Policy logic

Immuta row-level policies compare data values with user metadata at query-time to determine whether or not the querying user should have access to the individual rows of data.

Referencing data values

The values contained in one or many columns in the table in question (or a separate joined table) need to be referenced by the policy for its logic to take effect.

For example, consider the policy below:

Only show rows where user is a member of a group that matches the value in the column tagged Department.

The data values (the values in the column tagged Department) are matched against the user attribute (their groups) to determine whether or not rows will be visible to the user accessing the data.

The policy targets columns tagged Department; this means that this policy can be applied globally across all tables and data platforms that have that tag with this single policy rather than having to build a separate policy for individual tables and columns.

Leveraging custom functions

It is also possible to use custom functions in custom WHERE row-level policies for more complex use cases.

These wrap Immuta context into free-form SQL logic for the row-level policy. That context can be things like the attributes (@attributeValuesContains()) or groups (@groupsContains()) possessed by the user or the username (@username) - injected into the SQL at runtime.

Avoid referencing explicit column names in custom functions and instead use the @columnTagged('tag name') function in SQL. In doing so, you can avoid having to reference the physical database world with the custom SQL policies and instead continue to target the metadata/tag world.

Orchestrated masking policies (OMP) reduce conflicts between masking policies that apply to a single column, allowing policies to scale more effectively across your organization. Furthermore, OMP fosters distributed data stewardship, empowering policy authors who share responsibility of a data set to protect it while allowing data consumers acting under various roles or purposes to access the data.

When multiple masking policies apply to a column, Immuta combines the exception conditions of the masking policy so that data subscribers can access the data when they satisfy one of those exception conditions. Multiple masking policies will be enforced on a column if the following conditions are true:

• Policies use the same masking type.

• Policies use the for everyone except condition.

Requirements

Databricks Spark or Starburst (Trino) integration

Supported masking policy types

OMP supports the following masking types:

• Constant

• Hashing

• Format preserving masking

• Null

• Regex

• Rounding

Global policy logic

Previous policy logic

Governors can apply policies to all columns in a data source or target specific columns with tags or a regular expression. Without orchestrated masking policies enabled, when multiple global policies apply to the same columns, Immuta could only apply one of those policies.

Consider the following example to examine how policies behaved when one tag is used in two different policies:

• Mask PII Global Policy 1: Mask using hashing the value in columns tagged email except when user is acting under the purpose Email Campaign.

• Mask PII Global Policy 2: Mask using hashing the value in columns tagged email except when user is acting under purpose Marketing.

For columns tagged email, only one of these policies is enforced. The Mask PII Global Policy 2 is not applied to the data source, so Immuta is not enforcing the masking policy properly for users who should be able to see emails because they are acting under the Marketing purpose.

Consider the following example where multiple masking policies apply to columns that have multiple tags, resulting in one policy applying:

• Global Policy 3: Mask using hashing the value in columns tagged Employee Data unless users are acting under the purpose Retention Analysis.

• Global Policy 4: Mask using hashing the value in columns tagged HR Data unless users are acting under the purpose Employee Satisfaction Survey.

If a column is tagged Employee Data and HR Data, Immuta will only apply one of the policies.

Orchestrated masking policy logic

With orchestrated masking policies, Immuta applies multiple global masking policies that apply to a single column by combining the policy exceptions with OR. For these policies to combine, the masking type must be identical and the policy must use the for everyone except condition.

Consider the following example, both of these policies will apply to the data source:

• Mask PII Global Policy 1: Mask using hashing the value in columns tagged email except when user is acting under the purpose Email Campaign.

• Mask PII Global Policy 2: Mask using hashing the value in columns tagged email except when user is acting under purpose Marketing.

Users acting under the purpose Marketing or Email Campaign will be able to see emails in the clear.

However, in the following example, only one of these policies will apply to the data source because one masks using a constant and the other masks using hashing:

• Global Policy 5: Mask using the constant REDACTED the value in columns tagged Employee Data unless users are acting under the purpose Retention Analysis.

• Global Policy 6: Mask using hashing the value in columns tagged HR Data unless users are acting under the purpose Employee Satisfaction Survey.

