Symmetric-Key Based Privacy-Preserving Scheme For Mining Support Counts

Advanced Computing: An International Journal ( ACIJ ), Vol.4, No.2, March 2013

SYMMETRIC-KEY BASED PRIVACY-
PRESERVING SCHEME FOR MINING
SUPPORT COUNTS
Yu Li1 and Sheng Zhong
1
Department of Computer Science and Engineering, University at Buffalo, NY, USA

yli32@buffalo.edu

ABSTRACT
In this paper we study the problem of mining support counts using symmetric-key crypto which is more
efficient than previous work. Consider a scenario that each user has an option (like or unlike) of the
specified product, and a third party wants to obtain the popularity of this product. We design a much more
efficient privacy-preserving scheme for users to prevent the loss of the personal interests. Unlike most
previous works, we do not use any exponential or modular algorithms, but we provide a symmetric-key
based method which can also protect the information. Specifically, our protocol uses a third party that
generates a number of matrixes as each user’s key. Then user uses these key to encrypt their data which is
more efficient to obtain the support counts of a given pattern.

KEYWORDS
Privacy-Preserving, Symmetric-key encryption, Support Counts

1. INTRODUCTION

With the rapid development of information technology, personal data should be protected in order
to respect the people's private information and prevent the inappropriate usage. Most Internet
users will be asked to provide an opinion that whether they support some pattern or not, then a
data miner can count the number of this support. However, some websites collect, store and share
these personal private data in order to benefit themselves despite users' desire. Privacy-preserving
data mining has received a lot of attention in the past few years. This kind problem can be mainly
divided into two categories: (a) the data is divided among more than one different party and the
privacy preserving method needs to protect each party's private data; (b) the data is to be
published for research, but due to its sensitivity the data should be modified [1].

A lot of typical privacy-preserving data mining problems include the scenario that mentioned
above belongs to the first category. This kind of privacy-preserving data mining problem is called
secure multiparty computation [1]. That means different party wants to jointly compute the
correct result without revealing any private data. The most common way to provide privacy-
preserving method is using cryptographic technology. Huge amounts of research have been
investigated in this field after [2] provided an elegant privacy preserving data mining method
using cryptographic tools.

DOI : 10.5121/acij.2013.4201 1

In this paper, we propose a symmetric-key crypto based privacy-preserving scheme for mining
support counts using symmetric-key encryption. Our scheme is more efficient than previous work
[3]. The rest of this paper is organized as follows. Section II describes the related work. We
introduce in symmetric-key cryptosystem and block cipher in Section III. In Section IV, we
describe our symmetric-key based privacy preserving support counts protocol and analyse our
correctness of our protocol. Section V gives the discussion of work. Last, Section VI gives the
conclusion of this paper.

2. RELATED WORK
Providing privacy-preserving for Internet data is a longstanding goal of the computer research
community. It is has received considerable attention with the development of data mining and
network technology. Cryptograph based privacy-preserving method can provide a better
guarantee of the privacy when different institute want to cooperate in a common goal. In [4], Lea
Kissner and Dawn Song propose efficient techniques for privacy-preserving operations on
multisets using cryptosystem. In [5], the authors proposed a security protocol for the IVC
applications based on group signature and ID-based signature schemes. In [3] the authors propose
an identity-based protocol for support count that preserves privacy using public key cryptosystem.
Different from the above reported schemes, we propose a symmetric-key based privacy-
preserving protocol, which cannot only guarantee the requirements of security and privacy but
can also provide a very good efficiency. Furthermore, the protocol and infrastructure are easy to
be implemented. Thus, the proposed protocol can practically be launched for enabling the
application of Internet website.

3. TECHNICAL PRELIMINARIES
In this section, we give a brief review of symmetric-key cryptosystem and block cipher that used
in our privacy-preserving scheme for mining support counts.

3.1. Symmetric-key Cryptosystem

A cryptosystem [6] can be defined as a tuple . The denotations are shown in
table 1. The cryptosystem must be efficient and must be sufficiently random so that it is hard to
guess. The encryption and decryption process must be efficient. We
must have . We have a symmetric-key cryptosystem when encryption key is
same with decryption key which means .Without the decryption key, there is no efficient
algorithm that can decrypt ciphertexts. There are two requirements for a symmetric cryptosystem.
The first one is that it is impractical to get a plaintext only known the ciphertext and encryption
scheme. Another one is that all users must obtain the key in a safe way and keep their keys
secure.
Table 1. Denotations of Cryptosystem

Summary of Denotations
M: cleartext space
C: ciphertext space
K: key space
G: key generating algorithm
E: encryption algorithm
D: decryption algorithm
G: generate a pair of encryption/decryption keys for a given length
2

3.2. Blockcipher

A block cipher [7] is a function . There are two inputs in this
algorithm. The first input is the key whose length is . The second input is the plaintext whose
length is , and the output is the ciphertext whose length is same as the length of plaintext. The
length of the key and the length of the block are parameters associated to the blockcipher.
When using block cipher, a random key is assigned and kept secret between users. The function
is then used by a user to process data before she sends it to other user. Typically, we will
assume the adversary will be able to obtain some input-output examples for which are pairs of
where . But the adversary will not be shown the key because the security
relies on the secrecy of the key. So the goal of the adversary is recovering the key by given some
input-output examples of . However, a good blockcipher scheme should be designed to make
this task computationally difficult.

