Privacy Preserving Data Mining Research Articles

DATA PROCESSING THROUGH AN ADDITIVE ROTATIONAL PERTURBATION TECHNIQUE IN A SECURED ENVIRONMENT OF PPRIVACY

As the usage of internet and web applications emerges faster, security and privacy of the data is the most challenging issue which we are facing, leading to the possibility of being easily damaged. Various conventional techniques are used for privacy preservation like condensation, randomization and tree structure etc., the limitations of the existing approaches are, they are not able to maintain proper balance between the data utility and privacy and it may have the problem with privacy violations. This paper presents an Additive Rotation Perturbation approach for Privacy Preserving Data Mining (PPDM). In this proposed work, various dataset from UCI Machine Learning Repository was collected and it is protected with a New Additive Rotational Perturbation Technique under Privacy Preserving Data Mining. Experimental result shows that the proposed algorithm’s strength is high for all the datasets and it is estimated using the DoV (Difference of Variance) method.

WITHDRAWN: Security protocol of sensitive high utility itemset hiding in shared IoT environments

Abstract As Internet Technology (IT) rapidly grows, it is necessary to protect private and confidential information of collected data from both smart and Internet of Things (IoT) devices. It becomes a necessity to protect and hide confidential or private information for data sharing or publishing between collaborators and partners. The concept of Privacy-Preserving Data Mining, known as PPDM, is created to clean the database for hiding sensitive information. However, it applies only to binary databases. To handle Privacy-Preserving Utility Mining, known as PPUM, we present a Maximum Sensitive Utility-Maximum Sensitive conflIct algorithm, known as MSU-MSI, to find conflicting items within sensitive high-utility itemsets for sanitization of data. Transactions with sensitive high-utility itemsets are first projected for the sanitization process. Next, the number of conflicting items within the itemsets is calculated, and the designed progress is used to reduce the quantity value of processed items for sanitization. In our experimental results, the designed algorithm is compared with the standard algorithms like HHUIF, MSICF, and the state-of-the-art MSU-MAX and MSU-MIN approaches. Finally, the proposed MSU-MSI is shown to achieve better performance in terms of missing cost, especially in extreme datasets (very dense or very sparse), a true testament to the process. Furthermore, the designed MSU-MSI achieved an excellent performance in terms of database structure similarity and database utility similarity with a very slight difference compared to the previously achieved top results.

RETRACTED ARTICLE: An improved privacy-preserving data mining technique using singular value decomposition with three-dimensional rotation data perturbation

Recent advancements in data mining have given rise to a new channel of research, coined as privacy-preserving data mining (PPDM). PPDM technology allows us to derive useful information from vast amounts of data while protecting privacy of individual records. This paper proposed a methodology based on the machine learning algorithm called singular value decomposition (SVD) and 3D rotation data perturbation (RDP) for preserving privacy of data. Decomposition and dimensionality reduction helps to eliminate sensitive information, and perturbed matrix is generated. The original and perturbed data are classified using different classifiers, and the performance is measured in terms of accuracy rate. Accuracy is the degree of correlation between the absolute observation and the actual observations. Experimental results revealed that the proposed scheme outperforms by achieving excellent accuracy for matrices of different sizes.

A Machine Learning Model for Data Sanitization

Discovering important knowledge that may be available from databases while preserving the privacy of sensitive information can be considered a hot research subject of data mining in recent times. With the establishment of strong Internet of Things (IoT) networks globally, several data-intensive applications will be developed. Privacy of information over the network is increasingly relevant, and as edge computing has grown more important, applications running over networks require protection. The privacy of information while utilizing data is a trade-off that needs to be addressed. In the past, many heuristics and meta heuristics-based approaches were revealed to sensitize sensitive information in privacy-preserving data mining (PPDM). They perturb the original database to hide sensitive information using addition or deletion operations. This is a known NP-hard problem. In this paper, we propose data privacy of IoT connected devices over heterogeneous networks. A deep re-enforcement learning-based technique is applied to sensitize sensitive information from a given database while keeping the balance between privacy protection and knowledge discovery during the sanitization process. Furthermore, minimizing known side effects that can be caused in the sanitization process is also be considered. Substantial experiments are conducted on both synthetic and real-world datasets. Results are evaluated based on sanitization side effects which include failing to hide sensitive items as well as choosing not to hide sensitive items. The proposed approach shows significant performance improvement compared to meta-heuristics (Genetic Algorithm, Particle Swarm Optimization) and heuristics (Greedy) approaches by our evaluation.