Limitations

• No UI enhancements were made in this release. Multiple masking policies applied to the same column are visible on a data source, but there is no indication that the exceptions are combined with OR.

• Masking types must match exactly for the policies to be combined. For example, both policies must mask using rounding.

• Existing policies will not automatically migrate to the new policy logic when you enable the feature. To re-compute existing policies with the new logic, you must manually trigger global policy changes by staging and re-enabling each policy.

Masking policy conflicts

In some cases, two conflicting global masking policies apply to a single data source. When this happens, the policy containing a tag deeper in the hierarchy will apply to the data source to resolve the conflict.

Consider the following global data policies created by a data governor:

Data policy 1:

Mask columns tagged PII by making null for everyone on data sources with columns tagged PII

Data policy 2:

Mask columns tagged PII.SSN using hashing for everyone on data sources with columns tagged PII.SSN

If a data owner creates a data source and applies the PII.SSN tag to a column, both of these global masking policies will apply to the column with that tag. Instead of having a conflict, the policy containing a deeper tag in the hierarchy will apply.

In this example, data policy 2 will be applied to the data source because PII.SSN is deeper and thus considered more specific than PII. If data owners wanted to use data policy 1 on the data source instead, they would need to disable data policy 2.

Should two or more masking policies target the same column and have the same hierarchy depth, the policy that was authored first will win out. This is a conservative approach that avoids the original policy being changed unexpectedly.

Row-level policy conflicts

Similar to masking policies, it is possible for two or more row-level policies to target the same table. When this occurs, all row-level policies will be applied and AND'ed together, meaning the user will need to meet all in some capacity to see any rows in the table at all.

To OR separate row-level policies together, build them into a single Immuta policy together with an OR.

Masking policy intelligent fallbacks

When masking columns, the type of the column matters. For example, it is not possible to hash a numeric column, because the hash would render the number as a string.

Many data platforms make the user account for this by building separate data policies for every column type that could exist now or in the future, which is quite onerous.

Instead, Immuta has intelligent fallbacks. An intelligent fallback occurs when a masking type targets a column type that is incompatible with the masking type. In this case, Immuta will fall back to the most appropriate masking type which retains the level of privacy or better required by the previous type.

For example, if a hashing masking type hits a numeric type, it would intelligently fallback to nulling the column instead, since nulls are allowed in numeric types.

Lockout policies

Sometimes a global data policy will target a table and the policy cannot be applied as written. This can happen for several reasons, but the most common is that the row-level policy logic is not relevant to the table in question.

For example, with the following policy

@attributeValuesContains('Attribute Name', 'SOME_COLUMN')

If SOME_COLUMN does not exist in the table, the row-level policy will not work (this is why it is always recommended to use the @columnTagged('tag name') function instead of hard coding column names).

In the case where an error such as this occurs with a global data policy, the lockout policy will kick in. The lockout policy is a row-level policy that blocks any rows from returning for any users. This may seem extreme, but since Immuta does not know how to apply the policy, the lockout policy avoids data leaks until the policy is edited to work correctly.

Overview

The Immuta Policy Builder allows you to use custom functions that reference important Immuta metadata from within your where clause. These custom functions can be seen as utilities that help you create policies easier. Using the Immuta Policy Builder, you can include these functions in your policy queries by choosing where in the sub-action drop-down menu.

Custom Functions

The @attributeValuesContains() Function

This function returns true for a given row if the provided column evaluates to an attribute value for which the querying user has a corresponding attribute value. This function requires two arguments and accepts no more than three arguments.

Parameters

The @columnTagged() Function

This function returns the column name with the specified tag.

If this function is used in a Global Policy and the tag doesn't exist on a data source, the policy will not be applied.

Parameters

The @groupsContains() Function

This function returns true for a given row if the provided column evaluates to a group to which the querying user belongs. This function requires at least one argument.

Parameters

The @hasAttribute() Function

This function returns a boolean indicating if the current user has the specified attribute name and value combination. If the specified attribute name or attribute value has a single quote, you will need to escape it using a \'\' expression within a custom WHERE policy.

