Privacy Preserving Data Mining (PPDM)

In Privacy preserving has originated as an important concern with reference to the success of the data mining. Privacy preserving data mining (PPDM) deals with protecting the privacy of individual data or sensitive knowledge without sacrificing the utility of the data. People have become well aware of the privacy intrusions on their personal data and are very reluctant to share their sensitive information. This may lead to the inadvertent results of the data mining. Within the constraints of privacy, several methods have been proposed but still this branch of research is in its infancy.

A recent survey on the techniques used for privacy-preserving data mining may be found in Xueyun (2014) which reviews main PPDM techniques based on a PPDM framework and compare the advantages and disadvantages of different PPDM technique. It also discusses the open issues and future research (2012), describes the current scenario preserving data mining and propose some future Shweta (2014), all Cryptography and Random Data Perturbation methods techniques of PPDM is illustrates the application of certain techniques for preserving privacy on experimental dataset, and reveals their effects on the results. Jian (2009) intends to preserving data mining technologies clearly and then proceeds to analyze the merits and shortcomings of these technologies. Methods such as k-

The data is an essential element in any business domain, discloser or leakage of sensitive and private data to others can create a number of social and financial issues. In this context the data mining is shifting towards the privacy preserving data mining. In different real world conditions the end data owner submitted their private and confidential data to a business domain. But due to cretin requirements of business intelligence and marketing research the data discloser is required. In these conditions the discloser or leakage of the actual data owner’s data can create various issues. In this context the proposed work is intended to work with the privacy preserving data mining environment to preserve the data privacy. The proposed data model consists of the three main contributions first designing the noise based data transformation approach. Secondly the data model help to prevent the data discloser to another party. Third the technique by which the data publishing and it’s utility in other public domain becomes feasible. Therefore the proposed work introduces a lightweight privacy preserving data model that combines data from different data sources. Include the regulated noise to entire dataset to modify the values. Process the data using data mining model for finding combined data based decisions and help to publish the data for other marketing and research purposed without disclosing the actual data values. The implementation of the proposed technique is given using JAVA technology and their performance is measured. The obtained results demonstrate the proposed work is helpful for the PPDM based data processing and publishing.

Data mining aims to take out useful information from multiple sources, whereas privacy preservation in data mining aims to preserve these data against disclosure or loss. Privacy preserving data mining (PPDM) [1,2] is a new research direction in data mining and statistical databases [3], where data mining algorithms are analyzed for the side effects they acquire in data privacy. The main consideration of the privacy preserving data mining is twofold. First, sensitive raw data like identifiers, name, addresses and the like should be modified out from the original database, in order for the recipient of the data not to be able to cooperation another person’s privacy. Second, sensitive knowledge which can be mined from a database by using data mining algorithms should also be excluded, because such knowledge can equally well cooperation data privacy. The main purpose of privacy preserving data mining is to develop algorithms for modifying the original data in some way, so that the private data and the private knowledge stay private even after the mining process. In this paper, we provide a classification and description of the various techniques and methodologies. Agarwal and Srikant [3] and Lindell and Pinkas [4] introduced the first Privacy -preserving data mining algorithms which allow parties to collaborate in the extraction of knowledge, without any party having to reveal individual items or data. The goal of this paper is to give a review of the different dimension and classification of privacy preservation techniques used in privacy preserving data mining. Also aim is to give different data mining algorithms used in PPDM and related research in this field.

Data publishing is the process of making data, collected by various institutions such as hospitals, financial institutions, government etc. public, so that it can be used to find useful patterns for the purpose of data mining. But the growing concern with this is the violation of the individual’s privacy. When seen in isolation, the individual’s identity may not be found out. But this can be linked with other publicly available documents like voter list and the sensitive information of an individual may be compromised. One approach to overcome this is to apply transformation techniques to the sensitive attributes in a manner that the data published will be useful for analysis while still protecting the privacy. A term closely related to this is privacy preserving data mining (PPDM) that refers to development of data mining techniques that have privacy concerns incorporated. These involve data hiding, rule hiding etc. Data perturbation is a specific class of data hiding techniques.

