Challenges and Methods of Data Perturbation Techniques

International Journal of Emerging Technology and Advanced Engineering

Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 7, Issue 11, November 2017)

197

Challenges and Methods of Data Perturbation Techniques

Thanga Revathi S

, Dr. N. Ramaraj

¹MNM Jain Engineering College, Chennai, India ²Vignan University, Guntur, India

Abstract-- Data mining is the extraction of valuable knowledge from large data. Nowadays Privacy Preserving Data mining (PPDM) is the most important issue among the researchers. As the privacy preservation Data mining permits exchange of Data among all users for analytical purpose. And Nowadays Data mining is done on Dynamic data rather than the traditional static data. The Data Mining techniques used today consider one main issue, the privacy of the critical and sensitive Data. There are various techniques involved in the privacy preservation of critical data. Data Perturbation is an important method to preserve privacy of data. Data perturbation is the data security technique that modifies the database to preserve the privacy and confidentiality. It is used for both data privacy and accuracy. In this paper we are going to discuss about the various perturbation techniques that can be used to preserve data privacy and discuss about the implications of the techniques.

Keywords: Data Perturbation, Confidentiality, Data Mining, Privacy Preserving

I. INTRODUCTION

Today data from all the organizations are collected about the entire organization structure, Human resource, work flow, etc., And the organizations suffers to get the full information or knowledge from the data. Sophisticated organizations make use of the Data mining algorithms to extract the unknown patterns or knowledge from the data, which may also access the confidential data stored in the Database. This creates the need for the Database administrator to protect the confidential record about the individual in the organizational database from improper disclosure. Some of the commonly used techniques for privacy preserving in cloud[12] are

1. Reconstruction (or) Perturbation method

2. Anonymization method

3. Cryptographic method

The Reconstruction Method:

Reconstruction is popular method for the Privacy Preserving Data Mining technique. The sensitive data are transformed or masked by adding additional data to the original data[12]. This method provides greater security for the individual data but the efficiency depends on the noise added to the data.

The Anonymization Method:

In this method the individual record is made indistinguishable using Suppression and Generalization techniques. K- anonymity is a common algorithm for this process. K-anonymity conceals the records which are commonly used as the unique identifier for the data. This method is vulnerable to some attacks like linking attacks.

The Cryptographic Method:

This method is mainly used when data mining is mutually done by different parties on the same data. This case requires preserving the privacy data mining of the parties. Several algorithms have been developed for this privacy preserving data mining [12]. This method provides higher security and utility when compared to other methods. But, falls back in efficiency when compared with the other methods.

Data Perturbation:

A data perturbation procedure can be simply described as follows. Before the data owners publish their data, they change the data in certain ways to disguise the sensitive information while preserving the particular data property that is critical for building meaningful data-mining models. Perturbation techniques have to handle the intrinsic trade-off between preserving data privacy and preserving data utility, as perturbing data usually reduces data utility [3].

Types of Data Perturbation Techniques:

The different type of Perturbation techniques falls into two main categories. i) Probability Distribution category and ii) Fixed data perturbation category. In the Probability Distribution Category a sample of the whole population is considered and while performing the perturbation this sample is replaced by another set of samples in the database [13]. In the Fixed method, the Data is replaced by other set of data and it is done only once. It is not dependent on the samples, it is done individually and it is done once for all.

International Journal of Emerging Technology and Advanced Engineering

198

In the next method the density function of the attributes is identified and the estimate the values of the functions. Then a series of samples is generated from the estimated values which replaces the actual database of the same rank order and same size[2].

Data perturbation approach is classified into two methods. First is the Probability Distribution were the original database is replaced by the sample in the distribution or the same distribution itself. Next is the Value Distortion approach were the data is perturbed directly by adding noise or multiplicative noise or by some randomized noise[1].

The Data Perturbation approach is classified as 1) Rotation Perturbation

2) Projection Perturbation 3) Geometric Data Perturbation.

Rotation Perturbation

In this method the value of the two attributes in the matrix is rotated but the meaning of the value is protected. The pair of attributes is first selected then the value distortion technique is applied for those values [7]. Suppose original dataset have d column and N records then it is represented as X d×n, the rotation perturbation of dataset X

will be defined as G(X)=RX, where R d×d is a random

rotation orthonormal matrix. [9]

Projection Perturbation:

The data perturbation is done by moving the data value in high dimensional space to the lower dimensional space randomly. This can be projected either column or row wise [8]. Let Pk×d be a random Projection matrix, where P’s rows are orthonormal.