Parameters

The @iam Function

This function returns the IAM ID for the current user.

Parameters

None.

The @isInGroups() Function

This function returns a boolean indicating if the current user is a member of all of the specified groups. If any of the specified groups has a single quote, you will need to escape it using a \'\' expression within a custom WHERE policy.

Parameters

The @isUsingPurpose() Function

This function returns a boolean indicating if the current user is using the specified purpose. If the specified purpose has a single quote, you will need to escape it using a \'\' expression within a custom WHERE policy.

Parameters

The @purposesContains() Function

This function returns true for a given row if the provided column evaluates to a purpose under which the querying user is currently acting. This function requires at least one argument and accepts no more than two arguments.

Parameters

The @username Function

This function returns the current user's user name.

Parameters

None.

When building a global data policy, governors can create custom certifications, which must then be acknowledged by data owners when the policy is applied to data sources.

For example, data governors could add a custom certification that states that data owners must verify that tags have been added correctly to their data sources before certifying the policy.

When a global data policy with a custom certification is cloned, the certification is also cloned. If the user who clones the policy and custom certification is not a governor, the policy will only be applied to data sources that user owns.

Once a user is subscribed to a data source, the data policies that are applied to that data source determine what data the user sees.

For all data policies, you must establish the conditions for which they will be enforced. Immuta allows you to append multiple conditions to the data. Those conditions are based on user attributes and groups (which can come from multiple identity management systems and applied as conditions in the same policy), or purposes they are acting under through Immuta projects.

Conditions can be directed as exclusionary or inclusionary, depending on the policy that's being enforced:

• exclusionary condition example: Mask using hashing values in columns tagged PII on all data sources for everyone except users in the group AUDIT.

• inclusionary condition example: Only show rows where user is a member of a group that matches the value in the column tagged Department.

Policy support matrix

The table below outlines the policy types supported by each integration. Details about each of these policies are included in the policy types section.

*Supported with Caveats:

• On Databricks data sources, joins will not be allowed on data protected with replace with NULL/constant policies.

• On Trino data sources, the Immuta function @iam for WHERE clause policies can block the creation of views.

Policy types

Inclusionary policies

For all policies except purpose-based restriction policies, inclusionary logic allows governors to vary policy actions with an Otherwise clause.

For example, governors could mask values using hashing for users acting under a specified purpose while masking those same values by making null for everyone else who accesses the data.

This variation can be created by selecting for everyone who when available from the condition dropdown menus and then completing the Otherwise clause.

Limit to purpose policies

Purposes help define the scope and use of data within a project and allow users to meet purpose restrictions on policies. Governors create and manage purposes and their sub-purposes, which project owners then add to their project(s) and use to drive Data Policies.

Purposes can be constructed as a hierarchy, meaning that purposes can contain nested sub-purposes, much like tags in Immuta. This design allows more flexibility in managing purpose-based restriction policies and transparency in the relationships among purposes.

For example, if the purpose Research included Marketing, Product, and Onboarding as sub-purposes, a governor could write the following global policy:

Limit usage to purpose(s) Research for everyone on data sources tagged PHI.

This hierarchy allows you to create this as a single purpose instead of creating separate purposes, which must then each be added to policies as they evolve.

Now, any user acting under the purpose or sub-purpose of Research - whether Research.Marketing or Research.Onboarding - will meet the criteria of this policy. Consequently, purpose hierarchies eliminate the need for a governor to rewrite these global policies when sub-purposes are added or removed. Furthermore, if new projects with new Research purposes are added, for example, the relevant global policy will automatically be enforced.

Refer to the data governor policy guide for a tutorial on purpose-based restrictions on data.

Masking policies

Masking policies hide values in data, providing various levels of utility while still preserving privacy.

Hashing

This policy masks the values with an irreversible sha256 hash, which is consistent for the same value throughout the data source, so you can count or track the specific values, but not know the true raw value.

Hashed values are different across data sources, so you cannot join on hashed values unless you enable masked joins on data sources within a project. Immuta prevents joins on hashed values to protect against link attacks where two data owners may have exposed data with the same masked column (a quasi-identifier), but their data combined by that masked value could result in a sensitive data leak.