WITHDRAWN: Demonstrating impact of noise over classifier performance in a privacy-preserving data mining application

Abstract The Privacy-Preserving Data Mining (PPDM) is respectively new domain of data mining. In this mining technique, different parties combine their data with unknown parties or data from other parties for finding a standard solution of data intelligence. But the data modeling and their outcomes are varying according to different factors. In this paper, the investigation is made to found the impact of dimensions and noise over classifiers' performance in PPDM systems. In this context, first, a large dimension dataset is captured from KDD CUP for experimentation; additionally, the PCA (Principle component analysis), KPCA (kernel principle analysis), and CRC (correlation coefficient) based dimensionality reduction techniques are used for the study. On the other hand, for demonstrating the impact of noise, the first study of different noise used in the PPDM system, random noise, is used to manipulate data. During this, we found the random noise is not functional for categorical attributes. Therefore an extended controlled noise algorithm is introduced. That algorithm is used to generate a new dataset from original data without disturbing the data utility. To justify this fact, two supervised learning classifiers are implemented, namely C4.5 and CART. Additionally, five publically available datasets are used for experiments. According to the obtained results, we found that the classical random noise greatly influences classifier performance inaccuracy in an uncontrolled manner. On the other hand, controlled noise-based manipulated data is less influential for classifier performance because we found less difference between original dataset classification performance and controlled noise-based manipulated dataset classification. But the controlled noise increases the time and memory usages due to alteration of all the categorical attributes.

Anonymized noise addition in subspaces for privacy preserved data mining in high dimensional continuous data

Data privacy is a major concern in data mining. Privacy-preserving data mining algorithms have been used for preserving privacy in data mining. However, privacy-preserving data mining on high dimensional continuous data leads to high data loss, information loss and identifying clusters are very difficult. In this paper, a novel technique Anonymized Noise Addition in Subspaces (ANAS) is proposed, which reduces data loss, information loss and enhances identification of clusters and privacy. Anonymization using aggregation is performed in dense and non-dense subspaces considering Euclidean distances to reduce data loss and enhance privacy. Random noise within the subspace limits is then applied to anonymized subspaces to enhance identification of clusters and reduce data loss. ANAS is run on benchmark datasets, and results show that ANAS can identify 80% of the original dataset clusters on sparse datasets, whereas the existing techniques do not identify any clusters. ANAS reduces data loss by 50%, information loss by 20% and enhances privacy by 40%.

PPDM-TAN: A Privacy-Preserving Multi-Party Classifier

Distributed medical, financial, or social databases are analyzed daily for the discovery of patterns and useful information. Privacy concerns have emerged as some database segments contain sensitive data. Data mining techniques are used to parse, process, and manage enormous amounts of data while ensuring the preservation of private information. Cryptography, as shown by previous research, is the most accurate approach to acquiring knowledge while maintaining privacy. In this paper, we present an extension of a privacy-preserving data mining algorithm, thoroughly designed and developed for both horizontally and vertically partitioned databases, which contain either nominal or numeric attribute values. The proposed algorithm exploits the multi-candidate election schema to construct a privacy-preserving tree-augmented naive Bayesian classifier, a more robust variation of the classical naive Bayes classifier. The exploitation of the Paillier cryptosystem and the distinctive homomorphic primitive shows in the security analysis that privacy is ensured and the proposed algorithm provides strong defences against common attacks. Experiments deriving the benefits of real world databases demonstrate the preservation of private data while mining processes occur and the efficient handling of both database partition types.

