Privacy protection of grid service in
distributed architectures
Jiaying Shi
Department of Computer Science
University of Windsor
Windsor, ON, N9B 3P4 Canada
shiv@uwindsor.ca
Abstract
Grid computing has extended the web
application in a large scale for a
request-response interaction between service
providers and service consumers. Yet, the
wide acceptance of the grid technology has
created pressure to add some features that
were not part of its original design, such as
security, privacy, and quality-of-service
support [4].
In this paper, we try to emerge a different
approach in protecting privacy during the grid
service in distributed architectures.
Keywords
Grid computing, privacy, onion routing,
distributed system
1. Introduction
The advent of Internet with its immense
usage around the world has lead computing
shift from traditional mainframe based
computing
and
subsequent
personal
computing to distributed computing and grid
computing. Aggregation of large number of
powerful nodes can achieve a huge computing
power by collaborating with each other. And
this same versatility makes it rather easy to
compromise data privacy in grid applications
[4].
The aim of our study on performance and
privacy in the distributed architecture is to
find out a better solution to achieve good
computing power while maintain good privacy
for the users.
In distributed architectures, to achieve a
better performance, we should not only take
the request-response time between providers
and consumers into consideration, but also the
local process time. To reduce the response
time, the recent trend of modern web system is
towards highly distributed intermediate
architectures that exploit server nodes located
close to the clients to replicate contents and
services, as can be observed in recent
literature [2].
Besides the cost, most adaptation
services are based on information concerning
the user, the so called user profile, which may
include sensitive information, such as
geographic location, navigation history and
personal preferences. Privacy issues related to
the management of user profiles are extremely
critical in the case of a system consisting of
geographically distributed servers, due to the
difficulty in guaranteeing a high level of
security for immense distributed nodes [2].
In this paper, we try to explore the
performance of using onion routing in the
distributed architecture through comparison
with using recent encryption method.
The remaining of this paper is organized
as follows. Section 2 discusses some related
works. Section 3 describes distributed-core
architecture, while section 4 describes the
onion routing. Section 5 briefly describes AES.
Section 6 explains our own approach. Finally,
section 7 contains some concluding remarks
and future works.
2. Related works
Traditional well-known identity
management and access control technologies,
e.g. X.509 certificates, Secure Sockets Layer
(SSL) communication protocol etc.
The earlier work [7] used dynamic user
credential management by using dynamic
token generation for the communicating
session. That scheme makes the grid
environment more security and less
hack-proneness. The dynamic token generated
forms parts of the private key and user id of
the client provides the public key in terms of
PKI (Public Key Infrastructure) [7].
There was another work using encryption
or signing a portion of a XML document so
that same XML can float through multiple
service providers with different sections
locked for view by a particular server. The
particular server can unlock only his
designated portion. Then the XML content
floats to next server in the chain. Finally, it
comes back with all the reply, keeping
confidentiality and privacy of the individual
service providers.
3. Distributed architecture
After doing some research on other’s
works, we realize the importance of providing
privacy in the user profile management. The
information contained in the user file can be
divided into two parts: personal, the
information considered to be confidential and
impersonal, information can be made public.
So how to protect the confidential
information safe is what we considered. We
need to build architecture to put this problem
in.
The architecture we used called
distributed-core architecture which is assumed
by Claudia in their paper [2]. They divided the
server nodes into edge and core nodes.
For edge level, they are considered
located close to network edge.
For core level, they are placed in
strategic positions and well connected.
The goal is to achieve high
user-perceived performance. One hand, the
distributed nodes will allow parallel
adaptation involved in a client request, the
other hand the edge nodes will allow to move
service close to the clients.
Figure 1. Distributed-core architecture
Figure 1 describes how the service of a client
request goes in the distributed-core
architecture.
(Step 1) The client sends a request to edge
node, it extracts the user ID and applies the
hash function H(x)1 to identify the
authoritative core node that owns the user
1
The user profiles are mapped on the core nodes through a hash
function H(x), that receives a (unique) user ID and returns a core
node identifier k = H(ID), where k ∈ [1, . . . , n], and n is the
number of core nodes.
profile.
(Step 2) The request is forwarded to the
authoritative node.
(Steps 3 and 4) fetches the content from the
provider server (and, if required, carries out
content adaptation (step (5) - numbers within
brackets are optional steps). Then, the
authoritative core node extracts the embedded
Web resources. On the basis of this list and of
the user profile information, the authoritative
core node decides how to distribute the
adaptation of the embedded resources among
the core nodes of the architecture, including
itself, and the edge node contacted by the
client (step 6). [2]
(Step 7& 8) The information about the
resources distribution is sent back to the edge
node and then to the client. In this way, the
edge node knows which nodes are responsible
for the adaptation of the different embedded
resources and is able to forward to the
appropriate node the subsequent requests
coming from the client [2]. This will reduce
the request-response time greatly.
4. Onion routing
Routing onions are data structures used to
create paths through which many messages
can be transmitted. To create an onion, the
router at the head of a transmission selects a
number of onion routers at random and
generates a message for each one, providing it
with symmetric keys for decrypting messages,
and instructing it which router will be next in
the path. Each of these messages, and the
messages intended for subsequent routers, is
encrypted with the corresponding router's
public key. This provides a layered structure,
in which it is necessary to decrypt all outer
layers of the onion in order to reach an inner
layer [6].
Figure 2. Onion routing path
Figure 2 describes how the onion routing
works. The client initializes communication
and make request send to application proxy
and onion proxy gives the order to choose
random router to form a random route, which
is based on the encryption algorithm. Once the
path has been specified, it remains active. The
sender can transmit equal-length messages
encrypted with the symmetric keys specified
in the onion, and they will be delivered along
the path. As the message leaves each router, it
peels off a layer using the router's symmetric
key, and thus is not recognizable as the same
message. The last router peels off the last
layer and sends the message to the intended
recipient [4].
