Mitigation of a denial of existence attack in a

P2P network

A. Ruiz Calderon, C. Mex-Perera

1
Department of Computer Systems ITNL
Av. Eloy Cavazos 2001 Col. Tolteca CD Guadalupe, N.L. Mexico
2
Center for Electronics and Telecommunications, ITESM, Campus Monterrey
Av. Eugenio Garza Sada 2501 Sur, Col. Tecnologico
Monterrey, N.L., CP 64849 Mexico
armandoruizmex@gmail.com,carlosmex@itesm.mx

Abstract. Currently, networking and Internet activities are common

in our society. Such activities represent an important way of commu-
nication, commonly we use the client server model, it has been widely
studied and tested, however that model has several limitations that affect
its performance. There are several alternatives to solve those limitations.
One of them is the P2P architecture. Such architecture provides high ac-
ceptance, and enhance the network performance. The use of specialized
overlays such as Pastry is a feasible alternative to exchange messages in
a wide distribution applications. But on P2P architectures the security
aspect have not been solved. The Denial of Existence attack is a security
problem that requires to be focused. Such attack increases the quantity
of lost messages. It causes more traffic on the network. Our analysis and
modifications on P2P Pastry networks shows there are an important im-
provement with regards to lost messages.

Key words: Attacks, Denial of Service, Denial of Existence, P2P

1 Introduction

RELATED WORK.

the behavior of the impersonate user is taken as a characteristic pattern

which is valid only for this user.

The present work is organized as follows. In Section 2 are described the

common schemes use for detecting illegitimate users. Section ?? includes the
proposed methodologies for building masquerade sessions. Section 4 presents the
obtained results with each proof. Finally, in Section 6 are given the conclusions
and some ideas for future work.

2 Schemes for Detecting Attackers

2 A. Ruiz Calderon, C. Mex-Perera

3 Methodology
In this work we develop a set of simulations. Considering the following facts and
conditions:

1. All the messages are routed in the correct way without delays
2. There are no bottlenecks
3. A node receive a request and responds to it
4. The network does not have bad segments

To build the environment, there were considered the following aspects:

1. b is used as b = 2 where b is a configuration parameter of Pastry

2. The generated environment has a capacity of N elements where N = 5000
3. There were created n elements where n = 500 and is the number of nodes
that will be used
4. The assignment of an ID to the nodes in was performed in a randomly way
5. Each ID is unique
6. The nodes were sorted in a logical ring based on the assigned ID
7. The leaf set and the routing table were generated according to Pastry rules

On the generated network environment there are not incoming or outgoing

nodes while it is in operation. So, the neighborhood set was not developed. For
the network, such table is considered as empty table. In the case of a routed
message falls in the rare case, we joined both the leaf set and the routing table.
After that, we looked for the best match on both tables.

Once generated the environment and their behavior, we consider the following
variables to analyze.

1. The probability of failure of a routed message Pf . This probability is defined

by the amount of lost messages during the experiments expressed as:

mp
Pf = (1)
n

where mp , is the amount of lost messages on the test and n is the amount
of used nodes.
2. The average of hops is defined by the amount of visits that a single message
do on a single node before it arrives to the final destiny

In this scenario, the attack is generated by a coalition of nodes (E1 , ..., EnC ).

These nodes begin the attack in a coordinated way. Besides, this attack is
considered selective. If the selected target node is denied in a DHT of a node,
the attack is completed. Otherwise, the routing process will follow the Pastry
rules. Performing this behavior, the malicious node can hide itself and it is more
difficult to find and discover the coalition of nodes.
The attack works in the following way:
 Mitigation of a denial of existence attack in a P2P network 3

Coalitioned nodes

Target

Direction of the attack

Fig. 1. Scheme of an attack with a coalition of malicious coordinated nodes in a P2P

1. The network is working in a normal way

2. There is a trigger that begins the attack
3. The target is selected randomly
4. The coalition of nodes are selected randomly
5. The malicious nodes are randomly on the complete network

4 Results

The conditions on the Pastry architecture were simulated on a Linux environ-

ment by using PERL v5.8 language. The generated environment had N = 5000,
and n = 500, according to the theory of the DHT. It recommends that solely
the 60% of the capacity on the Pastry tables should be used. Following such
recommendation, only the 10% of the capacity on the Pastry tables was used.
The tests had different amount of nodes with coalition behaviour. For each test
50,000 messages were sent.
The results are shown in the table 1. It enlists the probability of routing fail
shown by percentages for each different tests. It was based on the number of
nodes with coalition behaviour as consequence of the Selective attack.