Replace with NULL

This policy makes values null, removing any utility of the data the policy applies to.

Replace with constant

With this policy, you can replace the values with the same constant value you choose, such as 'Redacted', removing any utility of that data.

Regular expression (regex)

This policy is similar to replacing with a constant, but it provides more utility because you can retain portions of the true value. When authoring the policy in Immuta, the regex and the replacement value do not need to be in single or double quotes.

The image below illustrates authoring a regex global policy that will apply to Databricks Unity Catalog data sources:

Rounding

This is a technique to hide precision from numeric values while providing more utility than simply hashing. For example, you could remove precision from a geospatial coordinate. You can also use this type of policy to remove precision from dates and times by rounding to the nearest hour, day, month, or year.

With reversibility

This option masks the values using hashing, but allows users to submit an unmasking request to users who meet the exceptions of the policy.

Note: The user receiving the unmasking request must send the unmasked value to the requester.

With Reversible Masking, the raw values are switched out with consistent values to allow analysis without revealing the underlying sensitive data. The direct identifier is replaced with a token that can still be tracked or counted.

With format preserving masking

This option masks the value, but preserves the length and type of the value.

This option also allows users to submit an unmasking request to users who meet the exceptions of the policy.

Preserving the data format is important if the format has some relevance to the analysis at hand. For example, if you need to retain the integer column type or if the first 6 digits of a 12-digit number have an important meaning.

Custom function

This option uses functions native to the underlying database to transform the column. Single quotes enclosing the regex and escaping special characters are required. The following example masks telephone numbers variably depending on the presence of a dash (implying a prefix), space, or only digits:

REGEXP_REPLACE(@column, '(\\+?\\d{0,3}[-\\s]?)?\\d{4}', '****')

The image below illustrates authoring a global policy using this custom function:

Limitations

• The masking functions are executed against the remote database directly. A poorly written function could lead to poor quality results, data leaks, and performance hits.

• Using custom functions can result in changes to the original data type. In order to prevent query errors you must ensure that you cast this result back to the original type.

• The function must be valid for the data type of the selected column. If it is not

  • Local policies will error and show a message that the function is not valid.

  • Global policies will error and change to the default masking type (hashing for text and NULL for all others).

Conditionally masking

For all of the policies above, both at the local and global policy levels, you can conditionally mask the value based on a value in another column. This allows you to build a policy that looks something like: "Mask bank account number where country = 'USA'" instead of blindly stating you want bank account masked always.

Note: When building conditional masking policies with custom SQL statements, avoid using a column that is masked using randomized response in the SQL statement, as this can lead to different behavior depending on your data platform and may produce results that are unexpected.

With k-anonymization

K-anonymity is measured by grouping records in a data source that contain the same values for a common set of quasi identifiers (QIs) - publicly known attributes (such as postal codes, dates of birth, or gender) that are consistently, but ambiguously, associated with an individual.

The k-anonymity of a data source is defined as the number of records within the least populated cohort, which means that the QIs of any single record cannot be distinguished from at least k other records. In this way, a record with QIs cannot be uniquely associated with any one individual in a data source, provided k is greater than 1.

In Immuta, masking with k-anonymization examines pairs of values across columns and hides groups that do not appear at least the specified number of times (k). For example, if one column contains street numbers and another contains street names, the group 123, "Main Street" probably would appear frequently while the group 123, "Diamondback Drive" probably would show up much less. Since the second group appears infrequently, the values could potentially identify someone, so this group would be masked.

After the fingerprint service identifies columns with a low number of distinct values, users will only be able to select those columns when building the policy. Users can either use a minimum group size (k) given by the fingerprint or manually select the value of k.

Note: The default cardinality cutoff for columns to qualify for k-anonymization is 500.

Masking multiple columns with k-anonymization

Governors can write global data policies using k-anonymization in the global data policy builder.

When this global policy is applied to data sources, it will mask all columns matching the specified tag.

If you select multiple columns to mask with k-anonymization in the same policy, the policy is driven by how many times these values appear together. If the groups appear fewer than k times, they will be masked.