4. SYMMETRIC-KEY CRYPTO BASED SUPPORT COUNT PROTOCOL

In this section, we go into the detail of our symmetric-key crypto based support counts mining
protocol and prove its correctness.

4.1. Protocol

In our symmetric-key crypto based support count protocol, we use a third party key-generator.
First, let the generator generate two matrixes. The dimensions are based on the number of users.
Assume there are users in total, the dimension of the two matrixes should be . Then the
generator fills the first matrix with random numbers. After that, generator fills the second matrix
with the negative values of first one. Note that the filling order of the second matrix should be
permuted. Then we define the symmetric-key for each user as the row of and
column of .

Then each user performs our protocol shown in algorithm 1.

Algorithm 1 Symmetric-key based privacy preserving scheme
Initialize
Step 1
User 1 will be asked whether he support one given pattern or not. Define support as 1
and not support as 0. Then he uses the two keys distributed to him to encrypt his answer. He
will get the value of and sends the value to User 2.
Step 2
For each user , User receives User computes
and sends it to user
Step 3
User receives . User adds to the received
number, and sends the results to all other parties.
Step 4
Each user uses the block cipher to update their keys.

3

Here, we specify that how to update user 's keys using blockcipher in step 4. Since is the
sum of elements in row in matrix which is . Therefore we use
algorithm 2 to update the .

Algorithm 2 Keys updating algorithm
Initialize
Step 1
Each user will be assigned a blockcipher as well as a key during the
initialization stage.
Step 2
Using blockcipher method, we define a function which makes

Step 3
User uses the same method to update the second key . First the user converts the
value of to absolute value. Then make them as negative after update the value use step 2.
Step 4
After mining rounds, the server reassigns another and to each user.

In keys updating process, blockcipher and key are chosen randomly by third party key-
generator and make known to all users. In order to enhance the security level of our protocol, we
let the generator create a new blockcipher and key and send them to all users after
rounds of support counts mining.

Unlike most previous works, we do not use any exponential or modular algorithms, we provide a
symmetric-key based method which can also protect the information and the execution time is
faster than public-key based algorithm. Since in modern network service, the number of user can
be as large as millions. Therefore the efficiency is most essential when designing of the privacy-
preserving algorithm.

4.2. Correctness

In this section, we prove the correctness of our symmetric-key crypto based support counts
mining protocol.

Theorem 1. Our symmetric-key crypto based support counts mining protocol correctly computes
the sum of all users' outputs.
Proof. Now, we prove that the user can obtain the final result by following our protocol.

Since we have

Therefor we can get

It is proved that .

4

5. DISCUSSION

Our proposed symmetric-key crypto based protocol for mining support counts has the following
characteristics.

It preserves the data privacy for each user without revealing it to others in the process of mining
support counts. Our privacy-preserving protocol is accurate which means the result obtained by
our protocol is the same as those obtained by having all data collected without privacy-
preserving.

We are utilizing a symmetric-key based method in our models. The protocol is much more
efficient compared with public-key based approached. An existing work [3] in mining support
counts proposed the privacy-preserving was based on public-key based approach. Different from
their protocol, we focus on another cryptographic method---symmetric-key cryptosystem.
Furthermore, our protocol is more efficient than previous work.

6. CONCLUSION

In this paper, we proposed a symmetric-key crypto based privacy preserving protocol for mining
support counts, in order to obtain the result without revealing users' personal private information.
We provided a very efficient method based on symmetric-key cryptosystem and cipher block to
encrypt each user's data and proved the correctness of our protocol.

REFERENCES
[1] Y. Lindell & B. Pinkas, (2009) Secure multiparty computation for privacy-preserving data mining.
Journal of Privacy and Confidentiality, Vol. 1, No.1, pp5.
[2] Y. Lindell & B. Pinkas , (2000) Privacy preserving data mining. In Advances in Cryptology
CRYPTO, Springer, pp36-54.
[3] F.Wu, J. Liu & S. Zhong. (2009) An efficient protocol for private and accurate mining of support
counts. Pattern Recognition Letters, Vol. 30, No.1, pp80-86.
[4] L.Kissner & D.Song. (2005) Privacy-preserving set operations. In Advances in Cryptology
CRYPTO, Springer, pp241-257.
[5] X.Lin, X.Sun, P.H. Ho & X.Shen. (2007) Gsis: a secure and privacy-preserving protocol for vehicular
communications. Transactions on Vehicular Technology, IEEE, Vol. 56, No.6, pp3442-3456.
[6] H.Delfs & H.Knebl. (2007) Symmetric-key encryption. Introduction to Cryptography, Springer ,
pp11-31.
[7] T.W. Cusick & P.Stanica. (2009) Cryptographic Boolean functions and applications. Academic Press.

5

Symmetric-Key Based Privacy-Preserving Scheme For Mining Support Counts

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Symmetric-Key Based Privacy-Preserving Scheme For Mining Support Counts (20)

More from acijjournal (20)

Recently uploaded (20)

Symmetric-Key Based Privacy-Preserving Scheme For Mining Support Counts