Skip to content

Policy Handler HTTP API

Audience: Data Owners and Data Governors

Content Summary: This page describes how to update policies using the Policy Handler API.

Create Policy Handler

Method Path Successful Status Code
POST /policy/handler 200

Request Parameters

The create policy handler endpoint must be a policy handler object.

Update Policy Handler

Method Path Successful Status Code
PUT /policy/handler 200

Request Parameters

The update policy handler endpoint must be a policy handler object.

Policy Handler Object

  • dataSourceId (integer): ID of the data source the policy will be applied to.
    • Example: 1
  • jsonRules (array[object]): Array of JSON rules objects.

Defining Policy Rules

The jsonRules array contains rules objects. The following types of policy rules are supported:

Not all combination of policy rules are valid. The examples below are supported policy rule combinations:

  • Prerequisite, Visibility, Masking
  • Prerequisite, Masking, Minimization
  • Differential Privacy

Prerequisite Policy Rule Type

Prerequisite policies are used to limit usage to one or more purposes.

  • type (string): Policy rule type. Must be prerequisite for prerequisite policy rules.
    • Example: "prerequisite"
  • operator (string): Operator to be applied on conditions. Possible values: and, or.
    • Example: "or"
  • conditions (array[object]): Conditions to be applied for the rule. Multiple values will be evaluated according to the operator.

Example:

In this example, users will only have access to data from this data source when they are acting under the purpose named Purpose Name.

{
    "type": "prerequisite",
    "operator": "or",
    "conditions": [{
        "type": "purposes",
        "value": "Purpose Name"
    }]
}

Visibility Policy Rule Type

Visibility policies are used to enforce row-level security.

  • type (string): Policy rule type. Must be visibility for row-level security policy rules.
    • Example: "visibility"
  • operator (string): Operator to be applied on conditions. Possible values: and, or.
    • Example: "or"
  • conditions (array[object]): Conditions to be applied for the rule. Multiple values will be evaluated according to the operator.

Note: When adding conditions to a visibility policy rule, the field is required, and the condition value should be left empty. For example, for a group policy condition, the group name is not specified.

{
    "type": "groups",
    "field": "organization",
    "group": {
        "iam": "active_directory"
    }
}

The user must possess the group, authorization, or purpose that matches the value stored in the field.

Example:

In this example, users will only see rows when they have an authorization that matches the value in the field department and they belong to a group that matches the value in the field organization.

{
    "type": "visibility",
    "operator": "and",
    "conditions": [{
        "type": "authorizations",
        "field": "department",
        "authorization": {
            "auth": "accesses",
            "iam": "active_directory"
        }
    }, {
        "type": "groups",
        "field": "organization",
        "group": {
            "iam": "active_directory"
        }
    }]
}

Masking Policy Rule Type

Masking policy rules will mask the value in one or more columns.

  • type (string): Policy rule type. Must be masking for masking policy rules.
    • Example: "masking"
  • fields (array[string]): Fields that will be masked when a user does not fulfill policy conditions.
    • Example: ["email", "locastion"]
  • operator (string): Operator to be applied on conditions. Possible values: and, or.
    • Example: "or"
  • conditions (array[object]): Conditions to be applied for the rule. Multiple values will be evaluated according to the operator.

Note: When adding conditions to a masking policy rule, the field will be left blank, and the condition value should be populated.

{
    "type": "groups",
    "group": {
        "name": "users",
        "iam": "active_directory"
    }

When using a masking rule, there is an additional field that needs to be sent in the update data source request in the policyHandler.maskingConfiguration array field.

  • name (string): Name of the field being masked.
    • Example: "social"
  • type (string): Type of masking to apply. Supported values are "Consistent Value", "Grouping", "Regular Expression"
    • Example: "Consistent Value"
  • metadata (object): Extra metadata used when masking the value.