For example, if Policy A

Policy A: Mask with k-anonymization the values in the columns gender and state requiring a group size of at least 2 for everyone

was applied to this data source

the values would be masked like this:

Note: Selecting many columns to mask with k-anonymization increases the processing that must occur to calculate the policy, so saving the policy may take time.

However, if you select to mask the same columns with k-anonymization in separate policies, Policy C and Policy D,

Policy C: Mask with k-anonymization the values in the column gender requiring a group size of at least 2 for everyone

Policy D: Mask with k-anonymization the values in the column state requiring a group size of at least 2 for everyone

the values in the columns will be masked separately instead of as groups. Therefore, the values in that same data source would be masked like this:

Using randomized response

This policy masks data by slightly randomizing the values in a column, preserving the utility of the data while preventing outsiders from inferring content of specific records.

For example, if an analyst wanted to publish data from a health survey she conducted, she could remove direct identifiers and apply k-anonymization to indirect identifiers to make it difficult to single out individuals. However, consider these survey participants, a cohort of male welders who share the same zip code:

All members of this cohort have indicated substance abuse, sensitive personal information that could have damaging consequences, and, even though direct identifiers have been removed and k-anonymization has been applied, outsiders could infer substance abuse for an individual if they knew a male welder in this zip code.

In this scenario, using randomized response would change some of the Y's in substance_abuse to N's and vice versa; consequently, outsiders couldn't be sure of the displayed value of substance_abuse given in any individual row, as they wouldn't know which rows had changed.

How the randomization works

Immuta applies a random number generator (RNG) that is seeded with some fixed attributes of the data source, column, backing technology, and the value of the high cardinality column, an approach that simulates cached randomness without having to actually cache anything.

For string data, the random number generator essentially flips a biased coin. If the coin comes up as tails, which it does with the frequency of the replacement rate configured in the policy, then the value is changed to any other possible value in the column, selected uniformly at random from among those values. If the coin comes up as heads, the true value is released.

For numeric data, Immuta uses the RNG to add a random shift from a 0-centered Laplace distribution with the standard deviation specified in the policy configuration. For most purposes, knowing the distribution is not important, but the net effect is that on average the reported values should be the true value plus or minus the specified deviation value.

Preserving data utility

Using randomized response doesn't destroy the data because data is only randomized slightly; aggregate utility can be preserved because analysts know how and what proportion of the values will change. Through this technique, values can be interpreted as hints, signals, or suggestions of the truth, but it is much harder to reason about individual rows.

Additionally, randomized response gives deniability of record content not dataset participation, so individual rows can be displayed.

Mixing masking policies on the same column

In some cases, you may want several different masking policies applied to the same column through Otherwise policies. To build these policies, select everyone who instead of everyone or everyone except. After you specify who the masking policy applies to, select how it applies to everyone else in the Otherwise condition.

You can add and remove tags in Otherwise conditions for global policies (unlike local policy Otherwise conditions), as illustrated above; however, all tags or regular expressions included in the initial everyone who rule must be included in an everyone or everyone except rule in the additional clauses.

Complex data types: masking fields within struct columns (public preview)

Spark supports a class of data types called complex types, which can represent multiple data values in a single column. Immuta supports masking fields within array and struct columns:

• array: an ordered collection of elements

• struct: a collection of elements that are primitive or complex types

Without this feature enabled, the struct and array columns of a data source default to jsonb in the Data Dictionary, and the masking policies that users can apply to jsonb columns are limited. For example, if a user wanted to mask PII inside the column patient in the image below, they would have to apply null masking to the entire column or use a custom function instead of just masking name or address.

After Complex Data Types is enabled on the App Settings page, the column type for struct columns for new data sources will display as struct in the Data Dictionary. (For data sources that are already in Immuta, users can edit the data source and change the column types for the appropriate columns from jsonb to struct.) Once struct fields are available, they can be searched, tagged, and used in masking policies. For example, a user could tag name, ssn, and street as PII instead of the entire patient column.

