Dynamic data masking - Amazon Redshift
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Dynamic data masking

Overview

Using dynamic data masking (DDM) in Amazon Redshift, you can protect sensitive data in your data warehouse. You can manipulate how Amazon Redshift shows sensitive data to the user at query time, without transforming it in the database. You control access to data through masking policies that apply custom obfuscation rules to a given user or role. In that way, you can respond to changing privacy requirements without altering underlying data or editing SQL queries.

Dynamic data masking policies hide, obfuscate, or pseudonymize data that matches a given format. When attached to a table, the masking expression is applied to one or more of its columns. You can further modify masking policies to only apply them to certain users, or to user-defined roles that you can create with Role-based access control (RBAC). Additionally, you can apply DDM on the cell level by using conditional columns when creating your masking policy. For more information about conditional masking, see Conditional dynamic data masking.

You can apply multiple masking policies with varying levels of obfuscation to the same column in a table and assign them to different roles. To avoid conflicts when you have different roles with different policies applying to one column, you can set priorities for each application. In that way, you can control what data a given user or role can access. DDM policies can partially or completely redact data, or hash it by using user-defined functions written in SQL, Python, or with Amazon Lambda. By masking data using hashes, you can apply joins on this data without access to potentially sensitive information.

End-to-end example

The following is an end-to-end example showing how you can create and attach masking policies to a column. These policies let users access a column and see different values, depending on the degree of obfuscation in the policies attached to their roles. You must be a superuser or have the sys:secadmin role to run this example.

Creating a masking policy

First, create a table and populate it with credit card values.

--create the table CREATE TABLE credit_cards ( customer_id INT, credit_card TEXT ); --populate the table with sample values INSERT INTO credit_cards VALUES (100, '4532993817514842'), (100, '4716002041425888'), (102, '5243112427642649'), (102, '6011720771834675'), (102, '6011378662059710'), (103, '373611968625635') ; --run GRANT to grant permission to use the SELECT statement on the table GRANT SELECT ON credit_cards TO PUBLIC; --create two users CREATE USER regular_user WITH PASSWORD '1234Test!'; CREATE USER analytics_user WITH PASSWORD '1234Test!'; --create the analytics_role role and grant it to analytics_user --regular_user does not have a role CREATE ROLE analytics_role; GRANT ROLE analytics_role TO analytics_user;

Next, create a masking policy to apply to the analytics role.

--create a masking policy that fully masks the credit card number CREATE MASKING POLICY mask_credit_card_full WITH (credit_card VARCHAR(256)) USING ('000000XXXX0000'::TEXT); --create a user-defined function that partially obfuscates credit card data CREATE FUNCTION REDACT_CREDIT_CARD (credit_card TEXT) RETURNS TEXT IMMUTABLE AS $$ import re regexp = re.compile("^([0-9]{6})[0-9]{5,6}([0-9]{4})") match = regexp.search(credit_card) if match != None: first = match.group(1) last = match.group(2) else: first = "000000" last = "0000" return "{}XXXXX{}".format(first, last) $$ LANGUAGE plpythonu; --create a masking policy that applies the REDACT_CREDIT_CARD function CREATE MASKING POLICY mask_credit_card_partial WITH (credit_card VARCHAR(256)) USING (REDACT_CREDIT_CARD(credit_card)); --confirm the masking policies using the associated system views SELECT * FROM svv_masking_policy; SELECT * FROM svv_attached_masking_policy;

Attaching a masking policy

Attach the masking policies to the credit card table.

--attach mask_credit_card_full to the credit card table as the default policy --all users will see this masking policy unless a higher priority masking policy is attached to them or their role ATTACH MASKING POLICY mask_credit_card_full ON credit_cards(credit_card) TO PUBLIC; --attach mask_credit_card_partial to the analytics role --users with the analytics role can see partial credit card information ATTACH MASKING POLICY mask_credit_card_partial ON credit_cards(credit_card) TO ROLE analytics_role PRIORITY 10; --confirm the masking policies are applied to the table and role in the associated system view SELECT * FROM svv_attached_masking_policy; --confirm the full masking policy is in place for normal users by selecting from the credit card table as regular_user SET SESSION AUTHORIZATION regular_user; SELECT * FROM credit_cards; --confirm the partial masking policy is in place for users with the analytics role by selecting from the credit card table as analytics_user SET SESSION AUTHORIZATION analytics_user; SELECT * FROM credit_cards;

Altering a masking policy

The following section shows how to alter a dynamic data masking policy.

--reset session authorization to the default RESET SESSION AUTHORIZATION; --alter the mask_credit_card_full policy ALTER MASKING POLICY mask_credit_card_full USING ('00000000000000'::TEXT); --confirm the full masking policy is in place after altering the policy, and that results are altered from '000000XXXX0000' to '00000000000000' SELECT * FROM credit_cards;

Detaching and dropping a masking policy

The following section shows how to detach and drop masking policies by removing all dynamic data masking policies from the table.

--reset session authorization to the default RESET SESSION AUTHORIZATION; --detach both masking policies from the credit_cards table DETACH MASKING POLICY mask_credit_card_full ON credit_cards(credit_card) FROM PUBLIC; DETACH MASKING POLICY mask_credit_card_partial ON credit_cards(credit_card) FROM ROLE analytics_role; --drop both masking policies DROP MASKING POLICY mask_credit_card_full; DROP MASKING POLICY mask_credit_card_partial;

Considerations when using dynamic data masking

When using dynamic data masking, consider the following:

  • When querying objects created from tables, such as views, users will see results based on their own masking policies, not the policies of the user who created the objects. For example, a user with the analyst role querying a view created by a secadmin would see results with masking policies attached to the analyst role.

