登录查看更多内容

How can you mask sensitive data in your datasets?

由人工智能和领英社区提供技术支持

Data is a valuable asset for any data science project, but it can also contain sensitive information that needs to be protected from unauthorized access or disclosure. Data masking is a technique that replaces or obscures some or all of the original data with realistic but fictitious values, while preserving its structure and meaning for analysis. Data masking can help you comply with data privacy regulations, protect your customers' and employees' identities, and reduce the risk of data breaches. In this article, you will learn about some common methods and tools for data masking, as well as some best practices and challenges.

此文章中的业界达人

由社区从 4 条内容中精选。了解更多

Jaganadh Gopinadhan (Jagan)

Associate Director - Engineering @ Cognizant | Trusted AI Advisor, Generative AI Strategist, AI Infrastructure, Cloud…
Joshua Paiva

Principal Data Scientist

1 Why data masking?

Data masking can provide several benefits for data science projects, such as enhancing data security and privacy, enabling data sharing and collaboration, and preserving data quality and usability. Data masking can prevent sensitive data from being exposed or stolen by hackers, malicious insiders, or third parties. It can also help you comply with data protection laws and standards, like GDPR, HIPAA, or PCI DSS. Furthermore, it allows you to share your data with other teams, partners, or vendors without compromising the confidentiality or integrity of the original data. Moreover, data masking can retain the essential characteristics and relationships of the original data so that the masked data is still suitable and meaningful for data analysis, modeling, and visualization.

添加您的观点

Jaganadh Gopinadhan (Jagan)

Associate Director - Engineering @ Cognizant | Trusted AI Advisor, Generative AI Strategist, AI Infrastructure, Cloud Engineering, AI Engineering, GenAI/AI/ML and Data Science
举报内容
Data masking is the process of hiding protected and sensitive data by modifying its original value. Data masking allows Data Scientists and Data Analysts to collaborate without compromising the organization's data. This ensures we comply with regulatory requirements such as GDPR, HIPAA, and CCPA. Regarding Data Science, the data is accessed as raw and visualized in notebooks. It is one of the high-risk activities where protected/sensitive data may get exposed and stay unprotected. Ensuring the data is masked and discretized eliminates the risk. Assessing the data access risk at every stage of Data Science and ML projects is essential.

已翻译

赞

2 How data masking works?

Data masking is an effective method for protecting data at various stages of the data life cycle. It can be classified into two main types: static data masking (SDM) and dynamic data masking (DDM). SDM involves masking the data at rest, such as in databases, files, or backups. This creates a copy of the original data with the masked values, and stores it in a separate location. Usually done once or periodically, SDM is suitable for long-term or permanent data protection. DDM, on the other hand, involves masking the data in transit, such as in queries, reports, or applications. This does not alter the original data, but only modifies what is displayed or delivered to users or systems. Typically done on-demand or in real-time, DDM is suitable for short-term or temporary data protection.

添加您的观点

Jaganadh Gopinadhan (Jagan)

Associate Director - Engineering @ Cognizant | Trusted AI Advisor, Generative AI Strategist, AI Infrastructure, Cloud Engineering, AI Engineering, GenAI/AI/ML and Data Science
举报内容
There are various techniques to protect the data at rest, including Dynamic Data Masking (DDM) and Static Data Masking (SDM). The data miners' discretization methodology is an effective way to hide the original value statistically. Discretization is very effective in protecting personal attributes such as date of birth. When data Science tools access the data from a data system, the security of both platforms and communication protocols like TLS ensures the safe travel of data. Beyond all of these safeguards, the Data Scientists should avoid making offline copies of the data involved sensitive data.