After a global or local policy masks the columns containing PII, users who do not meet the exception specified in the policy will see these values masked:

Note: Immuta uses the > delimiter to indicate that a field is nested instead of the . delimiter, since field and column names could include ..

Caveats

To get column information about a data source, Immuta executes a DESCRIBE call for the table. In this call, Spark returns a simple string representation of the schema for each column in the table. For the patient column above, the simple string would look like this:

struct<name:string,ssn:string,age:int,address:struct<city:string,state:string,zipCode:string,street:text>>

Immuta then parses this string into the following format for the data source's dictionary:

{
  dataType: 'struct',
  children: [
    {
      name: 'name',
      dataType: 'text'
    },
    {
      name: 'ssn',
      dataType: 'text'
    },
    {
      name: 'age',
      dataType: 'integer'
    },
    {
      name: 'address',
      dataType: 'struct',
      children: [
        {
          name: 'city',
          dataType: 'text'
        },
        {
          name: 'state',
          dataType: 'text'
        },
        {
          name: 'zipCode',
          dataType: 'text'
        },
        {
          name: 'street',
          dataType: 'text'
        },
      ]
    }
  ]
}

However, if the struct contains more than 25 fields, Spark truncates the string, causing the parser to fail and fall back to jsonb. Immuta will attempt to avoid this failure by increasing the number of fields allowed in the server-side property setting, maxToStringFields; however, this only works with clusters on a Spark 3 runtime. The maxToStringFields configuration in Spark 2 cannot be set through the ODBC driver and can only be set through the Spark configuration on the cluster with spark.debug.maxToStringFields 1000 on cluster startup.

Row-level security policies

These policies hide entire rows or objects of data based on the policy being enforced; some of these policies require the data to be tagged as well.

Note: When building row-level policies with custom SQL statements, avoid using a column that is masked using randomized response in the SQL statement, as this can lead to different behavior depending on whether you’re using the Spark or Snowflake and may produce results that are unexpected.

Matching

These policies match a user attribute with a row/object/file attribute to determine if that row/object/file should be visible. This process uses a direct string match, so the user attribute would have to match exactly the data attribute in order to see that row of data.

For example, to restrict access to insurance claims data to the state for which the user's home office is located, you could build a policy such as this:

Only show rows where user possesses an attribute in Office Location that matches the value in the column State for everyone except when user is a member of group Legal.

In this case, the Office Location is retrieved by the identity management system as a user attribute or group. If the user's attribute (Office Location) was Missouri, rows containing the value Missouri in the State column in the data source would be the only rows visible to that user.

WHERE clause policy

This policy can be thought of as a table "view" created automatically for the user based on the condition of the policy. For example, in the policy below, users who are not members of the Admins group will only see taxi rides where passenger_count < 2.

Only show rows where public.us.taxis.passenger_count <2 for everyone except when user is a member of group Admins.

You can put any valid SQL WHERE clause in the policy. See the Custom WHERE clause functions for a list of custom functions.

Time-based restrictions

These policies restrict access to rows/objects/files that fall within the time restrictions set in the policy. If a data source has time-based restriction policies, queries run against the data source by a user will only return rows/blobs with a date in its event-time column/attribute from within a certain range.

The time window is based on the event time you select when creating the data source. This value will come from a date/time column in relational sources.

Minimization

These policies return a limited percentage of the data, which is randomly sampled, at query time. but it is the same sample for all the users. For example, you could limit certain users to only 10% of the data. Immuta uses a hashing policy to return approximately 10% of the data, and the data returned will always be the same; however, the exact number of rows exposed depends on the distribution of high cardinality columns in the database and the hashing type available. Additionally, Immuta will adjust the data exposed when new rows are added or removed.

Masked columns as input for row-level policies

If a global masking policy applies to a column, you can still use that masked column in a global row-level policy.

Consider the following policy examples:

• Masking policy: Mask values in columns tagged Country for everyone except users in group Admin.

• Row-level policy: Only show rows where user possesses an attribute in OfficeLocation that matches the value in column tagged Country for everyone.

Both of these policies use the Country tag to restrict access. Therefore, the masking policy and the row-level policy would apply to data source columns with the tag Country for users who are not in the Admin group.

