The Normalisation Approach: A Step-by-Step Guide to Protecting Your Data When Mining

To procure valuable insights from existing databases, data mining is becoming increasingly popular. However, given that databases can consist of extensive amounts of data, the conventional database management system (DBMS) may not be an efficient way of uncovering information pertaining to a specific query. Data normalisation is a critical preliminary step in the data mining process to ensure personal information is secure. Proper normalisation of data is of utmost importance to prevent potential breaches and safeguard highly sensitive information such as medical records, criminal histories, and business financial records. In this blog, we explore the role of data normalisation in securing personal information and preventing potential breaches during data mining.

Various Methods of Normalisation

To establish a new range from existing data, data normalisation is an indispensable scaling technique. It serves as a crucial step for any future processing that involves forecasting or prediction. Different normalising methods such as Min-max, Z-score, and decimal scaling are available.

Application of Min-max Distribution

In data mining, min-max normalisation is an effective technique to linearly transform the initial data range. This method is accurate and allows data to be precisely moulded so that it falls within a specified boundary.

The min-max normalisation technique is particularly advantageous as it preserves the relationship between the initial data values. This method involves transforming the original data to maintain its privacy while retaining the same inter-value distance. By utilising this approach, organisations can secure their data while ensuring data points remain accurately represented.

To execute min-max normalisation, the following steps must be taken:

• Step 1:

  The data owner collects information from the databases.

• Step 2:

  The data owner identifies any potentially problematic details in the dataset.

• Step 3:

  The min-max normalisation technique is employed to modify confidential data. The cleansed information is then returned to the data miner.

Utilising the Z-Score for Normalisation

Z-score normalisation can be utilised to generate normalised values from unstructured data by utilising statistical methods such as the mean and standard deviation. This procedure involves calculating the Z-score for each data point, which is the point’s distance from the dataset’s mean depicted in terms of standard deviation. The objective of this normalisation method is to establish a standard for the data, making it more accessible to interpret and compare with other datasets.

Decimal System Normalisation

This technique yields information that lies between -1 and 1, including the boundaries.

Data Privacy versus Security

It is important to note that “data privacy” and “security” are two distinct terms that cannot be used interchangeably.

• The security of any system is determined by three core components: privacy, trustworthiness, and accessibility. To guarantee a secure system, it is critical to prevent sensitive information from being disclosed, modified, or lost as a result of outside influences. Privacy entails safeguarding the data from unauthorised access, trustworthiness guarantees that the data stored is truthful and trustworthy, and accessibility allows the data to be accessed by authorised personnel. Together, these three security components provide a structure to safeguard data.
• The term “privacy” pertains to an individual’s capacity to determine who learns specific information about them within established laws and protocols.

Protecting Personal Information During Data Mining (PPDM)

The aim of Privacy-Preserving Data Mining (PPDM) is to extract valuable insights from voluminous datasets while upholding data privacy. Below are several strategies that may be employed to achieve this objective:

• Disguising Sensitive Information:

  In this approach, the input data is adjusted and filtered to ensure that confidential information remains undisclosed. This technique assures that all data is preserved since nothing is removed entirely.

• Information Disguising Techniques:

  The use of such techniques further reduces the possibility of a data breach that could reveal sensitive information employed in inferring confidential data.

• Combining Multiple Techniques:

  The hybrid approach leverages the benefits of both data masking and knowledge hiding, both of which have their own limitations.

Techniques for Concealing Sensitive Data

Data obfuscation is a technique used to protect individuals’ privacy and prevent potential misuse of sensitive input data from a database by modifying, blocking, or removing it. By masking the data, unauthorised individuals are prevented from making any inferences or generating any insights that could potentially harm an individual. Data obfuscation is a critical tool in preserving the privacy of individuals and their data.

1. Noise in Data:

   This case study demonstrates how raw data can be altered to protect user privacy. There are various techniques that may be employed to achieve this, such as introducing random variations to the data or rearranging existing values. It is important to note that the quality of the data is upheld when it is released to the public.
   
   With noise addition, the data owner inserts a random number, known as ‘noise,’ into the original set of numbers. Typically, normal distributions with a zero mean and minimal standard deviation are used to generate the random numbers to ensure that the noise closely mirrors the statistical properties of the original data. This process conceals the original data points to safeguard data privacy.
   