Abstract:Due to the increase in sharing sensitive data through networks among businesses, governments and other parties, privacy preserving has become an important issue in data mining and knowledge discovery. Privacy concerns may prevent the parties from directly sharing the data and some types of information about the data. This paper proposes a solution for privately computing data mining classification algorithm for horizontally partitioned data without disclosing any information about the sources or the data. The proposed method (PPDM) combines the advantages of RSA public key cryptosystem and homomorphic encryption scheme. Experimental results show that the PPDM method is robust in terms of privacy, accuracy, and efficiency. Data mining has been a popular research area for more than a decade due to its vast spectrum of applications. However, the popularity and wide availability of data mining tools also raised concerns about the privacy of individuals. The aim of privacy preserving data mining researchers is to develop data mining techniques that could be applied on databases without violating the privacy of individuals. Privacy preserving techniques for various data mining models have been proposed, initially for classification on centralized data then for association rules in distributed environments.

methods might result in information loss and side-effects in some extent, such as data utility-reduced, data mining efficiency- downgraded, etc. That is, an essential problem under the context is trade-off between the data utility and the disclosure risk. This paper provides an analysis of the Euclidean distance preserving methods for privacy preserving data mining, and points out their merits and demerits.

Data mining is an approach for mining the centralized database for extracting the valuable patterns from data. This process needs to implement the different computational algorithms for finding the required patterns. But sometimes when we work with the real world datasets the privacy and sensitivity of data are needed to be maintained [1]. The branch of data mining which preserves privacy during the mining of data is known as privacy- preserving data mining techniques [2]. In this work, the main area of study is privacy-preserving data mining. In this context, the multiparty data association and association rule mining technique is

Nowadays there is huge amount of data being collected and stored in databases everywhere across the globe. Recently, there has been a growing interest in the data mining area, where the objective is the discovery of knowledge that is correct and of high benefit for users. Data miming consists of a set of techniques that can be used to extract relevant and interesting knowledge from data. Data mining has several tasks such as association rule mining, classification and prediction, and clustering. Classification techniques are supervised learning techniques that classify data item into predefined class label. It is one of the most useful techniques in data mining to build classification models from an input data set. The used classification techniques commonly build models that are used to predict future data trends. In this research paper we analysis CAMDP (Combination of Additive and Multiplicative Data Perturbation) technique for KNN classification as a tool for privacy-preserving data mining. We can show that KNN Classification algorithm can be efficiently applied to the transformed data and produce exactly the same results as if applied to the original data.

Ying-hua et  al. (2011) surveyed the Distributed Privacy Preserving Data Mining (DPPDM) depending on different underlying technologies. Existing techniques are categorized into three groups such as (1) secure multi-party computation, (2) pertur- bation and (3) restricted query. Li (2013) elucidated the advantages and drawbacks of each method by developing and analyzing a symmetric-key based privacy-preserving scheme to support mining counts. An incentive consideration is proposed to the study the secure computation by presenting a reputation system in wireless network. The pro- posed system offered an incentive for misbehaving nodes to behave properly. Experi- mental results revealed the system effectiveness in detecting the misbehaving nodes and enhancing the average throughput in the whole network. Furthermore, Dev et al. 2012) acknowledged the privacy risks related to data mining on cloud system and presented a distributed framework to remove such risks. The proposed approach involved classifica- tion, disintegration, and distribution. This avoided the data mining by preserving the pri- vacy levels, splitting the data into chunks and storing them into suitable cloud providers. Though, the proposed system offered a suitable way to safe privacy from mining based attacks, but it added a performance overhead as client accessed the data frequently. For instance, client had to run a global data analysis for a complete dataset, where the analy- sis required accessing the data through different locations with a degraded performance. Tassa (2014) developed a protocol for secured mining of association rules in horizon- tally distributed database. The proposed protocol possessed advantages over leading protocols in terms of performance and security. It included two set of rules including (1) a multi-party protocol to compute the union or intersection of private subsets pos- sessed by each client and (2) a protocol to test the presence of an element held