Masking Configuration Metadata

Consistent Value
  • constant (string|null): Constant value to mask to. If this field is not defined, the value will be hashed.
    • Example: "REDACTED"
Regular Expression
  • regex (string): Regex to match against when masking columns.
    • Example: "[0-9]{3}-[0-9]{2}"
  • replacement (string): String used to replace the matched regex.
    • Example: "xxx-xx"
Grouping
  • bucketSize (integer): For number fields. Size of buckets to round numbers to.
    • Example: 100
  • timePrecision (string): For time fields. Time precision to round to. Possible values: "MIN", "HOUR", "DAY", "WEEK", "MONTH", "YEAR"
    • Example: "HOUR"

Example Policy Handler Update with Masking Configuration Metadata:

{
    "policyHandler": {
        "maskingConfiguration": [{
            "type": "Consistent Value",
            "name": "age",
            "metadata": {
                "constant": "REDACTED"
            }
        }]
    }
}

Example:

In this example, the fields email and location will be masked unless the user belongs to the group admins.

{
    "type":"masking",
    "fields": ["email", "location"],
    "operator":"or",
    "conditions":[{
        "type":"groups",
        "group":{
            "name":"admins",
            "iam":"active_directory"
        }
    }]
}

Differential Privacy Rule Type

Differential privacy rules will make the data source differentially private by adding noise to queries based on a high cardinality column for everyone except when the user fulfills the policy conditions. See the Differential Privacy Policy Guide for further information.

  • type (string): Policy rule type. Must be additional for minimization policy rules.
    • Example: "additional"
  • name (string): Name of additional policy. Must be differentialPrivacy for differential privacy policy rules.
    • Example: "differentialPrivacy"
  • operator (string): Operator to be applied on conditions. Possible values: and, or.
    • Example: "or"
  • conditions (array[object]): Conditions to be applied for the rule. Multiple values will be evaluated according to the operator.
{
    "type": "groups",
    "group": {
        "name": "users",
        "iam": "active_directory"
    }
}

When using a differential privacy rule, there is an additional field that needs to be sent in the update data source request in the policyHandler.additionalFilters.differentialPrivacy field.

  • epsilon (number): Epsilon value to use for the differential privacy rule
    • Example: 3
  • delta (number): Delta value to use for the differential privacy rule
    • Example: 0.001
  • samples (integer): The number of buckets used to randomly partition the data during sample and aggregate.
    • Example: 99
  • hashPhrase (string): : Column to base differential privacy on. This should be a high cardinality column in the data source.
    • Example: "name"

Example Policy Handler Rule:

In this example, the data source will be made (\varepsilon, \delta)-differentially private, with \varepsilon=3, \delta=0.001, and random partitioning based on the name field into 99 buckets. It will not apply if the users belong to the group admins.

{
    "type": "additional",
    "name": "differentialPrivacy",
    "operator": "or",
    "conditions": {
        "type":"groups",
        "group":{
            "name":"admins",
            "iam":"active_directory"
        }
    }
}

Example Data Source Update (partial):

{
    "policyHandler": {
        "additionalFilters": {
            "differentialPrivacy": {
                "epsilon": 3,
                "delta": 0.001,
                "samples": 99,
                "hashPhrase": "name"
            }
        }
    }
}

Minimization Rule Type

Minimization policy rules will show a limited percentage of the data, based on a high cardinality column, for everyone unless the user fulfills the policy conditions.

  • type (string): Policy rule type. Must be additional for minimization policy rules.
    • Example: "additional"
  • name (string): Name of additional policy. Must be minimization for minimization policy rules.
    • Example: "minimization"
  • operator (string): Operator to be applied on conditions. Possible values: and, or.
    • Example: "or"
  • conditions (array[object]): Conditions to be applied for the rule. Multiple values will be evaluated according to the operator.

Note: When adding conditions to a minimization policy rule the field will be left blank.

{
    "type": "groups",
    "group": {
        "name": "users",
        "iam": "active_directory"
    }

When using a minimization rule, there is an additional field that needs to be sent in the update data source request in the policyHandler.additionalFilters.minimization field.

  • percent (integer): Percentage of the data to show to the users. This percentage will be based off of unique values in the hashPhrase column.
    • Example: 50
  • hashPhrase (string): Column to base the percentage off of. This should be a high cardinality column in the data source.
    • Example: "name"

Example Policy Handler Rule:

In this example, 50 percent of the data, based on the name field, will be visible to users unless they fulfill the policy conditions.

