Skip to main content

Data Privacy Mechanisms provided by VulcanSQL for Easier Data Sharing

· 7 min read
Jimmy Yeh

cover

What is data sharing and why it's important?

Data sharing is a behavior that we share data to other parties which maybe other departments in the same company, or customers outside the company, etc. The reason why we agree to share data is that we finally realize data is a valuable asset especially to business, since it can make the business process more smooth and enable better decision making results!

When thinking about how to share data with others, there are some common requirements coming up to our minds, such as what are the formats of data being stored, what kind of storage should persist the data, and how to share data with other parties. However, there are also other directions we need to consider and I would like to discuss data privacy with you in this article.

Why data privacy is important to data sharing?

In terms of sharing data with others, there are certainly some scenraios that we can treat everyone we share with equally.

For example, datasets found in Kaggle competitions are normally the same to all competition participants. However, in the business world, we should especially obey the principle of least privilege(PoLP). It's an information security concept which maintains that a user or entity should only have access to the specific data, resources and applications needed to complete a required task1.

Given that we understand data sharing is a great opportunity to share valuable resources with others, and we also realize we need to control how different parts of data can be accessed from different parties. Now the question comes to how do we share data with others in a more contrallable way that can fulfill the data privacy requirement?

How VulcanSQL can help?

VulcanSQL comes with several built-in data privacy mechanisms to enable a more controllable and scalable data sharing use case!

As of now, VulcanSQL has 5 techniques for handling data privacy, here are some brief introduction to each technique and in the Showcase section, we'll show you the code and explain further accordingly.

Authentication

Authentication is the first layer of data privacy protection. Any entity that is not authenticated is not allowed to access any resources. At the moment, VulcanSQL only accepts three authentication methods, namely HTTP Basic, password file and simple token. Since we realize there should be a more mature and easy way for users to authenticate, we plan to support OpenID Connect in the future.

info

You can understand more about the authentication mechanism in VulcanSQL here!

Authorization

With authorization, VulcanSQL applies an attribute-based access control(ABAC) approach to control access based on user attributes provided by Authenticator. In VulcanSQL, we can configure each user's attributes in vulcan.yaml; then we can define different policies for different users based on their attributes for each data source in profiles.yaml. With this mechanism, different users would see different parts of the data based on their attributes defined in VulcanSQL!

info

You can understand more about the authorization mechanism in VulcanSQL here!

Dynamic Data Masking

Sometimes, we want to share masked data to users. The purpose is to protect the actual data while having a functional substitute for occasions when the real data is not required!

With dynamic data masking, we can define a specific pattern for masking the real data, such as transforming an ID from F123456789 to F12xxxx89 using a partial(3, 'xxxx', 2) function.

info

Column-level Security

If we need to have fine grained control over some specific columns, we can use the Column-level Security(CLS) mechanism to achieve the goal. In VulcanSQL, we can decide who can access the specific column based on their user attributes.

Row-level Security

Similar to the case of Column-level Security, if we need to have fine grained control over some specific rows, we can use the Row-level Security(RLS) mechanism to achieve the goal. In VulcanSQL, we can decide who can access the specific row based on their user attributes.

info

Showcase

Now we are going to show you the code to demonstrate how you can deliver data privacy mechanisms in VulcanSQL! For those who may not familiar with VulcanSQL yet, VulcanSQL is a Data API framework for data folks to create REST APIs easily by writing templated SQL! It's mainly used for sharing data from databases, data warehouses and data lakes!

If you would like to read the source code or try the example by yourself, welcome to check it out here!

Below is the dataset we'll use in the showcase:

iddepartmentlast_namecompany_roleannual_salary
JDK32424engineeringJamesengineer"$100,000"
EKJ34124salesHardensales"$120,000"
MKO56124salesMichaelmanager"$110,000"
ONP01124engineeringCindymanager"$115,000"
NZP59124ceoRosaboss"$150,000"

Below is the code you may write in SQL templates in VulcanSQL:

SELECT
-- dynamic data masking
{% masking id partial(2, 'xxxxxxx', 2) %} as id,
department,
last_name,
company_role,
-- column level security
{% if context.user.attr.role == 'employer' %}
annual_salary
{% else %}
NULL AS annual_salary
{% endif %}
FROM read_csv_auto('departments.csv', HEADER=True)
-- row level security
{% if context.user.attr.role != 'employer' %}
WHERE department = {{ context.user.attr.department }}
{% endif %}

Here is the auth configuration in vulcan.yaml:

auth:
enabled: true
options:
basic:
# Read users and passwords from a text file.
htpasswd-file:
path: passwd.txt # Path to the password file.
users: # (Optional) Add attributes for users
- name: james
attr:
role: employee
department: engineering
- name: harden
attr:
role: employee
department: sales
- name: michael
attr:
role: employee
department: sales
- name: cindy
attr:
role: employee
department: engineering
- name: rosa
attr:
role: employer
department: ceo

The REST API results you'll see based on different users:

James

iddepartmentlast_namecompany_roleannual_salary
JDxxxxxxx24engineeringJamesengineer
ONxxxxxxx24engineeringCindymanager

Harden

iddepartmentlast_namecompany_roleannual_salary
EKxxxxxxx24salesHardensales
MKxxxxxxx24salesMichaelmanager

Michael

iddepartmentlast_namecompany_roleannual_salary
EKxxxxxxx24salesHardensales
MKxxxxxxx24salesMichaelmanager

Cindy

iddepartmentlast_namecompany_roleannual_salary
JDxxxxxxx24engineeringJamesengineer
ONxxxxxxx24engineeringCindymanager

Rosa

iddepartmentlast_namecompany_roleannual_salary
JDxxxxxxx24engineeringJamesengineer$100,000
EKxxxxxxx24salesHardensales$120,000
MKxxxxxxx24salesMichaelmanager$110,000
ONxxxxxxx24engineeringCindymanager$115,000
NZxxxxxxx24ceoRosaboss$150,000

After observing the result tables shown above and the given SQL template, we can clearly figure out several data privacy mechanisms provided by VulcanSQL:

  1. Authentication: In the above example, we used HTTP Basic as the authentication method and the password of each user was stored in a text file called passwd.txt.
  2. Authorization: You can find that we had defined user attributes in vulcan.yaml. With these user attributes defined, we can have more fine grained control on what kind of data each user should access.
  3. Dynamic Data Masking: {% masking id partial(2, 'xxxxxxx', 2) %} as id makes only the first two and last two digits of id visible, and the rest is masked.
  4. Column-level Security: We can see that only the user who is in the employer role can see the data in the salary field, so Rosa is the only person who can see other people's salary.
  5. Row-level Security: We can see a user who is not in the employer role can only see the data in the same department as he/she.

Conclusion

Data privacy is more important than ever. We may regard it as a special kind of human rights, so we should protect the data from being abused!

We hope this blog post effectively highlights the significance of data privacy when sharing data with other parties. It also showcases that VulcanSQL offers user-friendly solutions that are certainly worth exploring.


  1. The definition of the principle of least privilege(PoLP) is referenced from the article from paloalto networks.