   Data mining techniques may be used to accurately reconstruct the distribution of the initial dataset by analysing the noise distribution introduced to the original data and the noisy data. This enables data sets that have been compromised by noise to be restored.
   
   When data is exchanged, it is highly probable that records with similar properties will be exchanged, ensuring that data privacy is preserved, as it is exceedingly difficult to re-identify individuals.

2. Information Encryption Techniques:

   This approach utilises cryptography and secure multiparty computation (SMC) to maintain data confidentiality. SMC enables each participant to maintain their own confidential data while collaborating with others to achieve a common goal. Data protection is guaranteed, with the assurance that no individual will have access to the data or calculation results of others.

3. Confidentiality Techniques:

   The purpose of this technique is to preserve an individual’s privacy by concealing their identity in the input data. One of the most widely used methods to accomplish this is called k-anonymization. To achieve k-anonymity for a table, it is critical to ensure that each row is indistinguishable from the other K-1 rows when using the same attributes.

4. The Condensation Method:

   This technique preserves covariance in the collected data. Firstly, the raw data is divided into subsets of uniform sizes. The statistical attributes of each group, such as the mean and covariance, are then indefinitely preserved. This information is then deployed to generate anonymised data that possesses the same statistical features as the initial dataset. Lastly, the collected data is anonymised and made accessible for data mining purposes.

Tactics for Concealing Known Facts

Knowledge concealment involves concealing or eliminating personally identifiable information from retrieved data. Various techniques may be employed, such as filtering data to remove any personally identifiable information, altering data to render it unrecognisable, or de-identifying data by removing any identifying characteristics that could be used to trace the data back to an individual. These methods of knowledge concealment offer an essential way to safeguard the privacy of individuals while still allowing data to be used for research or other applications.

1. Concealing an Association Rule:

   To safeguard sensitive data, it is necessary to modify the data in such a way that only the pertinent rules are retained, while avoiding any impact on remaining data or rules. During a database query, association rule mining techniques are used to detect and extract rules that surpass a user-defined minimum support and confidence threshold. This threshold is steadily reduced until the support and confidence levels of the sensitive rule are no longer high enough for retrieval.

2. Data Query Audit:

   The suggested approach involves reviewing a user’s previous searches to determine whether the outcomes from the database contain any data that should not be disclosed publicly.

Integrated Approach

The integrated approach involves implementing both randomization and generalisation in sequence to raw data. Randomization is the first step applied to raw data, followed by generalisation, ensuring that no data is lost in the process. This technique has been demonstrated to be more dependable than other methods of knowledge concealment and data hiding, as the modified data may be used to reconstruct the original.

Maintaining data confidentiality during mining processes is critical. Protecting personal information is crucial, but accomplishing it during computing processes is complex. Researchers are presently investigating methods for achieving the optimal solution that balances computing costs and information loss. Currently, there is no single technique that can guarantee the best outcome in all situations.

FAQs

1. How can data mining be standardized?

   Ans: Normalizing data mining involves three key stages:
   • Determine the bounds, which are the highest and lowest values of x.
   • Calculate the difference between the minimum and V.
   • The resulting value of the second step is subsequently divided by the range obtained from the first step.

2. What are the methods for normalizing data?

   Ans: Min-max normalization and z-score normalization are the most frequently used methods of data normalization in data mining, although other techniques also exist.

3. In normalizing data, there are three main phases.

   Ans: The First Normal Form, Second Normal Form, and Third Normal Form are the three stages of database normalization. For data to qualify for the First Normal Form, each column must contain only one unique value. To progress to the Second Normal Form, data must first conform to the First Normal Form and exhibit partial dependency. Finally, in the Third Normal Form, the data must not have a transitive dependency.

4. Could you provide an example of data normalization and its function?

   Ans: In order to enable extensive analysis and retrieval, data normalization implements a scaling method that creates a new range within an existing range. By using min-max normalization, the values in the original data set (1000, 2000, 3000, 9000) can be transformed to have a minimum of 0 and a maximum of 1, resulting in (0, 0.125, 0.25, 1), making the data more useful for further analysis.