Rings for Privacy: An Architecture for Large Scale Privacy-Preserving Data Mining

This article proposes a new architecture for privacy-preserving data mining based on Multi Party Computation (MPC) and secure sums. While traditional MPC approaches rely on a small number of aggregation peers replacing a centralized trusted entity, the current study puts forth a distributed solution that involves all data sources in the aggregation process, with the help of a single server for storing intermediate results. A large-scale scenario is examined and the possibility that data become inaccessible during the aggregation process is considered, a possibility that traditional schemes often neglect. Here, it is explicitly examined, as it might be provoked by intermittent network connectivity or sudden user departures. For increasing system reliability, data sources are organized in multiple sets, called rings, which independently work on the aggregation process. Two different protocol schemes are proposed and their failure probability, i.e., the probability that the data mining output cannot guarantee the desired level of accuracy, is analytically modeled. The privacy degree, the communication cost and the computational complexity that the schemes exhibit are also characterized. Finally, the new protocols are applied to some specific use cases, demonstrating their feasibility and attractiveness.

Cluster-based convolution process on big data in privacy preserving data mining

Cluster-based convolution process on big data in privacy preserving data mining

Dynamic Parameters-Based Reversible Data Transform (RDT) Algorithm in Recommendation System

The protection and processing of the sensitive data in recommendation system are the major concern. Existing literature, used homomorphic encryption (HE), Reversible Data Transform (RDT), differential privacy (DP) and many more schemes to protect user information. Existing RDT scheme require prior sharing of the parameters and an alternative mechanism e.g., Shamir Threshold Protocol or Diffie-hellman algorithm are used to protect the sharing parameters. In this paper, we proposed a chaotic based RDT approach for privacy-preserving data mining (PPDM) in recommendation system. Using this approach, RDT parameter values will be generated locally and because of this, prior sharing of the parameter values for the recovery process will not be necessary. This approach can be used as an alternative to the standard-RDT algorithm where bandwidth and memory are considered important factors. Our results on the Iris data set clearly show that the proposed chaotic RDT shows similar results as standard-RDT. Secondly, in this paper, we explore the usage of the RDT algorithm on real app usage records in the mobile app recommendation (MAR) domain. Thirdly, we tested the application of the RDT algorithm for the standard MovieLens dataset to ensure the validity of results because app usage dataset is publicly not available. Our results show that the proposed RDT algorithm can replace HE if an adaptive recommendation approach is used. Similarly, we can safely use the RDT approach to any data including user rating, health data or app usage frequency to ensure user privacy before delivering it to the recommender-server.

Classification of Privacy Preserving Data Mining Algorithms: A Review

Nowadays, data from various sources are gathered and stored in databases. The collection of the data does not give a significant impact unless the database owner conducts certain data analysis such as using data mining techniques to the databases. Presently, the development of data mining techniques and algorithms provides significant benefits for the information extraction process in terms of the quality, accuracy, and precision results. Realizing the fact that performing data mining tasks using some available data mining algorithms may disclose sensitive information of data subject in the databases, an action to protect privacy should be taken into account by the data owner. Therefore, privacy preserving data mining (PPDM) is becoming an emerging field of study in the data mining research group. The main purpose of PPDM is to investigate the side effects of data mining methods that originate from the penetration into the privacy of individuals and organizations. In addition, it guarantees that the data miners cannot reveal any personal sensitive information contained in a database, while at the same time data utility of a sanitized database does not significantly differ from that of the original one. In this paper, we present a wide view of current PPDM techniques by classifying them based on their taxonomy techniques to differentiate the characteristics of each approach. The review of the PPDM methods is described comprehensively to provide a profound understanding of the methods along with advantages, challenges, and future development for researchers and practitioners.

A hybrid approach of homomorphic encryption and differential privacy for privacy preserving classification

Privacy preserving data mining is a substantial research area that aims at protecting the privacy of individuals while enabling to perform data mining techniques. In this study, we propose a secure protocol that fulfils the privacy restriction by combining homomorphic encryption with differential privacy and integrate this protocol into Holte’s One Rule which is a simple, but accurate and efficient classification algorithm. The proposed method allows a researcher to get the answers of his/her queries to build One Rule classifier by processing the encrypted training dataset under Paillier’s cryptosystem and also applies differential privacy to minimize the privacy leakage of individuals as much as possible in this training dataset. Therefore, both of security and privacy of the individuals in the training dataset for classification are provided thanks to our proposed method; since neither the parties, nor the researcher attain any information about the individuals in the database. Besides the One Rule classifier, we apply our proposed privacy preservation model to Naïve Bayes classification algorithm for the performance comparison, and show the efficiency of the proposed method through experiments on real data from UCI repository.