G(X) = (Sqrt(d)/k) PX

is applied to perturb the original data set X. [9]

Geometric Perturbation:

This is a hybrid technique with the combination of rotation, translation and adding random noise value to the given data value in the matrix to provide quality of data preservation mainly for clustering [9].

G(X) = RX+T+D;

Matrix (X) d*n indicates the original data set with d-number of columns and n records, (R) d*d be a random rotation matrix and D be a random noise matrix and T is the Translation matrix.

Transformation:

Translation Transformation is done by adding a constant value to all the attribute values.

The matrix can be moved from (X,Y) to (X’,Y’). The original data cannot be viewed directly. It is represented by v’ = Tv, where T is transformation matrix.

Rotation Transformation is performed by rotating a pair of attributes to a certain angle with respect to origin.

II. BDATA PERTURBATION METHODS

Geometric Perturbation

Random geometric perturbation[3[ is the linear combination of the three components: rotation perturbation, translation perturbation, and distance perturbation.

G(X)=RX+T+∆.

The Geometric properties are preserved by the modified data. In random rotation perturbation technique, multiple column data values are moved into single column transformation. But it leads to lack of privacy in multi dimensional rotation perturbation. In this paper a unified privacy metric was proposed to address the problem. And this metric was analyzed against the naive – interference attacks, ICA based attacks and distance interference attacks. And also proved for greater accuracy of rotation-invariant classifiers against other techniques.

In this paper[2] geometric transformation technique is used including translation, scaling, rotation, which transforms the data by preserving the similarity between data objects. Geometric data translation is a type of data perturbation where the original data is translated to other form by adding some noise. Addition of noise is possible to categorical data and also numerical data. Addition of random noise may cause some loss of data which leads to less accuracy of results. To minimize the loss of data this paper uses Gaussian noise to perturb data which provide less loss in data. The geometric data perturbation was done only for numeric data which can be extended to non-numeric data using K-anonymization technique and for streams using Stream Analysis tool.

Rotation Perturbation

International Journal of Emerging Technology and Advanced Engineering

199

Data perturbation approach was used in the single level trust of Data Miners. The perturbation approach uses random perturbation for the individual values to preserve the privacy of the data before it is published. But this is not suitable for the single level trust on data miners. This paper[3] extends the trust level to the multi level trust on data miners, were the higher trust level data miners will have less copies of perturbed data. But this MLT poses challenges based on the perturbation. Because of the different copies of perturbed data available, the data miner can utilize the diversity of the copies and can generate more accurate data as the original data. This is called the Diversity attack. In this paper this challenge in MLT – PPDM. The different copies of the perturbed data is generated by the Random Rotation process which varies in mean and co-variance. Based on these, the data owners will release only the mean and co-variance of the perturbed data for each trust levels to handle the Diversity attack. The rotation perturbation helps to reveal the statistical data instead of the original data.

Many Privacy preserving data mining techniques has been proposed to preserve the sensitive information in the original dataset. But these techniques hold only to the static data and not for the data streams. This data stream has many dimensions which differ from the static data. They differ with time, with speed of data, amount of data, need of immediate response and many more. And mainly the accuracy of the data decreases if the transformation is performed on the data streams. This paper[7] has performed the random rotation of the data followed by the clustering process to preserve the sensitive data. This paper has also proved for the accuracy of the data after perturbation process.

Projection perturbation

Random projection refers to the technique of projecting a set of data points from a high dimensional space to a randomly chosen lower dimensional space. Projecting the data from higher dimensional space to a lower dimensional random space, will modify the data from original form to perturbed form by preserving much of its distance-related

characteristics. This research paper[4] presents

experimental results on the accuracy and privacy of the random projection-based data perturbation technique.

The following steps reduce the dimensionality of the data by random projections:

Consider the data set X={x1, …. xn} where each data point is a p dimensional vector such that xi є Rp and the following steps reduces the data onto a q dimensional space such that 1 ≤ q < p.

1) Arrange the data into a p × n matrix where p is the

dimensionality of the data and n is the number of data points.

2) Generate a q × p random projection matrix R*

[MATLAB randn (q, p)].

3) Multiply the random projection matrix with the

original data.