Limitations

• This feature is only available for Snowflake and Databricks Unity Catalog integrations.

• This feature is only supported for global data policies, not local data policies.

New column added policy

This policy pairs with schema monitoring to mask newly added columns to data sources until data owners review and approve these changes from the requests tab of their profile page.

When this policy is activated by a governor, it will automatically be enforced on data sources that have the New tag applied to them.

To learn how to activate this policy, navigate to the tutorial.

Masking policies hide values in data, providing various levels of utility while still preserving privacy. Immuta offers and .

Column masking

Column masking policies allow you to hide the data in a column. However, there are several different approaches for masking data that allow you to make tradeoffs between privacy (how far you go with masking) vs utility (how much you want the masked data to be useful to the data consumer).

As with all Immuta policy types, it is recommended that you use when to manage policies at scale. When using global policies, tagging your data with metadata becomes critical and is described in detail in the use case.

Types

• Categorical Randomized Response: Categorical values are randomized by replacing a value with some non-zero probability. Not all values are randomized, and the consumer of the data is not told which values are randomized and which ones remain unchanged. Values are replaced by selecting a different value uniformly at random from among all other values. If a randomized response policy were applied to a “state” column, a person’s residency could flip from Maryland to Virginia, which would provide ambiguity to the actual state of residency. This policy is appropriate when obscuring sensitive values such as medical diagnosis or survey responses.

• Custom Function: This function uses SQL functions native to the underlying database to transform the values in a column. This can be used in numerous use cases, but notional examples include top-coding to some upper limit, a custom hash function, and string manipulation.

• K-Anonymization: Masking through k-anonymization is a distinct policy that can operate over multiple attributes. A k-anonymization policy applies rounding and NULL masking policies over multiple columns so that the columns contain at least “K” records, where K is a positive integer. As a result, attributes will only be disclosed when there is a sufficient number of observations. This policy is appropriate to apply over indirect identifiers, such as zip code, gender, or age. Generally, each of these identifiers is not uniquely linked to an individual, but when combined with other identifiers can be associated with a single person. Applying k-anonymization to these attributes provides the anonymity of crowds so that individual rows are made indistinct from each other, reducing the re-identification risk by making it unclear which record corresponds to a specific person. Immuta requires that you . Immuta supports k-anonymization of text, numeric, and time-based data types.

• Mask with Format Preserving Masking: This function masks using a reversible function but does so in a way that the underlying structure of a value is preserved. This means the length and type of a value are maintained. This is appropriate when the masked value should appear in the same format as the underlying value. Examples of this would include social security numbers and credit card numbers where Mask with Format Preserving Masking would return masked values in a format consistent with credit cards or social security numbers, respectively. There is larger overhead with this masking type, and it should really only be used when format is critically valuable, such as situations when an engineer is building an application where downstream systems validate content. In almost all analytical use cases, format should not matter.

• Mask with Reversibility: This function masks in a way that an authorized user can “unmask” a value and reveal the value to an authorized user. Masking with Reversibility is appropriate when there is a need to obscure a value while allowing an authorized user to recover the underlying value. All of the same use cases and caveats that apply to Replace with Hashing apply to this function. Reversibly masked fields can leak the length of their contents, so it is important to consider whether or not this may be an attack vector for applications involving its use.

• Randomized Response: This function randomizes the displayed value to make the true value uncertain, but maintains some analytic utility. The randomization is applied differently to both categorical and quantitative values. In both cases, the noise can be increased to enhance privacy or reduced to preserve more analytic value. Immuta requires that you .

• Datetime and Numeric Randomized Response: Numeric and datetime randomized response apply a tunable, unbiased noise to the nominal value. This noise can obscure the underlying value, but the impact of the noise is reduced in aggregate. This masking type can be applied to sensitive numerical attributes, such as salary, age, or treatment dates. Immuta requires that you .

• Replace with Constant: This function replaces any value in a column with a specified value. The underlying data will appear to be a constant. This masking carries the same privacy and utility guarantees as Replace with NULL. Apply this policy to strings that require a specific repeated value.