Providing security to delicate data against unapproved access has been a long term goal for the database security research group and for the administration statistical organizations. Subsequently, the security issue has gotten to be, recently, a significantly more critical territory of examination in data mining. Consequently, as of late, privacy- preserving data mining has been mulled over extensively. Data anonymization, one of the methods of privacy preserving data mining, is a sort of data purification whose expectation is security insurance. It is the methodology of either scrambling or expelling by and by identifiable data from information sets, with the goal that the individuals whom the information portrays stay unacknowledged. In this paper we have reviewed various techniques of data anonymization and have shown the comparative analysis of the same.

have met a serious challenge due to the increased concerning and worries of the privacy, that is, protecting the privacy of the critical and sensitive data. Data perturbation is a popular technique for privacy preserving data mining. The major challenge of data perturbation is balancing privacy protection and data quality, which normally considered as a pair of contradictive factors. Geometric data perturbation technique is a combination of Rotation, Translation and Noise addition perturbation technique. It is especially useful for data owners to publish data while preserving privacy –sensitive information. Typical examples include publishing micro data for research purpose or outsourcing the data to the third party that provides data mining services. In this paper we try to explore the latest trends in Geometric data perturbation technique.

noise. In the statistics community, this approach was primarily used to provide summary statistical information (e.g., sum, mean, variance, etc.) without disclosing individual’s confidential data. In the privacy preserving data mining area, this approach was considered [2,3] in for building decision tree classifiers from private data. Recently, many researchers have pointed out that additive noise can be easily filtered out in many cases that may lead to compromising the privacy [4,5]. Given the large body of existing signal-processing literature on filtered random additive noise, the utility of random additive noise for privacy-preserving data mining is not quite clear.

We have reviewed the multiplicative perturbation method as an alternative method to privacy preserving data mining. The design of this category of perturbation algorithms is based on an important principle: by developing perturbation algorithms that can always preserve the mining task and model specific data utility, one can focus on finding a perturbation that can provide higher level of privacy guarantee. We described three representative multiplicative perturbation methods − rotation perturbation, projection perturbation, and geometric perturbation. All aim at preserving the distance relationship in the original data, thus achieving good data utility for a set of classification and clustering models. Another important advantage of using these multiplicative perturbation methods is the fact that we are not required to re-design the existing data mining algorithms in order to perform data mining over the perturbed data. One observation is that both the above mentioned techniques require much more samples (or background knowledge) to work effectively in the high dimensional case. Thus, random projection techniques should generally be used for the case of high dimensional data, and only a smaller number of projections should be retained in order to preserve privacy. Thus, as with the additive perturbation technique, the multiplicative technique is not completely secure from attacks. A key research direction is to use a combination of additive and multiplicative perturbation techniques in order to construct more robust privacy preservation.

IV. CONCLUSTION AND FUTURE SCOPE The amount of digital Information generated by fast growing technology and science concerned in storing, retrying, sharing resources availability of data processing and securing the data increasing everywhere, however it is in the dark concerning how data is shared and utilized. Privacy preserving data mining initiated new direction and got serious attention by researchers. we try to solve the problem to some extent by proposed novel methods. In this paper we present and evaluate comparative based analytical results of privacy preserving data mining algorithms empirically that extent to protects the sensitive data. we observe the computation time, communication cost to share the sensitive data among parties will be reduced and provides trade of between data privacy, utility.

collector gets data from data providers so as to support the subsequent data mining operations. In order to invent useful knowledge which is expected by the decision maker, the data miner applies data mining algorithms to data obtained from data collector. A decision maker can get the data mining results directly from the data miner, or from some Information Transmitter. The focus of this topic is to achieve privacy at data miner level. Data miner will get data to mine from data collector to be able to in not unique layout and by means of applying one-of-a-kind data mining strategies; data miner can discover sensitive information. So venture of data miner is to hold the privacy of received result and pass the consequences to decision maker that doesn't bring about any security breach. Several studies on PPDM have been conducted [5] [6]. But none of the modern-day proposals provide privacy to unwanted disclosure of sensitive information. The paper presents a system architecture that provides privacy by the use data mining algorithms without affecting the security of sensitive information contained in the data.