{
    "type": "additional",
    "name": "minimization",
    "operator": "or",
    "conditions": {
        "type":"groups",
        "group":{
            "name":"admins",
            "iam":"active_directory"
        }
    }
}

Example Data Source Update (partial):

{
    "policyHandler": {
        "additionalFilters": {
            "minimization": {
                "percent": 50,
                "hashPhrase": "name"
            }
        }
    }
}

Time-Based Rule Type

Time-based rules will make a limited portion of the data available based on event time. The data source must contain an event time column in order for this policy type to be valid. For instance, users who do not fulfill the policy conditions will only see data from within the defined time window.

  • type (string): Policy rule type. Must be additional for minimization policy rules.
    • Example: "additional"
  • name (string): Name of additional policy. Must be time for time based policy rules.
    • Example: "time"
  • operator (string): Operator to be applied on conditions. Possible values: and, or.
    • Example: "or"
  • conditions (array[object]): Conditions to be applied for the rule. Multiple values will be evaluated according to the operator.

Note: When adding conditions to a time based policy rule the field will be left blank.

{
    "type": "groups",
    "group": {
        "name": "users",
        "iam": "active_directory"
    }

When using a time based rule, there is an additional field that needs to be sent in the update data source request in the policyHandler.additionalFilters field.

  • time (integer): Age in seconds of the oldest data a user will be allowed to see. This counts backward from the present.
    • Example: 14400

Example Policy Handler Rule:

In this example, only data from the last 4 hours will be visible to users unless they fulfill the policy conditions.

{
    "type": "additional",
    "name": "time",
    "operator": "or",
    "conditions": {
        "type":"groups",
        "group":{
            "name":"admins",
            "iam":"active_directory"
        }
    }
}

Example Data Source Update (partial):

{
    "policyHandler": {
        "additionalFilters": {
            "time": 14400
        }
    }
}

Policy Conditions

There are three types of policy conditions:

  1. Groups
  2. Authorizations
  3. Purposes

Group Policy Condition

The group policy condition restricts access to the condition when a user is a member of a group.

  • type (string): Type of policy condition. Must be "groups" for the group policy condition.
    • Example: "groups"
  • group (object): Object describing group user must belong to in order to satisfy the policy condition.
  • field (string): Data field to match group name against when checking policy.
    • Example: "department"

Example:

{
    "type": "groups",
    "group": {
        "name": "users",
        "iam": "active_directory"
    },
    "field": "department"
}

Group Object

  • name (string): Name of group user must belong to in order to satisfy the policy condition.
    • Example: "users"
  • iam (string): ID of the IAM containing the group.
    • Example: "active_directory"

Authorization Policy Condition

The authorization policy condition restricts access to the condition when a user possesses an authorization.

  • type (string): Type of policy condition. Must be "authorizations" for the authorization policy condition.
    • Example: "authorizations"
  • authorization (object): Object describing authorization user must possess in order to satisfy the policy condition.
  • field (string): Data field to match authorization value against when checking policy.
    • Example: "department"

Example:

{
    "type": "authorizations",
    "authorization": {
        "auth": "accesses",
        "value": "PII",
        "iam": "active_directory"
    },
    "field": "department"
}

Authorization Object

  • auth (string): Name of authorization to check for authorization value.
    • Example: "accesses"
  • value (string): Value of authorization user must possess in order to satisfy the policy condition.
    • Example: "PII"
  • iam (string): ID of the IAM user must receive the authorization from.
    • Example: "active_directory"

Purpose Policy Condition

The purpose policy condition restricts access to the condition when a user is acting under a purpose.

  • type (string): Type of policy condition. Must be "purposes" for the purpose policy condition.
    • Example: "purpopses"
  • value (string): Purpose name user must be acting under in order to satisfy the policy condition.
    • Example: "Purpose"
  • field (string): Data field to match purpose name against when checking policy.
    • Example: "department"

Example:

{
    "type": "purposes",
    "value": "Purpose",
    "field": "department"
}