Each member in Tr should have 0 or 3 child nodes
Call HMD child nodes as HGAdi (HGAd1, HGAd2, and
- Group member node: represents each group member
HGAd3 shown in Fig. 2)
as a leaf
SUBGROUP_FORMATION ()
- Sponsor node: corresponds with one key. This is HGAdi forms a set of subgroups
shared by all members of the subgroup rooted at this - HGAd1: {HM1, HM2, HM3},
key node (GM3, GM6 and GM9) - HGAd2: {HM4, HM5, HM6} and
- HGAd3: {HM7, HM8, HM9}
HGAd1, HGAd2, and HGAd3 computes their own secret
keys KH1, KH2, KH3 using TGDH() protocol.
HGAdi shares theirs secret keys with each other to
compute final group key
TGDH()
Assumption 1: HMD=1, HGAdi =3 and HMi= 9 as
shown in Fig. 2
Assumption 2: Rightmost group member of each
subgroup; HM3, HM6, HM9 act as a sponsor for
respective subgroups
Up flow:
1. HGAd1 determines a set {gKH1 mod p}
Broadcast
HGAd2: {HM4, HM5, HM6}
Fig. 2: Hospitals Group formation mechanism 2. HGAd2 determines a set {g KH1 KH2mod p,
gKH2mod p, g KH1 mod p} Broadcast HGAd3: {
Notation Description HM7, HM8, HM9}
n Number of current group members (here n=9) 3. HGAd3 determines a set {g KH1 KH2 KH3mod p, g
L Set of leaving members KH1KH2
mod p, g KH2 KH3mod p, gKH1 KH3mod p}
GMi i-th group member; i belongs to {1,2,..,n} where and computes group key since its having all
3h for all HM i
members secret keys
h height of tree
<l,v> v-th node at the l-th level of tree
4. HM9 acts as a main sponsor and it computes final
HMD Hospitals group managing director where HMD=1 group key i.e.,
HGAdi Hospitals group admins Key: g KH1 KH2 KH3mod p
BKi* set of HMis blinded keys
N Number of new member send requests to join/leave Down flow:
Broadcast
the original group(this value always 3h form) 1. HM9 HGAd1: {HM1, HM2, HM3},
{ g KH1KH2mod p,
n-N Number of remain member after leave request by N . HGAd2: {HM4, HM5, HM6}
members (3h form always) g KH2 KH3mod p, g KH1 KH3mod p}

PuB Public key of Bob 2. HGAd1 computes Key: g KH1(KH2KH3) mod p

PuKRe Public key of Receiver
3. HGAd2 computes Key: g KH2(KH1KH3) mod p
PrKGrp new group private key
PuKGrp group public key
End.

Table 1: Notation used with description Algorithm: Hospitals Group Formation Algorithm