Nodes with coalition behaviour Pf

0 0%
1 0.182%
2 0.348 %
5 1.02 %
10 1.99 %
20 3.78 %
50 9.99 %
100 20.02%

Table 1. The table shows the amount of lost messages based on the quantity of nodes
with coalition behaviour and the probability of routing fail N=50,000 and n=500
4 A. Ruiz Calderon, C. Mex-Perera

The table 1 shows that the percentage of lost messages is proportional to the
number of nodes with coalition behaviour. It can be seen on the figure 2, as a
result of a denial of existence attack and without ignoring Pastry rules.

2
10
Percentage of lost Messages

1
10

0
10

0 10 20 30 40 50 60 70 80 90 100
Number of Coalitionated Nodes

Fig. 2. Graphic that shows the percentage of lost messages, versus the number of nodes
with coalition behaviour

Once the routing failure probability was obtained, this probability was com-
pared with the theoretic routing failure probability given by the formulas

Enc
Ff = (2)
n

= (1 Ff )h1 (3)
Where

h = log2b (n) (4)

Pf is considered as the probability of routing failure where:

pf = (1 ) (5)
Considering:

1. n the quantity of nodes in the network

2. Enc the total number of nodes with coalition behaviour
3. Ff the fraction of nodes failing
4. is the probability of a satisfactory routing for a request between two work-
ing nodes with an Enc number of nodes with coalition behaviour
 Mitigation of a denial of existence attack in a P2P network 5

These formulas were taken from [16]. We appreciate that the routing failure
with coalitioned nodes probability by using these formulas is less than the Pastry
theory.
For the tests it was used the modified algorithm. It is important to remark
that with this algorithm malicious activity (denial of existence) is considered.
In the original algorithm of Pastry, the nodes do not have malicious activities,
just fails. With this fact, the results shown that the modified algorithm routing
failure probability is lower compared to the original algorithm [16], as These
results are displayed on the figure 3.

2
10
Percentage of lost messages

1
10

0
10

Pastry
Selective Attack

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Number of Coalitionated nodes

Fig. 3. The Graphic shows the comparison between the expected and observed routing
failure probability under a selective attack

The results from selective attack were compared with another attack called
brute forceThere are significative differences in the amount of lost messages.
This comparison was made to be able to observe the amount of lost messages
for each attack. These results are enlisted in the table 4.
The brute force attack always deny the existence of the target node without
care if the target node exists or not in the routing information. As consequence,
the nodes that performs this attack are easily discovered. The figure 4 shows the
comparison between both attacks.
The figure 4 shows the percentage of lost messages of both attacks.
To verify the randomness of the algorithms there is another test developed
with the following characteristics:

For each routing test, the target node is selected randomly

6 A. Ruiz Calderon, C. Mex-Perera

Coalitioned Nodes Selective Attack % Selective Attack Brute force % Brute force
0 0 0.00 % 0 0.00 %
1 91 0.182 % 460 0.92 %
2 174 0.348 % 923 1.846 %
5 508 1.02 % 2259 4.518 %
10 997 1.99 % 4435 8.87 %
20 1890 3.78 % 8415 16.83 %
50 4995 9.99 % 18702 37.404 %
100 10014 20.02 % 31402 62.80 %

Table 2. Difference of the quantity of lost messages between both kind of Attacks

2
10
Percentage of Lost Messages

1
10

0
10

Selective Attack
Brute Force Attack

0 10 20 30 40 50 60 70 80 90 100
Number of Coalitioned Nodes

Fig. 4. Graphic that shows the comparison of the lost messages with both attacks

5 Suggested Solution

As a result of the tests, it was observed that over 20% of the messages were lost
when the network had 100 nodes of 500 coalitionated, that represents 20% of
the total amount of nodes on the network.
The suggested solution is as follows:

If a message arrives to a node without information of the destiny, the message

returns to the previous node. On the current node in a randomly way, a new
node is selected from the leaf set, and then the message is routed to that
node. We consider to discard the node that do not have the information of
the destiny.
Once the new node is selected, then the message continue to that node and
follow the Pastry rules

This algorithm has a condition, and is the next:

 Mitigation of a denial of existence attack in a P2P network 7

The routing
B process conclude
satisfactory
5
Malicious Coalitioned
node
4
Verify if the target exists
The information of the Z in the information of Z
Ec2 and continue routing to
leaf set of F is used and
the message is routed to destiny B
Z

