Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system/method for enhancing the privacy preserving of data using data perturbation technique, said system/method comprising the steps of:
a) The system has GNDP_Categorical method (1) takes input as data set and generates a perturbed data (8) as an output.
b) The second technique is GDP_RS (2) also takes data set as input and perturbed data as an output.
c) The third technique OABE (4), consists of three sub process to enhance the security of the data by using clustering phase (5), map-reduce functionality (6) and privacy validation (7).
2. As mentioned in claim 1, the data’s are browse and given input to the GNDP_Categorical method and also GDP_RS method and generate secured data.
3. As per claim 1, the OABE techniques processed in three steps like clustering the data based on different criteria, then eliminate the some redundant data using map-reduce function after that, privacy of the data is validated using the OABE method. , Description:Field of Invention
The present invention relates to, controlling information leaking in the field of data mining. By choosing proper Privacy preserving techniques, the confidential information is secured while publishing the private information.
Background of the Invention
To construct privacy data protection algorithms with low data mining outcomes, The patent (US2015/0242648A1) Providing data negotiation control allows a person or entity to negotiate the use of data collected beyond the needs of a third party transaction. The patent (US2017/9626528B2) focuses on how to classify documents depending on enforcement tactics. The patent (US2015/0254469A1) describes automatically learning and adjusting to classify protected data. Various systems, methods, and/or algorithms can learn and adapt to perform classification of protected data.
The patent (WO2017/214608A1), elaborated integrating a privacy management system with DLP tools using DLP tools to identify sensitive information stored in computer memory outside the context of the privacy management system displaying each area of sensitive data to a privacy officer (e.g., similar to pending transactions in a checking account that have not been reconciled) and designated privacy officer to either reconcile an entry with an existing data flow or campaign in the privacy management system, or initiate a new privacy assessment on the data to capture the appropriate privacy attributes and data flow information.
Personal information is protected by divorcing it from user identification (US2010/0199098A1). In some embodiments, each user is assigned an anonymous token that is not linked to their identity. Information (e.g., a user's physical or geographic location) is stored with this anonymous token, but not the user. Those with access to personal data, including the owner, can use a variety of methods to link the anonymous token to the owner. Locating the data associated with the anonymous token in the data repository retrieves the personal information.
Techniques for safeguarding the privacy and security of web document data are presented (US2017/0012980A1). A web browser is set up to change a preview web page document's URL (which contains an access token) before loading external resources (e.g., web page content). For example, the browser may replace a current page URL with a sacrificial URL that does not contain the token. So the browser won't expose your access token to resources while the web page content is loaded, and won't communicate your original URL as a referrer to resources that provide web page content. After loading the web page content, the browser's current page URL is changed back to the original URL.
The objective of this invention is used to safeguards the personal data using Gaussian Noise based Data Perturbation for Categorical (GNDP C). The GNDP C technique retains individual sensitive information by adding some noise (Gaussian Noise) to the original data.Geometric Data Perturbation with Random Swapping (GDP RS) protects sensitive data through random swapping. This GDP RS technique improves accuracy and privacy for both categorical and numerical datasets.

Summary of the Invention
In light of the above mentioned drawbacks in the prior art, the present invention aims to protect sensitive personal data through the use of data mining techniques that don't compromise privacy.
According to the findings of this invention, different privacy-preserving framework models have important features that must be preserved in order to protect sensitive information. Recent privacy concerns and a need for better data mining techniques have been cited as major obstacles.
Brief Description of Drawings
The invention will be described in detail with reference to the exemplary embodiments shown in the figures wherein:
Figure 1 Proposed method for securing privacy in data mining.
Detailed Description of the Invention
In privacy-preserving data posting, data is changed into some form before being published in such a way that untrusted viewers cannot easily identify a person. Previously, the researchers concentrated on a variety of data categorization techniques with varying degrees of generalization. High, medium, and low level generalization parameters have been tested and reviewed to ensure that various forms of data can be visible while remaining private. The PPDP's appropriateness can be maintained if the system allows us to evaluate the data's dependability and validity, as well as the user, prior to data transmission. Such work, on the other hand, falls within the umbrella of empirical research, as new parameters are required to test the system. Loss, entropy, and other characteristics may not be sufficient to determine the legitimacy of the system.
With rapid advancement in information and communication technology over the last decade, we've now reached the big data stage. A lot of data is generated from various sources (e-commerce sites, social media platforms, and medical records), making it possible to use and uncover new information from the vast amounts of data available. The ability to extract relevant patterns comes with the responsibility of keeping information private and secure at the same time. As a result, the offering privacy of a data using data mining is a difficult task. Military, financial, and banking applications, as well as security applications, all require the preservation of privacy.
Data modification, on the other hand, is the process of altering the database's real values. The sensitive features are preserved in such updated data, making it valuable for data analysis. To safeguard the organization's privacy policy, data modification is a necessary internal activity. Individual users may not provide adequate or complete data due to privacy concerns. This leads to flawed analysis. Data mining requires precise input to analyse data. The privacy of sensitive user data must be protected. This issue inspires a new research era known as Privacy-Preserving Data Mining (PPDM).
The goal of privacy-preserving data mining is to provide huge aggregate results while protecting personal sensitive information. Data perturbation, randomization, and anonymization are popular and well-researched strategies for maintaining an individual's sensitive information. In this context, three approaches are used to construct a novel privacy-preserving data mining architecture.
The Gaussian Noise based Data Perturbation for Categorical (GNDP C) method safeguards personal data. The GNDP C technique retains individual sensitive information by adding some noise (Gaussian Noise) to the original data. The proposed GNDP_Categorical method achieved the accuracy value of 83.26% for NB and 85.80% for J48. Geometric Data Perturbation with Random Swapping (GDP RS) protects sensitive data through random swapping. This GDP RS technique improves accuracy and privacy for both categorical and numerical datasets. The proposed GNDP_Categorical method achieved the accuracy value of 83.26% for NB and 85.80% for J48. Finally, an OABE technique for massive data privacy preservation is defined. Initially, the PFCM (Probabilistic Fuzzy C-Means) method grouped the input data. The clustered data is then supplied into the map-reduce framework. In the proposed OABE method, riders are optimized for privacy validation. The proposed OABE technique attains more security than the existing GNDP_Categorical and the GDP_RS methods. The proposed OABE scheme achieved the accuracy of 90.42%.
3 Claims & 1 Figure