已翻译

赞

3 What are some data masking techniques?

Data masking is a process of obscuring sensitive data, and there are various techniques that can be used depending on the data type, sensitivity, and masking level. Substitution replaces original data with random or predefined values from a list or dictionary. Shuffling rearranges the original data within a column or table, dissociating it from its original records. Masking out hides or deletes some or all of the original data with blanks, asterisks, or other symbols. Encryption transforms data into an unreadable form using a mathematical algorithm and key. Pseudonymization replaces the original data with a unique identifier or token that can be linked back to the original data through a mapping table or function. This technique provides a balance between data protection and utility as masked data can be re-identified under certain conditions.

添加您的观点

4 What are some data masking tools?

Data masking can be performed with a variety of tools, depending on the type and source of the data, and the features and functionalities you need. Popular options include SQL Server, which offers built-in dynamic data masking capabilities, Oracle Data Masking and Subsetting, IBM InfoSphere Optim, and Delphix. SQL Server Integration Services (SSIS) can also be used to perform static data masking with custom scripts or components. Oracle Enterprise Manager allows for masking and subsetting of Oracle and non-Oracle databases with predefined or custom masking formats. IBM InfoSphere Optim provides data masking and subsetting capabilities for databases, files, or applications using predefined or custom algorithms. Delphix enables data masking and virtualization for databases, files, or cloud services with predefined or custom policies. All of these tools can create masked data subsets that are consistent, representative, scalable, optimized, fast, and efficient for testing and development purposes, data sharing and collaboration.

添加您的观点

5 What are some best practices and challenges for data masking?

Data masking can be a powerful and effective technique for data security and privacy. However, it is important to assess your data sensitivity and risk, and determine the appropriate masking technique and level for each data element. Additionally, it is essential to test and validate the masked data to ensure that it meets your data quality and usability requirements, as well as monitor and audit the masked data to comply with data privacy regulations. Lastly, there needs to be a balance between data protection and data utility when masking your data, as it may affect the accuracy, completeness, or relevance of the original data. It is necessary to consider the purpose and scope of the data analysis, as well as the needs and expectations of the data users.

添加您的观点

Joshua Paiva

Principal Data Scientist
举报内容
A quick tip: When using common masking techniques, such as a string over-write, always consider problems with duplicity. A simple example: if we're masking sensitive information such as credit card numbers, replacing string '8000567891' with '8000XXXXXX', consider the implications of all records beginning with the '8000' string. It's quite possible that string masking may result in overlapping records on the masked-level, and this should always be considered.

已翻译

赞

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?

添加您的观点

Jaganadh Gopinadhan (Jagan)

Associate Director - Engineering @ Cognizant | Trusted AI Advisor, Generative AI Strategist, AI Infrastructure, Cloud Engineering, AI Engineering, GenAI/AI/ML and Data Science
举报内容
Data Science projects with sensitive data require more than data protection at rest and in transit. Data protection in use is very critical in Data science with sensitive data. One of the emerging paradigms in space is 'Confidential Computing,' which ensures data security and algorithms while 'data in use.' Confidential Computing ensures that the algorithm and data are in use within a trusted execution environment (TES) and that the hardware is trusted and protected. Beyond all, these adversaries may not be able to access the data while it is in use. The paradigm expands its offering to containerized workloads, which helps deploy Data Science and AI workloads in containers (for processing sensitive data).

已翻译

赞

Data Science

+ 关注

给文章评分

我们借助人工智能创建了此文章。您认为这篇文章怎么样？

很棒不太好

举报此文章

查看全部

How can you mask sensitive data in your datasets?

1

2

3

4

5

6

1 Why data masking?

2 How data masking works?

3 What are some data masking techniques?

4 What are some data masking tools?

5 What are some best practices and challenges for data masking?

6 Here’s what else to consider

Data Science

给文章评分

感谢您的反馈

更多Data Science相关文章

更多相关阅读内容

How can you mask sensitive data in your datasets?

1

2

3

4

5

6

1 Why data masking?

2 How data masking works?

3 What are some data masking techniques?

4 What are some data masking tools?

5 What are some best practices and challenges for data masking?

6 Here’s what else to consider

Data Science

给文章评分

感谢您的反馈

查看其他技能