F Malicious Coalitioned
A Node
Ec1
The message arrives
Sends a message that to node F and sends 3
income to the network by it to Ec1 The EC1 deny the target
node A and returns the message
to F

Fig. 5. The figure shows the scheme of the proposed solution to the routing algorithm

When a message fall in a coalitionated node, the message returns to the

previous node and select another node of the leaf set, but if in the second round
fall again in a coalitionated node, the message is discarded. This condition was
made to prevent unnecessary traffic over the network, or if the message is jumping
over all the network it could be corrupted and increase the traffic. The table 3
shows the number of lost messages with the suggested solution.

Coalitionated Nodes Percentaje of Lost Messages

0 0%
1 0.008%
2 0.0.16 %
5 0.032 %
10 0.1 %
20 0.288 %
50 1.31 %
100 4.56 %

Table 3. The table shows the quantity of lost messages applying the suggested solution

The figure 6 shows the percentage of lost messages with both attacks and
with the suggested solution.

To verify the helpfulness of the solution, there were counted the number of
hops with a selective attack and with the suggested solution, the results are
shown in table 4.
8 A. Ruiz Calderon, C. Mex-Perera

2
10

1
10

Percentaje of Lost Messages

0
10

1
10

2
10
0 10 20 30 40 50 60 70 80 90 100
Number of Coalitioned Nodes

Fig. 6. Graphic that shows the percentage of lost messages, and the number of coali-
tionated nodes, with the brute force attack, selective attack and the suggested solution

Coalitionated Nodes Num. of Hops with attack Num of hops with the solution
0 2.83 2.83
1 2.8298 2.8513
2 2.8385 2.9127
5 2.8346 3.0192
10 2.8384 3.0802
20 2.8389 3.2438
50 2.8386 3.5274
100 2.84 3.7219

Table 4. The table shows the average of hops under a selective attack and with the
suggested solution

6 Conclusions

Based on the results of the test we can conclude that P2P networks, are valuable
alternative for storing, sharing and retrieving message applications.

The use of specialized overlays like Pastry for applications that store and
retrieve information, is a good alternative to solve storing problems, that the
Client-Server networks have, because P2P data manipulation is better than
the traditional model.

The balance of the network is good, the observed randomness is good, and it
can seem on the amount of lost messages is proportional to the coalitioned nodes.

The suggested solution is very efficient, because with the solution the amount
of lost messages with 100 coalitioned nodes is only 4.56%, comparing with the
 Mitigation of a denial of existence attack in a P2P network 9

20% of lost messages with the selective attack; this reduction is significantly.

The effectiveness of the suggested solution is given by the little increase of

the average number of hops that comes from 2.81 to 3.7219 with 100 coalitioned
nodes.

This increase does not affect the performance of the network.

This kind of networks are useful to deploy wide distribution applications like
DNS, however it is important to establish the distribution of the information
that will be used, overall in applications where the resources are not uniform.

References
1. A. B.; Dreger, H.; Feldmann, A.; Predicting the DNSSEC overhead using DNS
traces;Information Sciences and Systems, 2006 40th Annual Conference on 22-24
March 2006 Page(s):1484 - 1489
2. A. Passarella, Franca Delmastro, Marco Conti; XScribe: a stateless, cross-layer
approach to P2P multicast in multi-hop ad hoc networks; Proceedings of the 1st
international workshop on Decentralized resource sharing in mobile computing and
networking MobiShare 06; September 2006
3. A. Rowstron and Peter Druschel; Storage management and caching in PAST, a
large scale, persistent peer to peer storage utility; In Proc ACM SOSP 1; Banff
Canada Oct 2001.
4. A. Rowstron and Peter Druschel; Pastry: Scalable, distributed object location and
routing for large-scale peer-to-peer systems; In Proc. IFIP/ACM Middleware 2001,
Heidelberg, Germany; November 2001.
5. A. Friedlander, Allison Mankin, W. Douglas Maughan, Stephen D. Crocker;
DNSSEC: a protocol toward securing the internet infrastructure; Communications
of the ACM, Volume 50 Issue 6; June 2007
6. Ben Y. Zhao, John D. Kubiatowicz, and Anthony D. Joseph.; Tapestry: An
infrastructure for fault-tolerant wide-area location and routing; Technical Report
UCB//CSD-01-1141, U. C. Berkeley, April 2001.
7. D. Ervin Khuth; The art of computer programming; Stanford University Addison
Wesley; Vol 3 sorting and searching; U.S.A. 1973.
8. E. Damiani, Stefano Paraboschi, Pierangela Samarati, Fabio Violante; Peer to peer
networks: A reputation-based approach for choosing reliable resources in peer-to-
peer networks; Proceedings of the 9th ACM conference on Computer and commu-
nications security CCS 02; November 2002
9. G. Alexis; DNSSEC Operational Impact and Performance;Computing in the Global
Information Technology, 2006. ICCGI 06. International Multi-Conference on Aug.
2006 Page(s)55 - 63
10. I. Stoica, R. Morris, D. Karger, M Frans H. Balakrishnan; Chord: a scalable
Peer-to-Peer lookup service for Internet applications; SIGCOMM 01 ; ACM; San
Diego California; 2001
11. J. Kyeong Kim a, Hyea Kyeong Kim a, Yoon Ho Cho; A user-oriented contents
recommendation system in peer-to-peer architecture; Expert Systems with Appli-
cations 34 (2008) 300-312.
12. K. Hui, John C.S. Lui, David K.Y. Yau; Small-world overlay P2P networks: Con-
struction, management and handling of dynamic flash crowds; computer networks
num 50 2006.
10 A. Ruiz Calderon, C. Mex-Perera