Initialization HMD generates a ternary tree Tr. Each node in Tr contains 0

Group MD (GMD) takes charge of group initialization. or 3 nodes. Child nodes of HMD called as Hospitals group
Since it is a ternary tree based approach, the group members, admins HGAdi. Now HGAdi forms a set of subgroups
which are participating in the group, are 3h form. Where h is containing three nodes and uses TGDH protocol to compute
any integer number. Group initialization done as follows: the group key as discussed in previous sections.
Fig. 2 shows the group formation mechanism by using ternary
TGDH technique and group formation algorithm. Since, it is
INITIALIZE () ternary tree it shows one group leader three group admins i.e.,
HMD: Construct a ternary tree Tr with members 3 h form
nine group members for consideration. The root node is
located at the 0-th level (HMD) and the lowest leaves are at
the h-th level. Every member can be a leaf or a parent of three point in the group (tree). Otherwise, if the ternary tree is fully
members. The members are denoted by <l, v>, where 0<= balanced, then the new member joins at the new level of tree.
v<= 3h-1 since each level l hosts at most 3h nodes. Each The sponsor is the rightmost leaf in the subgroup (sub tree)
member is associated with the key KH<l, v> and the blinded key rooted at the insertion point. If the intermediate node in the
BKH<l, v>. We assume that the member at leaf <l, v> is rightmost has 0 nodes (members), the sponsor inserts the new
associated with HMi, then the member <l, v> has HMis member under this intermediate node. The tree becomes fully
session random key KH<l, v>. Furthermore, the member HMi at balanced. Otherwise each member creates a new intermediate
node <l, v> knows every key along the path from <l, v> to node and a new member node, and promotes the new
<0, 0>. In Fig 3 HM3, HM6 and HM9 are considered as intermediate node to be the parent of both the insertion node
sponsors that owns subgroups (sub tree). Then, HM3 knows and the new member node. After updating the tree, the
every key {KH<2, 2>, KH<1, 0>, KH<0, 0>. } and every blinded key { sponsor proceeds to update his share and computes the new
BKH<2,2v>, BKH<1, 0>, BKH<0,0v>}of that subgroup. Similarly for group key, while computing new group key includes all
HM6 and HM9. To compute single group key for all the members. Finally, sponsor will broadcast the updated group to
members, following two rounds need to follow: to all other members in the group including new member.
Sponsor is one of the group members from that group which
Round 1: Here, HGAd1: {HM1, HM2, HM3}, HGAd2: {HM4, having authorities to add or delete the group member from
HM5, HM6} and HGAd3: {HM7, HM8, HM9} are three set of group such as, HGAdi, HMD, or HMi. Let us consider HGAdi
group members that forms a new subgroups. These new is a sponsor for new group member to join. Then, HGAdi tries
subgroups compute their own secret keys KH1, KH2, KH3 using to find the leaf node. Every node in the group is under the
TGDH and shares between HGAd1, HGAd2, and HGAd3. administration of one of the group member in the group,
Round 2: To compute key for subgroups HGAd1, HGAd2 and which is having additional authorities than normal group
HGAd3 there are two stages member. Then found node is set as the associated one of the of
1. Up-flow and the new joining group member. If not found then HGAdi
2. Down-flow. creates a new node to the tree and associate it with one of the
The rightmost group member of each subgroup (i.e. HM3, HGAds and attaches the new member to that newly created
HM6, HM9) are assumes as a sponsors for HGAd1, HGAd2 and node with smallest depth in tree structure.
HGAd3. Now, new group member have to follow the following process
Up-flow: for its security:
HGAd1=> Determine a set {g KH1 mod p} and broadcast to all x Select a random security key.
members of HGAd2 i.e. HM4, HM5, and HM6 x Selects his/her path from its associated node (to cloud
HGAd2 => Determine a set {g KH1 KH2mod p, gKH2mod p, g KH1 server) and get the blinded keys of all sibling nodes
mod p} and broadcast to all members of HGAd3 i.e., HM7, in path.(Sponsor will have all keys of that root)
HM8, and HM9 x Calculates new security and blinded keys of each
HGAd3 => Determine a set {g KH1 KH2 KH3mod p, g KH1KH2mod node in path.
p, g KH2 KH3mod p, g KH1 KH3mod p}. x Send all the blinded keys from the associated node to
Now, g KH1 KH2 KH3 is a group key i.e. K: g KH1 KH2 KH3 root node in the path to the HGAdi in an
Down-flow: authentication tunnel.
HGM9 => Broadcasts {g KH1KH2mod p, g KH2KH3mod p, g
KH1KH3
mod p} to all other members of groups, from this Now, HGAdi uploads all these blinded keys in the cloud server
HGAd1 and HGAd2 can also determine the group key path, cloud server updates the tree structure and blinded keys
of each node in path.
New Member Join Whenever new member is joining the group then every time
In proposed model, any new member can join the the ternary key tree and group key pair should be updated to
group by sending the new join request to the group. New prevent the forward secrecy of data. However, all digital
group member need to follow the procedure below to join enveloped do not need to be updated.
group:
Consider the group has n members: {M1, HM2, ....HMn}. The Existing Member Leaving
new member HMn+1 initiates the protocol by new joining The proposed is made flexible for the group members,
request including its blinded key (gk mod p) should reach to group members can join the group by sending the join request
every member of that group and first determines the insertion and can leave the group whenever they want. To prevent the
backward secrecy, the group key pair and all digital envelopes KSS1: When new member joins the group, key synchronization
related to data sharing should be updated. In addition, the new is necessary, because common group key is derives with the
digital envelopes should be encrypted by new group public contribution of all group members in the group. In addition, to
key. provide the forward secrecy.
Consider HGAdi is a mandator and sponsor for group member KSS2: When an existing group member leaves the group, to
leaving event. HGAdi have to follow the below process: provide backward secrecy key synchronizing has to be done.
x If the siblings of the leaving group members KSS3: When offline member comes online again, key
associated node are also mandated by a HGA, these synchronizing will be done to get the current group key of
nodes (the leaving group members associated node group.
and its siblings) and their parent node should be KSS4: When any group member rejoins the group, to provide
merged to one leaf node. HGAdi mandates this new the backward secrecy, key synchronizing will be implemented.
leaf node. Otherwise, if the sibling node is associated
with a group member, HGAdi straightly mandates the
V. SECURITY ANALYSIS
leaving group members associated leaf node.
x Randomly choose a new secret key K for the new Backward secrecy: When any group member leaves the
mandated node i, j, k, and then update secret keys group, HGAd or the HMD mandates its position in the ternary
and blinded keys in the path from i, j, k to the root tree. Every group member can compute the security keys of
every node in the path from leaf to root node. Because of this
node 0, 0, 0. The updated root security key and
reason, the security key of every node in the path updated
blinded key are the new group private key (PrKGrp)
and group public key (PuKGrp). when any member leaves the group from that path. After
leaving the group, that member cannot able to compute the
Then, HGAdi computes the proxy re-encryption key and then group key whereas; all other members in the group can
uploads the updated information into the cloud. compute the new group key [13].