X * q × n = R* q × p * Xp × n

Thus the data is projected to the lower random projection space.

This technique allows many algorithms to be applied directly on the perturbed data and perform analysis. The Euclidean distance is preserved in the perturbed data when this random projection technique is used.

III. REVIEW AND COMPARISON

This section discusses the various data perturbation techniques used for data publishing in the cloud computing.

Yang Pan et al. [15] have proposed the data perturbation

method for preserving the high sensitive data in the cloud database. The proposed noise generator introduces the noise content to the original database thus hides the original data. The noise is added to the database with the perturbation algorithm. The mean and the covariance of the original data are preserved during the perturbation process. The retrieval algorithm retrieves the original data content from the database. The proposed model uses the entropy measure for noise addition, and hence it is robust against noise. But, the data retrieval fails when the order of perturbed data changes.

Yuan-Hung Kao et al. [18] have proposed the reversible privacy contrast mapping (RPCM) algorithm based on the data perturbation method. This method performs the data hiding through the reversible technique as used in the image processing. The proposed RPCM enhances the privacy disclosure risk, and the original data recovery efficiency gets reduced if seed data is unavailable.

Keke Chen and Ling Liu [19] have proposed Geometric Data Perturbation (GDP) for the data publishing in the cloud database. The multi-column privacy evaluation framework evaluates the proposed GDP model. The experimental results show that the GDP model has higher privacy guarantee. It suffers from a disadvantage when the order of original data is altered.

International Journal of Emerging Technology and Advanced Engineering

200

Huiqi Xu et al. [20] have proposed RASP data perturbation method for maintaining the data records in the cloud computing. The proposed model performs the data perturbation based on the range and the KNN query. The KNN-R algorithm processes the KNN queries. This proposed model has improved the security for data publishing. The record preservation for the data retrieval forms the data cloud has poor quality.

Wei Song et al. [16] have proposed Bloom filter based tree index for the full-text retrieval from the cloud

database. The data perturbation depends on the query and

the encrypted documents in the cloud data. This proposed technique has a good time efficiency and low storage requirement. The data decryption during the data retrieval process increases the computational overhead.

Huang Qin-long et al. [21] have proposed the digital rights management (DRM) scheme for the data encryption. This method ensures the data perturbation be effective since the encrypted contents from the process can be retrieved only through the key issued by the server. The license server issues the key to the user on demand and thus prevents the data hacking from the cloud database. The proposed model has low computational complexity, high efficiency, and security. The disadvantage of this model is that the license acquisition from the user increases the time duration to the data access.

Ming-Quan Hong et al. [17] have proposed the homomorphic encryption scheme based on the ECC for the data perturbation. This model protects the data from the user through the homomorphic encryption scheme. This model overcomes the limitations in the multiparty computations. The proposed model is highly feasible and has high security. But, the model does not support the floating point calculation.

IV. CHALLENGES

The major challenges involved in the publishing the data in the cloud computing are enlisted as follows:

 The cloud computing has a very large database and

hence the trust for cloud computing can arise only if the data stored in the cloud database are secured [17].

 The high sensitive and the private information

contents in the data can be preserved through the various encryption algorithms. At the same time, the data retrieval from the database through the

decryption process should not degrade the

information content in the database [16].

 The cloud computing should allow the maximum

utility to the users. The raw data content should be disclosed to the user by the cloud user only on the request of the data owner [15].

 The data should be available for the user all the time

in the cloud database.

 The encryption process in the data perturbation should

not alter the original data content.

V. CONCLUSION

Data publishing is gaining more value nowadays; in turn the importance of the Data Privacy has also improved to a greater extent which is the major part in data publishing. In this paper, we have discussed in detail about the various Data Perturbation techniques and have discussed about their merits. We have detailed about the random perturbation technique which has improved the efficiency of the perturbation techniques [4] [5] [1]. From the analysis of all the perturbation techniques, it is considered that the geometric perturbation provides a higher level of privacy and utility of the data to the users. We have also identified the challenges involved in publishing the data in cloud environment. Many existing PPDM techniques have been discussed in this paper. From the survey on existing literature about privacy preserving it can be concluded that there is no single technique which is consistent in all domains. Researchers are involved in extensive research to conceal the security of the data without compromising the utility of the data.