In different business or company homes and government bodies possess sure framework or infrastructures for maintaining brooding again knowledge collections for watching and processing it. The assorted knowledge extracted from its native or confined databases aren’t comfortable to full pity given accomplishing the specified results. Therefore, such shortcomings do require a system or platform that could effectively collect the massive distributed knowledge and may perform the data mining to want information that would be analyzed expeditiously and exactly provide desire result, Such a situations place forth the requirement for privacy preservation in data processing (PPDM) systems. The main objective of the PPDM systems is to keep maintains the integrity of the data revealed and to realize economic data processing results. The privacy preservation in knowledge mining systems that presently exist secure the info exploitation varied mechanisms. Cryptography is the most widely used mechanism to increases data integrity as well as security[10]. Cryptography techniques in which the key generation and key exchange functions are performed. But now a days cryptography technique becomes problematic because as a result of this method has some disadvantages. To beat this disadvantages and enhancing the adversaries ineffective in this paper we tend to introduce a Key Distribution-Less Privacy Preserving Data Mining system (KDLPPDM). The accepting of the optimal data mining technique is also critical and must facilitate accurate analysis with the use of limited data which not provide exact solution and results of information.

Sailaja. R. J. L, P. Dayaker [11] “Preventing Diversity Attacks in Privacy Preserving Data Mining” in September 2013 implement multilevel trust based PPDM which enables data owners to have freedom to choose the level of privacy needed. Based on this trust level perturbations are made. We built a prototype application that demonstrates the proof of concept. The existing perturbation based PPDM models assume single level trust on data miners. It focus on multilevel trust based PPDM which provides more flexibility to data owner in choosing the level of privacy to data. A challenge in doing so is that malicious data miners can use multiple copies of perturbed data in order to establish the original data. This kind of attack is prevented by using noise correlation matrix across the copies to deny the attackers not to have diversity option.

While k-anonymity protects against identity disclosure, it does not provide sufficient protection against attribute disclosure. There are two attacks: the homogeneity attack and the background knowledge attack. Because the limitations of the k-anonymity model stem from the two assumptions. First, it may be very hard for the owner of a database to determine which of the attributes are or are not available in external tables. The second limitation is that the k-anonymity model assumes a certain method of attack, while in real scenarios there is no reason why the attacker should not try other methods. Example1. Table4 is the Original data table, and Table5 is an anonymous version of it satisfying 2-anonymity.The Disease attribute is sensitive. Suppose Manu knows that Ranu is a 34 years old woman living in ZIP 93434 and Ranu's record is in the table. From Table5, Manu can conclude that Ranu corresponds to the first equivalence class, and thus must have fever. This is the homogeneity attack. For an example of the background knowledge attack, suppose that, by knowing Sonu's age and zip code, Manu can conclude that Sonu's corresponds to a record in the last equivalence class in Table5. Furthermore, suppose that Manu knows that Sonu has very low risk for cough. This background knowledge enables Manu to conclude that Sonu most likely has fever.

Privacy protection is a basic right, stated of the universal declaration of human rights. it is also an important concern in today's digital world. data security and privacy are two concepts that are often used in conjunction; however, they represent two different facets of data protection and various techniques have been developed for them privacy is not just a goal or service like security, but it is the people's expectation to reach a protected and controllable situation, possibly without having to actively look for it by themselves. therefore, privacy is defined as "the rights of individuals to determine for themselves when, how, and what information about them is used for different purposes" in information technology, the protection of sensitive data is a crucial issue, which has attracted many researchers. in knowledge discovery, efforts at guaranteeing privacy when mining and sharing personal data have led to developing privacy preserving data mining (ppdm) techniques. ppdm have become increasingly popular because they allow publishing and sharing sensitive data for secondary analysis. different ppdm methods and models (measures) have been proposed to trade of the utility of the resulting data/models for protecting individual privacy against different kinds of privacy attacks.