Admin Leaving Cloud provider cannot penetrate to group: Cloud provider

Each HGAdi mandates more than one leaf node, and knows blinded keys of every node in a ternary tree, but it do
he/she knows the blinded keys of its leaf nodes. When a not knows the security keys of any internal nodes, so it cannot
HGAdi leaves, another HGAdi or HMD should mandate these get the final group private key to penetrate itself into group.
leaf nodes and change the security keys instead of him/her. As As we consider cloud providers are semi-trusted third party,
the new mandating HGAdi or HMD chooses a random secret we cannot let permit cloud provider in the group for security
key for each of the leaving HGAdis mandated leaf nodes, and and privacy concerns.
computes the secret keys and blinded keys of all node in the
path from each of these new mandated leaf node to the root Data Confidentiality: to prevent the data confidentiality a
node. All the paths from each of these leaf nodes to the root traditional way used while sending data from sender to
node form a sub tree of the ternary tree. The mandating receiver is, sender chooses a random session key and uses
HGAdi or HMD first lists all internal nodes in the sub tree but symmetric encryption algorithm (such as AES, DES) to
whose siblings are not in sub tree, and gets the blinded keys of encrypt a file:{FILE}. Then, sender uses asymmetric
their siblings in Tr from cloud Servers. encryption algorithm (such as RSA) to encrypt Session Key: E
PuKRe. Now, the Key: E PuKRe is called as digital envelope