Privacy Preserving Data Mining: Techniques and Algorithms

Privacy Preserving Data Mining: Techniques and Algorithms

Privacy-preserving data mining for open government data from heterogeneous sources

Open data is a global movement with the potential to generate significant social and economic benefits. Policies on open government data (OGD) inspire the development of new and innovative services that government agencies may lack. The International Open Data Charter adequately describes the importance of data mining. Governments that have signed this charter should focus on the following areas—(i) data mining, (ii) linkage, and (iii) in-depth analysis, i.e., distribution of open data that is freely accessible for elaborate analysis using machine reading. However, a series of practical difficulties is observed in connection with the data mining of OGD for in-depth analysis. First, most OGD do not have identifiers to prevent privacy disclosure. Second, owing to the nature of siloed data, the data sharing and collection methods vary with respect to heterogeneous OGD, and administrative or institutional barriers need to be overcome. This has created a demand for a novel technical solution that applies micro-aggregation and distance-based record linkage to address the aforementioned issues. Thus, in this study, a method capable of integrating two or more de-identified OGDs into one dataset to enable OGD data mining is proposed. In addition, the proposed method allows users to adjust the privacy threshold level to determine an appropriate balance between privacy disclosure risk and data utility. The effectiveness of the method is evaluated in terms of several metrics via extensive experimentation. This study emphasizes the importance of the research on efficient utilization of already-published OGDs, which has been relatively neglected in the past. Further, it broadens the research area for privacy-preserving data mining by proposing a method capable of mining heterogeneous data even in the absence of identifiers.

A privacy-preserving protocol for continuous and dynamic data collection in IoT enabled mobile app recommendation system (MARS)

User trust is an important factor in the success of recommendation systems, including Internet of Things (IoT)-based recommendation systems. However, such trust can be eroded in many different ways (e.g., unauthorized data modifications). Several privacy-preservation schemes have been designed for specific data and/or require strict assumptions (e.g., a private/secure communication channel between client-server and third-party authentication). However, these may limit their application in practice. Hence, in this paper we propose the Reversible Data Transform (RDT) algorithm based privacy-preserving data collection protocol. Our protocol allows us to achieve privacy preservation against beyond the scope processing and does not require a private channel or rely on a third-party authentication. Due to group formation, the disclosure probability of the internal disclosure attack will not be greater than 1/k. Similarly, the reversible privacy-preserving data mining approach protects beyond the scope processing. Findings from the experimentation demonstrates the utility of the proposed protocol and its potential to be deployed in a mobile app recommendation system.

Remodeling: improved privacy preserving data mining (PPDM)

The data provided by individuals and various organizations while using internet applications and mobile devices are very useful to generate solutions and create new opportunities. The data which is shared needs to be precise to get the quality results. The data which may contain an individual’s sensitive information cannot be revealed to the world without applying some privacy preserving technique on it. Privacy preserving data mining (PPDM) and Privacy preserving data publishing (PPDP) are some of the techniques which can be utilized to preserve privacy. There are some positives and negatives for every technique. The cons frequently constitute loss of data, reduction in the utility of data, compromised diversity of data, reduced security, etc. In this paper, the authors propose a new technique called Remodeling, which works in conjunction with the k-anonymity and K-means algorithm to ensure minimum data loss, better privacy preservation while maintaining the diversity of data. Network data security is also handled by this proposed model. In this research paper, theoretically, we have shown that the proposed technique addresses all the above-mentioned cons and also discusses the merits and demerits of the same.