  • To prevent the EXPLAIN command from potentially exposing sensitive masking policy filters, only users with the SYS_EXPLAIN_DDM permission can see masking policies applied in EXPLAIN outputs. Users don't have the SYS_EXPLAIN_DDM permission by default.

    The following is the syntax for granting the permission to a user or role.

    GRANT EXPLAIN MASKING TO { username | ROLE rolename }

    For more information about the EXPLAIN command, see EXPLAIN.

  • Users with different roles can see differing results based on the filter conditions or join conditions used. For example, running a SELECT command on a table using a specific column value will fail if the user running the command has a masking policy applied that obfuscates that column.

  • DDM policies must be applied ahead of any predicate operations, or projections. Masking polices can include the following:

    • Low cost constant operations such as converting a value to null

    • Moderate cost operations such as HMAC hashing

    • High cost operations such as calls to external Lambda user defined functions

    As such, we recommend that you use simple masking expressions when possible.

  • You can use DDM policies for roles with row-level security policies, but note that RLS policies are applied before DDM. A dynamic data masking expression won't be able to read a row that was protected by RLS. For more information about RLS, see Row-level security.

  • When using the COPY command to copy from parquet to protected target tables, you should explicitly specify columns in the COPY statement. For more information about mapping columns with COPY, see Column mapping options.

  • DDM policies can't attach to the following relations:

    • System tables and catalogs

    • External tables

    • Datasharing tables

    • Materialized views

    • Cross-DB relations

    • Temporary tables

    • Correlated queries

  • DDM policies can contain lookup tables. Lookup tables can be present in the USING clause. The following relation types can’t be used as lookup tables:

    • System tables and catalogs

    • External tables

    • Datasharing tables

    • Views, materialized views, and late-binding views

    • Cross-DB relations

    • Temporary tables

    • Correlated queries

    Following is an example of attaching a masking policy to a lookup table.

    --Create a masking policy referencing a lookup table CREATE MASKING POLICY lookup_mask_credit_card WITH (credit_card TEXT) USING ( CASE WHEN credit_card IN (SELECT credit_card_lookup FROM credit_cards_lookup) THEN '000000XXXX0000' ELSE REDACT_CREDIT_CARD(credit_card) END ); --Provides access to the lookup table via a policy attached to a role GRANT SELECT ON TABLE credit_cards_lookup TO MASKING POLICY lookup_mask_credit_card;
  • You can't attach a masking policy that would produce an output incompatible with the target column's type and size. For example, you can’t attach a masking policy that outputs a 12 character long string to a VARCHAR(10) column. Amazon Redshift supports the following exceptions:

    • A masking policy with the input type INTN can be attached to a policy with size INTM as long as M < N. For example, a BIGINT (INT8) input policy can be attached to a smallint (INT4) column.

    • A masking policy with the input type NUMERIC or DECIMAL can always be attached to a FLOAT column.

  • DDM policies can't be used with data sharing. If the datashare's data producer attaches a DDM policy to a table in the datashare, the table becomes inaccessible to users from the data consumer who are trying to query the table. Tables with DDM policies attached can't be added to a datashare.

  • You can't query relations that have attached DDM policies if your values for any of the following configuration options don't match the default value of the session:

    • enable_case_sensitive_super_attribute

    • enable_case_sensitive_identifier

    • downcase_delimited_identifier

    Consider resetting your session’s configuration options if you attempt to query a relation with a DDM policy attached and see the message "DDM protected relation does not support session level config on case sensitivity being different from its default value."

  • When your provisioned cluster or serverless namespace has any dynamic data masking policies, the following commands are blocked for regular users:

    ALTER <current_user> SET enable_case_sensitive_super_attribute/enable_case_sensitive_identifier/downcase_delimited_identifier

    When you create DDM policies, we recommend that you change the default configuration option settings for regular users to match the session’s configuration option settings at the time the policy was created. Superusers and users with the ALTER USER privilege can do this by using parameter group settings or the ALTER USER command. For information about parameter groups, see Amazon Redshift parameter groups in the Amazon Redshift Management Guide. For information about the ALTER USER command, see ALTER USER.

  • Views and late-binding views with attached DDM policies can't be replaced by regular users using the CREATE VIEW command. To replace views or LBVs with DDM policies, first detach any DDM policies attached to them, replace the views or LBVs, and reattach the policies. Superusers and users with the sys:secadmin permission can use CREATE VIEW on views or LBVs with DDM policies without detaching the policies.

  • Views with attached DDM policies can't reference system tables and views. Late-binding views can reference system tables and views.

  • Late-binding views with attached DDM policies can't reference nested data in data lakes, such as JSON documents.

  • Late-binding views can't have DDM policies attached if that late-binding view is referenced by any view.

  • DDM policies attached to late-binding views are attached by column name. At query time, Amazon Redshift validates that all masking policies attached to the late-binding view have been applied successfully, and that the late-binding view's output column type matches the types in the attached masking policies. If the validation fails, Amazon Redshift returns an error for the query.