13. L. Liu, N. Antonopolus, S. Mackin; Fault tolerant peer-to-peer search on small-

world networks; Future Generation Computer Systems; vol 23; march 2007; 921-931.
14. L. Zambenedetti Granville, D. Moreira da Rosa, C. Melchiors, M. J. Bosquiroli
Almeida, and L. M. Rockenbach Tarouco; Managing Computer Networks Using
Peer-to-Peer Technologies; IEEE Communications Magazine ; October 2005
15. M. Castro and Barbara Liskov. Practical byzantine fault tolerance. In Proceed-
ings of the Third Symposium on Operating Systems Design and Implementation
(OSDl99), New Orleans, Louisiana, February 1999.
16. M. Castro, P. Druschel, A. Ganesh, A. Rowstron, and D. Wallach. Secure Routing
for Structured Peer-to-Peer Overlay Networks. Symposium on Operating Systems
Design and Implementation, Boston MA, Dec 2002.
17. M. Naor and Udi Wieder. Novel Architectures for P2P Applications: the
Continuous-Discrete Approach. Proc. SPAA, 2006
18. N.Saxena, G. Tsudik, J. Hyun Yi.; Threshold cryptography in P2P and MANETs:
The case of access control; Computer Networks vol 51; 2007
19. P. Fltstrm, Daniel Massey, Vasileios Pappas, Lixia Zhang; Distributed DNS
troubleshooting; Proceedings of the ACM SIGCOMM workshop on Network trou-
bleshooting: research, theory and operations practice meet malfunctioning reality
NetT 04; September 2 004 .
20. R. Venugopalan, Emin Gn Sirer; The design and implementation of a next gen-
eration name service for the internet; ACM SIGCOMM Computer Communication
Review; Proceedings of the 2004 conference on Applications, technologies, architec-
tures, and protocols for computer communications SIGCOMM 04, Volume 34 Issue
4; August 2004
21. S. Andr outsellis-Theotokis, Diomidis Spinellis; A survey of peer-to-peer content
distribution technologies; ACM Computing Surveys (CSUR), Volume 36 Issue 4;
December 2004.
22. S. Ariyapperuma, C.J Mitchell; Security vulnerabilities in DNS and DNSSEC;
Availability, Reliability and Security, 2007; The Second International Conference on
10-13 April 2007 Page(s):335 - 342; ARES 2007
23. S. Iyer, Antony Rowstron, Peter Druschel; Squirrel: a decentralized peer-to-peer
web cache; Proceedings of the twenty-first annual symposium on Principles of dis-
tributed computing PODC 02 Publisher: ACM Press; July 2002
24. S. Ratnasamy, Paul Francis, Mark Handley, Richard Karp, Scott Schenker; A
scalable content-addressable network;; ACM SIGCOMM Computer Communica-
tion Review , Proceedings of the 2001 conference on Applications, technologies,
architectures, and protocols for computer communications SIGCOMM 01, Volume
31 Issue 4; August 2001
25. Zhang Rongmei and Charlie Hu; Borg: A Hybrid protocol for scalable application
level multicast in peer to peer networks; Nossdaw 03; Monterey California; June 03.