Data mining privacy preserving: Research agenda

AbstractIn the modern days, the amount of the data and information is increasing along with their accessibility and availability, due to the Internet and social media. To be able to search this vast data set and to discover unknown useful data patterns and predictions, the data mining method is used. Data mining allows for unrelated data to be connected in a meaningful way, to analyze the data, and to represent the results in the form of useful data patterns and predictions that help and predict future behavior. The process of data mining can potentially violate sensitive and personal data. Individual privacy is under attack if some of the information leaks and reveals the identity of a person whose personal data were used in the data mining process. There are many privacy‐preserving data mining (PPDM) techniques and methods that have a task to preserve the privacy and sensitive data while providing accurate data mining results at the same time. PPDM techniques and methods incorporate different approaches that protect data in the process of data mining. The methodology that was used in this article is the systematic literature review and bibliometric analysis. This article identifieds the current trends, techniques, and methods that are being used in the privacy‐preserving data mining field to make a clear and concise classification of the PPDM methods and techniques with possibly identifying new methods and techniques that were not included in the previous classification, and to emphasize the future research directions.This article is categorized under: Commercial, Legal, and Ethical Issues > Security and Privacy

Improving healthcare services using source anonymous scheme with privacy preserving distributed healthcare data collection and mining

The trends of data mining on healthcare data for improving medical services have increased because of the electronic healthcare record(EHR) system, which collects a massive amount of data on a daily basis. In the current scenario, hospital maintains its EHR system and stores the detailed information of patients. Data mining for healthcare improvement requires the data from all the EHR systems located at a different location to be stored at the central data mining server. Collection of healthcare data at some untrusted central data mining server raises privacy threats. Healthcare data contains patients’ private information and sharing this information for data mining creates privacy issues. Most of the previous research either focused on k-anonymity technique which causes information loss and decreases data mining accuracy or privacy preserving data mining which is focused on only specific data mining technique. We adopt source anonymous technique as privacy preserving scheme and present a novel scheme for healthcare data collection and mining in this paper. Our scheme collects data from all EHR systems without any information loss and stores at a single central data mining server, also ensuring privacy is preserved. Central data mining server helps to analyze the collected data with different data mining techniques (Association rule mining, Classification, Clustering, etc.) without the involvement of EHR systems. Our scheme is collusion resilient against central data mining server and EHR systems. Theoretical and experimental analysis show the efficiency of our scheme in terms of computation and communication cost. The experimental results using Heart disease dataset show the advantage to EHR systems using the proposed approach in terms of disease prediction accuracy.

Method of improving the performance of public-private innovation networks by linking heterogeneous DBs: Prediction using ensemble and PPDM models

Due to the complexity of the innovation process and intensifying competition, small and medium enterprises (SMEs) have increased their participation in public-private innovation networks (PPINs). This study used both inference and prediction models by linking two heterogeneous databases (DBs), consisting of the responses of 1,439 manufacturing SMEs to the Korean Innovation Survey and the financial information of approximately 119,890 companies. In the inference model, we analyzed the determinants that affect the business performance and R&D investment performance of SMEs in PPINs, using generalized linear models. The prediction model utilized a machine learning based ensemble model and the method of linking heterogeneous DBs based on privacy-preserving data mining (PPDM). The findings of this study indicate that while PPINs do not have a significant effect on business performance, they do have a positive correlation to R&D investment. This study also proposes two prediction models for forecasting increases in R&D investment by SMEs, which is considered to be an indicator of PPIN performance. These two models can be respectively used in cases where the features of the companies targeted for prediction can be known in advance and in cases where the features are unknown.

Privacy-Preserving Association Rule Mining Using Homomorphic Encryption in a Multikey Environment

With the development of big data and cloud computing, data analysis technologies play an important role to produce huge market values. Customers with limited computing resources may resort to the cloud to perform some association rule mining tasks. Data owners may have a risk of personal sensitive information leakage in this process. To preserve privacy in outsourced data, data owners may encrypt raw data before uploading. Data analysis of encrypted data is a challenge that has attracted the attention of many researchers in recent years. Homomorphic encryption is a cryptographic tool, which is one of the ways to solve this challenge. It allows data processing of encrypted data without decryption. Researching homomorphic encryption schemes that support privacy-preserving data mining in a multikey environment has become a significant direction. In this article, we propose a novel homomorphic cryptosystem, which supports multiple cloud users to have different public keys. Besides, we propose a privacy-preserving association rule mining scheme on outsourced data uploaded from multiple parties in a twin-cloud architecture. Our scheme uses a transaction record representation method in databases for large shopping malls based on real-world situations, and our experiments on a real transaction database show that our technology is reasonably feasible.