ARA-C01 Question 28
Single answer1.2 Design an architecture that meets data security, privacy, compliance, and governance requirements.A healthcare company is migrating patient analytics workloads to Snowflake. The platform must support analysts in multiple business units, but only authorized users should be able to see personally identifiable information (PII) such as patient name, email, and phone number. Security administrators also require that masking logic be centrally governed, consistently enforced across many tables, and manageable without rewriting every BI query. Which approach BEST meets these requirements?
- A
Create secure views for each consuming team and manually exclude PII columns from every view definition
- B
Apply dynamic data masking policies to the sensitive columns and use role-based conditions in the policy body to determine who can see unmasked values
- C
Encrypt PII columns in source files before loading them into Snowflake and give analysts the decryption keys when they need access
- D
Use row access policies on the patient tables so unauthorized users cannot see PII values in masked columns
- E
Store PII in a separate database and grant USAGE on that database only to the security team, while analysts query replicated copies without governance policies
Show answer and explanation
Correct answer: B
Explanation
The best answer is to use dynamic data masking policies on sensitive columns. In Snowflake, masking policies provide centralized, policy-based control over how column values are exposed to different roles. This is well aligned to architectures that must meet data privacy, compliance, and governance requirements, especially for regulated data such as healthcare PII. Because the policy is attached to the column, downstream queries and BI tools do not need to be rewritten to implement masking behavior. Architects can combine masking policies with role-based access control (RBAC), schema design, and tagging strategies for broader governance. Row access policies are complementary for row-level filtering, but they do not replace masking for column-level privacy protection. Secure views can be useful in some cases, but they are generally less scalable than policy-based controls when many tables and consumers are involved. Relevant Snowflake best-practice areas include Dynamic Data Masking, Row Access Policies, Secure Views, and RBAC-based governance.
- A. Incorrect.
This is not the best approach. Secure views can help limit exposure and protect underlying logic, but manually maintaining many views across business units does not provide centralized, column-level governance at scale. It also increases administrative overhead and risks inconsistent implementation. The requirement specifically calls for centrally governed masking logic that does not require rewriting each BI query.
- B. Correct.
This is correct. Dynamic data masking allows architects to attach masking policies directly to sensitive columns and determine whether values are masked based on the querying role, current role hierarchy, or other policy logic. This approach is centralized, reusable, and transparent to consuming queries and BI tools, making it a strong fit for privacy and governance requirements involving PII.
- C. Incorrect.
This is incorrect. Encrypting data before loading may protect data in transit or outside Snowflake, but giving analysts decryption keys undermines governance and makes access control difficult to centralize and audit. It also shifts enforcement outside Snowflake rather than using Snowflake's built-in policy framework for fine-grained access control.
- D. Incorrect.
This is incorrect. Row access policies control which rows a user can see, not how specific column values are masked. They are useful when access must be restricted by row-level entitlements, such as region or tenant. They do not solve the primary need here, which is selective visibility of PII columns for authorized users.
- E. Incorrect.
This is incorrect. Physically separating PII into another database may reduce accidental access in some designs, but replicated copies without governance policies create additional risk and operational complexity. This does not provide centralized, consistent masking enforcement across many tables and may lead to duplicate datasets with uneven controls.