Onion routing also includes a technique
allowing recipients to send responses back to
the sender, without compromising the identity
of either party. To initiate a two-way
conversation, a sender generates both an onion
and a reply onion. The reply onion is
transmitted to the recipient, who then uses it to
initiate the return path. Because the reply
onion is multiply-encrypted, it provides little
information that might compromise the sender
— an attacker must either break the
public-key encryption, or alternatively
compromise all of the routers in the return
path [4].
Since traffic analysis also is a serious
menace to agent-based applications. In order
to protect the communications between the
agents against the traffic analysis attacks like
communication pattern attack, timing attack,
etc. The primary goal of onion routing is to
provide strongly anonymous communications
in real time over a public network with
reasonable cost and efficiency. A secondary
goal is to provide anonymity to the sender and
receiver, so that the receiver may receive
messages but be unable to identify the sender,
even though the receiver may be able to reply
to those messages [4].
5. AES
AES (Advanced Encryption Standard) has
been analyzed extensively and is now used
widely worldwide. As of 2006, AES is one of
the most popular algorithms used in
symmetric key cryptography [8]. AES has a
fixed block size of 128 bits and a key size of
128, 192 or 256 bits. We decided to use 128
bit as the key size, as the longer the key size,
the longer time consumed in computation.
AES has four rounds like:
1. SubBytes — a non-linear substitution step
where each byte is replaced with another
according to a lookup table.
2. ShiftRows — a transposition step where
each row of the state is shifted cyclically a
certain number of steps.
3. MixColumns — a mixing operation which
operates on the columns of the state,
combining the four bytes in each column.
4. AddRoundKey — each byte of the state is
combined with the round key; each round
key is derived from the cipher key using a
key schedule [8].
6. Our approach
After we learnt about the distributed
architecture, and onion routing network, we
now can put our idea into implementation. As
we all talked about the server nodes on how to
unleash better performance, now we are
aiming to discuss the client nodes. We know
the client nodes may be distributed in a large
scale of geographical area. Using the
characteristic of onion routing, we pick up a
onion node in the network, initialize the
communication and pass it to an onion proxy
agent which will then choose several random
onion agents to form a random route. Each
agent in the route will encrypt data using a
AES and send to the next node. Unlike
making socket connection directly to the target,
that is why each onion router can only identify
the previous and net hops in a route. In the
anonymous connection, data passes along
without worrying about being tracked.
7.1 Conclusions
The goal of this paper is to produce an
analysis of the onion routing applied with AES,
one of the symmetric key cryptographies.
Although we can not totally sure the new
method we applied whether it will achieve a
better performance which would be seen after
the detailed experiment, we still can find out
the new method we used to protect the privacy
of users did much better compared to the
traditional methods which we mentioned in
the related works. It is more secure and
provides more confidential things for users.
And in our paper, we just use onion routing
for client nodes which is part of distributed
system. We need to investigate whether it is
feasible to extend it to sever nodes.
7.2 Future works
There are also many other kinds of
mathematical
algorithms
include
the
symmetric key cryptography like AES, IDEA,
3-DES we can put them into experiments of
onion routing and later compare the
robustness from the degree which can be
broken in shortest time, then pick up the best
symmetric key cryptography using in onion
routing.
And also the onion routing got its own
weakness like it did not provide much to
defend the timing analysis. To make sure
onion routing to be a safer one, we should
make improvement on the onion routers and
path formation methods. And it involves
immense
mathematic
algorithms
and
cryptographies, may be it’s not our capability
to do, and we need to wait until other’s work.
References
[1] Jana, D.
Chaudhuri, A.
Datta,
A. Bhaumik, B.B. Privacy Protection of
Grid Services in a Collaborative SOA
Environment, TENCON 2005 2005 IEEE
Region 10 Nov. 2005 Page(s):1 – 6
[2] Canali, C.; Colajanni, M.; Lancellotti, R.
Distributed
Architectures
for
High
Performance and Privacy-Aware Content
Generation
and
Delivery,
Automated
Production of Cross Media Content for
Multi-Channel Distribution, 2006. AXMEDIS
'06. Second International Conference on Dec.
2006 Page(s):11 – 18
[3] Porras, P.A. Privacy-Enabled Global
Threat Monitoring, Security & Privacy
Magazine, IEEE Volume 4, Issue 6, Nov.-Dec.
2006 Page(s):60 – 63
[4] Yu, Jiong; Cao, Yuanda; Lin, Yonggang;
Tan, Li. Research on Security Architecture and
Privacy Policy of Grid Computing System,
Semantics, Knowledge and Grid, 2005. SKG
'05. First International Conference on Nov.
2005 Page(s):3
[5] Smith, M.; Engel, M.; Friese, T.;
Freisleben, B.; Koenig, G.A.; Yurcik, W.
Security issues in on-demand grid and cluster
computing,
Sixth
IEEE
International
Symposium on Cluster Computing and the
Grid Workshops, 2006. Volume 2, 2006
Page(s):14 pp.
[6] Onion routing, from Wikipedia 2007,
http://en.wikipedia.org/wiki/Onion_routing
[7] Jana D., Chaudhuri A., Datta A., Bhaumik
B B. Dynamic User Credential Management
in Grid Environment, IEEE International
Region 10 Conference, Proceedings of the
IEEE TENCON 2005, Nov.21-24, 2005.
[8] AES, Advanced Encryption Standard, from
Wikipedia 2007,
http://en.wikipedia.org/wiki/Advanced_Enc
ryption_Standard
[9] Figure 2. From Internet 2007,
http://ntrg.cs.tcd.ie/undergrad/4ba2.05/grou
p10/OnionRouting.png