• Replace with Hashing: This function masks the values with an irreversible hash, which is consistent for the same value throughout the data source, so you can count or track the specific values, but not know the true raw value. This is appropriate for cases where the underlying value is sensitive, but there is a need to segment the population. Such attributes could be addresses, time segments, or countries. It is important to note that hashing is susceptible to inference attacks based on prior knowledge of the population distribution. For example, if “state” is hashed, and the dataset is a sample across the United States, then an adversary could assume that the most frequently occurring hash value is California. As such, it's most secure to use the hashing mask on attributes that are evenly distributed across a population.

• Replace with Null: This function replaces any value in a column with NULL. This removes any identifiability from the column and removes all utility of the data. Apply this policy to numeric or text attributes that have a high re-identification risk, but little analytic value (names and personal identifiers).

• Replace with REGEX: This function uses a regular expression to replace all or a portion of an attribute. REGEX replacement allows for some groupings to be maintained, while providing greater ambiguity to the disclosed value. This masking technique is useful when the underlying data has some consistent structure, the remasked underlying data represents some re-identification risk, and a regular expression can be used to mask the underlying data to be less identifiable.

• Rounding: Immuta’s rounding policy reduces, rounds, or truncates numeric or datetime values to a fixed precision. This policy is appropriate when it is important to maintain analytic value of a quantity, but not at its native precision.

  • Date/Time Rounding: This policy truncates the precision of a datetime value to a user-defined precision. `minute`, `hour`, `day`, `months`, and `year` are the supported precisions.

  • Numeric Rounding: This policy maps the nominal value to the ceiling of some specified bandwidth. Immuta has a recommended bandwidth based on the Freedman-Diaconis rule.

Cell-level masking

For example, a regular masking policy looks like the following:

Mask columns tagged Discovered.Entity.Social Security Number using hashing for everyone except members of group admins

The cells can be conditionally masked by changing the for to a where:

Mask columns tagged Discovered.Entity.Social Security Number using hashing where country_of_residence = 'US' for everyone except members of group admins

That policy will check the country_of_residence column in the table and if the value is US the cell will be masked, otherwise the data will be presented in the clear as usual.

Mask columns tagged Discovered.Entity.Social Security Number using hashing where @columnTagged('country') = 'US' for everyone except members of group admins

This example policy targets the column with the tag country in the policy logic dynamically.

Masking circumstances

The masking functions described above can be implemented in a variety of use cases. Use the table below to determine the circumstance under which a function should be used.

Circumstance descriptions

• Applicable to Numeric Data: The masking function can be applied to numeric values.

• Column-Value Determinism: Repeated values in the same column are masked with the same output.

• Introduces NULLs: The masking function may, under normal or irregular circumstances, return NULL values.

• Performance: How performant the masking function will be (10/10 being the best).

• Preserves Appearance: The output masked value resembles the valid column values. For example, a masking function would output phone numbers when given phone numbers. Here, NULL values are not counted against this property.

• Preserves Averages: The average of the masked values (avg(mask(v))) will be near the average of the values in the clear (avg(v)).

• Suitable for De-Identification: The masking function can be used to obscure record identifiers, hiding data subject identities and preventing future linking against other identified data.

• Provides Deniability of Record Content: A (possibly identified) person can plausibly attribute the appearance of the value to the masking function. This is a desirable property of masking functions that retain analytic utility, as such functions must necessarily leak information about the original value. Fields masked with these functions provide strong protections against value inference attacks.

• Preserves Equality and Grouping: Each value will be masked to the same value consistently without colliding with others. Therefore, equal values remain equal under masking while unequal values remain unequal, preserving equality. This implies that counting statistics are also preserved.

• Preserves Message Length: The length of the masked value is equal to the length of the original value.

• Preserves Range Statistics: The number of data values falling in a particular range is preserved. For strings, this can be interpreted as the number of strings falling between any two values by alphabetical order.

• Preserves Value Locality: The output will remain near the input, which may be important for analytic purposes.

• Reversible: Qualified individuals can reveal the original input value.

Masking policy support by integration