REFERENCES

[1] Agrawal, C. and Yu, P.S., “General Survey of Privacy Preserving Data Mining Models and Algorithms”, Privacy-Preserving Data Mining Advances in Database Systems Volume 34, pp 11-52,2008. [2] Fung, B.C.M., Wang, K., Chen, R., and Yu, P.S.,

“Privacy-Preserving Data Publishing: A Survey of Recent Developments,” ACM Computing Surveys”, Vol. 42, No. 4, pp. 1-53, 2010. [3] Keke Chen, Ling Liu, “Privacy Preserving Data Classification with

Rotation Perturbation”, Fifth IEEE International Conference on Data Mining, 2005.

[4] Keke Chen, Ling Liu, “Geometric Data Perturbation For Privacy Preserving Outsourced Data Mining”, Springer Knowl Inf Sys, 2010.

[5] Jie Liu, Yifeng XU, “Privacy Preserving Clustering by Random Response Method of Geometric Transformation”, Fourth International Conference on Internet Computing for Science and Engineering (ICICSE), pp. 181-188, 2009.

International Journal of Emerging Technology and Advanced Engineering

201

[7] Swapnil Kadam, Prof. Navnath Pokale, “Privacy Preserving through Data Perturbation using Random Rotation Based Technique in Data Mining “,International Journal of Advanced Research in Computer Engineering & Technology (IJARCET) Volume 5 Issue 1, January 2016

[8] Stanley, R. M., Oliveira and Osmar R. Za¨iane, “Privacy Preserving Clustering by Data Transformation”, Journal of Information and Data Management”, Vol. 1, No. 1, 2010.

[9] Chhinkaniwala H. and Garg S., “Privacy Preserving Data Mining Techniques: Challenges and Issues”, CSIT, 2011.

[10] Nimpal Patel and Shreya Patel,”A Study on Data Perturbation Techniques in Privacy Preserving Data Mining”, International Research Journal of Engineering and Technology , Volume 02, Issue 09, Dec 2015.

[11] H. Kargupta, S. Datta, Q. Wang, and K. Sivakumar, “On the privacy preserving properties of random data perturbation techniques,” in Proceedings of the IEEE International Conference on Data Mining, Melbourne, FL, November 2003.

[12] Z. Huang, W. Du, and B. Chen, “Deriving private information from randomized data,” in Proceedings of the 2005 ACM SIGMOD Conference, Baltimroe, MD, June 2005, pp. 37–48.

[13] S. Guo and X. Wu, “On the use of spectral filtering for privacy preserving data mining,” in Proceedings of the 21st ACM Symposium on Applied Computing, Dijon, France, April 2006, pp. 622–626.

[14] Majid,M.Asger,Rashid Ali, “Privacy preserving Data Mining Techniques:Current Scenario and Future Prospects”,IEEE 2012.

[15] Yang Pan, Gui Xiaolin, An Jian, Yao Jing, Lin Jiancai And Tian Feng, "A Retrievable Data Perturbation Method Used in Privacy-Preserving in Cloud Computing", China communications, vol. 11, no. 8, pp. 73-84, 2014.

[16] Wei Song, Bing Wang, Qian Wang, Zhiyong Peng, Wenjing Lou and Yihui Cui, "A privacy-preserved full-text retrieval algorithm over encrypted data for cloud storage applications", Journal of Parallel and Distributed Computing, vol. 99, pp. 14-27, 2016. [17] Ming-Quan Hong, Wen-bo Zhao, and Peng-yu Wang,

"Homomorphic Encryption Scheme Based on Elliptic Curve Cryptography for Privacy Protection of Cloud Computing", In proceedings of IEEE International Conference on Intelligent Data and Security (IDS), pp. 152-157, 2016.

[18] Yuan-Hung Kao, Wei-Bin Lee, Tien-Yu Hsu and Chen-Yi Lin, "Data Perturbation Method Based on Contrast Mapping for Reversible Privacy-preserving Data Mining", Journal of Medical and Biological Engineering, vol. 35, no. 6, pp. 789-794, 2015.

[19] Keke Chen and Ling Liu, "Geometric data perturbation for privacy preserving outsourced data mining", Knowledge and Information Systems, vol. 29, no. 3, pp. 657-695. 2011.

[20] Huiqi Xu, Shumin Guo and Keke Chen, "Building Confidential and Efficient Query Services in the Cloud with RASP Data Perturbation", IEEE Transactions on Knowledge and Data Engineering, vol. 26, no. 2, pp. 322-335, 2014.