A Stateful Multicast Key Distribution Protocol
Based on Identity-based Encryption
Yunyun Wu
School of Information
Science and Engineering
Yunnan University
Kunming, 650000, China
wuyunyun@mail.ynu.edu.cn
Jing Liu
School of Software
Yunnan University
Kunming, 650000, China
liujing@ynu.edu.cn
Jingyu Hou
School of information
Technology
Deakin University
Burwood, VIC 3125, Australia
jingyu.hou@deakin.edu.au
Shaowen Yao
School of Software
Yunnan University
Kunming, 650000, China
yaosw@ynu.edu.cn
Abstract—Group key is used to encrypt group data in group
communication. Multicast key distribution scheme updates and
distributes group keys safely when a member joins or leaves
the group. However it has problems when arbitrary group
members want to build a dynamic conference, because traditional
distribution scheme uses group controller to manage the work. In
fact, a n-member group will have 2
n
− n − 1 possible conferences.
Moreover, members may need to join and/or leave the group
dynamically. If a group controller deals with all the members’
requests, it will be the performance bottleneck.
In this paper, we propose a new stateful multicast key
distribution protocol based on identity-based encryption. This
protocol can distribute the group keys safely and allows the
group members to build dynamic conferences by themselves.
It greatly reduces the group controller’s workload. In addition,
compared with traditional rekeying messages contain identifier
and group key’s version, our protocol can reduce the size of
rekeying messages by customizing the public key. We provide
three algorithms (INIT, JOIN, LEAVE) for rekeying when a
group membership changes and a revocation algorithm for
building dynamic conferences without the group controller. We
also give the security proof of our protocol in a symbolic security
model and the implementation of the protocol.
Index Terms—Group key distribution, Identity-based encryp-
tion, Security, Multicast
I. INTRODUCTION
Confidentiality is a requirement of many group communi-
cation applications, such as software distribution, stock quote
streaming, web caching, and multimedia conferencing. Both
confidentiality and access control can be enforced by encrypt-
ing the group data and distributing the Group Key (GK)
to current members. Moreover, to support group changes,
the group key must be updated when group members join
or leave the multicast group to comply Forward Secrecy
and Backward Secrecy. When a group member leaves, the
departing member cannot decrypt the future group communi-
cation. This is called Forward Secrecy. When a new member
joins, the new member cannot decrypt the previous group
communication. This is called Backward Secrecy[1].
Group key distribution schemes emphasize on keeping the
GK fresh, when a group member joins or leaves a group.
Traditional schemes manage all keys, rekeying, and clients’
requests by the Group Controller (GC) only. However they
cannot handle such scenario where a group of members or
an arbitrary subset (of size at least 2) of members wants to
form a privileged subgroup. We call such a subgroup Dynamic
Conference. If the number of members in the group is n, there
will be up to 2
n
− n − 1 possible conferences in the group.
Moreover, the members may need to join and/or leave the
group dynamically. GC’s work is huge when dealing with all
these events, and its efficiency therefore.
Currently, there are some dynamic conference
schemes[2][3][4][5][6] that solve the above problems.
Different from the major key distribution schemes, dynamic
conference schemes focus on the process of building
conferences among group members. Those schemes are either
based on symmetric encryption or asymmetric encryption.
The former may overload GC, while almost all of the latter
have been proven that it couldn’t satisfy forward secrecy and
backward secrecy.
In this paper, we propose a new stateful multicast key
distribution protocol based on identity-based encryption. It can
update and distribute a new GK when a group membership
changes, and set up dynamic conferences without GC. In this
protocol, we use a binary tree (key tree) to maintain keys,
and the leaf nodes in the tree are associated with the group
members. Each node has a pair of public/private key which is
called Key Encryption Key (KEK). A group member knows
the structure of the key tree, KEKs of his/her corresponding
leaf node, Associated Nodes which are nodes in the path
from its leaf node to the root and the entire nodes’ public
keys. When a group member joins or leaves, all the nodes
including the member’s leaf node and associated nodes will
be updated. The root’s public key is used to encrypt the
GK. If a group member wants to build a conference with
other group members, he/she will find the sub-trees only
covering the necessary members. And the member who sets
up the conference will encrypt the conference key with the
root’s public keys of these sub-trees and distribute it. Our
contribution has two merits and one potential merit as follows:
• Group members can set up a dynamic conference by
themselves. In traditional methods, GC is the central part
dealing with all the requirements from members. This
merit can greatly reduce the GC’s workload.
• The real rekeying communication overhead can be re-
978-1-5090-5507-4/17/$31.00 ©2017 IEEE
ICIS 2017, May 24-26, 2017, Wuhan, China