Key synchronizing and this can be distributed among the group members. In our
To provide the enough security, key synchronizing proposed scheme, we design a group key pair (PrKgrp, PuKgrp).
[14] is necessary, key synchronizing needed when following This schemes guarantee the authenticated group members
scenarios comes: know the current group private key. The whole security relies
x When any new member joins the group (forward on the PrKgrps security. Every members of the group are
secrecy) accessible to the PuKgrp, hence only group members can
x When any group member or group admin leaves the compute the updated group private key. Moreover, only group
group members can decrypt and download the files from the group
x When an offline member becomes online again data sharing.
x When any member rejoins the group (backward
secrecy)
 Certificateless Authentication: In this proposed work, proxy Broadcast 2*n-2 3 Min(2*N, n-
N)
signature scheme is used to grant the data validity, which
Rounds hp 1 1
uploaded by the data owner to the cloud, and to grant the Proposed Messages 3/2*(n-1) 2 3/2*(n-N-1)
group admin (HGAd) privileges to the group members (HMi). Protocol Broadcast 5*n-9/6 1 (5*(n-N)-
With the use of this HMD can grant the privileges of group 9)/6
admins (HGAdi), for this privileges HGAdi needs to provide Table 3: comparison of proposed with existing TGDH
its public key PuKHGAdi and message warrantor i.e., MwHGAdi.
REFERENCES
Every members of the group has rights to verify the HGAdi
authorization and it can be done with the help of HMDs
[1] Mell P, Grance T (2010) The NIST definition of cloud computing.
public key. Cloud servers can maintain the public stored Commun ACM 53(6):50
privilege revocation list (PRL) which includes PuKHGAdi or [2] Koufi V, Malamateniou F, Vassilacopoulos G (2010) Ubiquitous access
MwHGAdi. And every verifier can check this PRL to verify the to cloud emergency medical services. In: 10th IEEE International
Conference on Information Technology and Applications in
authentication details. The status of the group members is in Biomedicine (ITAB), IEEE Press, New York, pp 14
the following format shown in Table 2. [3] Behrend TS, Wiebe EN, London JE, Johnson EC (2011) Cloud
computing adoption and usage in community colleges. Behav Inf
Technol 30(2):231240
Member Email Id Group Status Block Revoke
[4] Li Z, Cheng C, Wang K (2011) Cloud computing for agent-based urban
Name Name transportation systems. IEEE Intell Syst 26(1):7379
Alice alice16@gmai Cloud Joined Block Revoke [5] NSA embraces cloud computing,
l.com 1
http://www.darkgovernment.com/news/nsa-embracescloud-computing/
Steve steve16@gma Cloud Left Block Revoke
[6] Liu, B., Chen, Y., Hadiks, A., Blasch, E., Aved, A., Shen, D. and Chen,
il.com 2 G., 2014. Information fusion in a cloud computing era: a systems-level
Table 2: Group member status view perspective. Aerospace and Electronic Systems Magazine, IEEE, 29(10),
pp.16-24.
VI. CONCLUSION AND FUTURE WORK [7] Liu, H., Ning, H., Xiong, Q. and Yang, L.T., 2015. Shared authority
based privacy-preserving authentication protocol in cloud computing.
Data sharing in cloud environment by forming groups is Parallel and Distributed Systems, IEEE Transactions on, 26(1), pp.241-
considered as a common approach. But secure data sharing in 251.
[8] Xue, K. and Hong, P., 2014. A dynamic secure group sharing framework
groups is still a challenging issue for the researchers. There are in public cloud computing. Cloud Computing, IEEE Transactions on,
many frameworks available to share data in groups in the 2(4), pp.459-470.
previous researches listed in Related work section of this [9] Samanthula, B.K., Elmehdwi, Y., Howser, G. and Madria, S., 2015. A
secure data sharing and query processing framework via federation of
paper, one of them is dynamic secure group sharing cloud computing. Information Systems, 48, pp.196-212.
framework for public cloud proposed by Kaiping Xue and [10] Jaiswal, P., Kumar, A. and Tripathi, S., 2015. Design of Queue-Based
Peilin Hong, the framework combines proxy signature, Group Key Agreement Protocol Using Elliptic Curve Cryptography.
In Information Systems Design and Intelligent Applications (pp. 167-
TGDH-based binary tree, proxy re-encryption as a protocol. 176). Springer India.Wang, C.J. and Luo, J.F., 2012, November. A key-
But TGDH- based binary tree approach has some drawbacks policy attribute-based encryption scheme with constant size ciphertext.
In Computational Intelligence and Security (CIS), 2012 Eighth
which are explained in Related work. After analyzing the International Conference on (pp. 447-451). IEEE.
existing frameworks drawbacks, in this paper a dynamic [11] Seo, S.H., Nabeel, M., Ding, X. and Bertino, E., 2014. An efficient
certificateless encryption for secure data sharing in public clouds.
secure group sharing framework in public cloud computing Knowledge and Data Engineering, IEEE Transactions on, 26(9),
environment is proposed with TGDH-based ternary tree pp.2107-2119.
approach which effective than the existing frameworks. [12] Roman, A. and Szykula, M., 2015. Forward and backward
synchronizing algorithms. Expert Systems with Applications, 42(24),
Moreover, the proposed protocol is beneficial in healthcare pp.9512-9527.
cloud applications for secure data sharing. [13] Kudlacik, R., Roman, A. and Wagner, H., 2012. Effective synchronizing
As a further step, strong auditing and security techniques can algorithms. Expert Systems with Applications, 39(14), pp.11746-11757.
be added to enhance the internal security of this framework.
Table 3 shows the performance analysis of proposed protocol
in comparison with existing protocol.

Protocol Initialization Join Leave

Rounds h 2 Min(k+1, h)
TGDH Messages 2*n-2 3 Min(2*N, n-
N)
Fig. 2: Hospital Group formation mechanism