hash

hash replaces each entity with a SHA hash of its value. Because the hash is deterministic for the same value and operator params, it is useful for analytics joins on de-identified data while reducing direct exposure of the original value.

Parameters

Parameter	Type	Default	Description
`hash_type`	STRING	`"sha256"`	Hash algorithm: `"sha256"` or `"sha512"`.
`salt`	STRING	`""`	Optional salt prepended to each value before hashing (`hash(salt + value)`). Use a secret, stable salt to reduce rainbow-table and common-value matching risk while staying deterministic for a given salt + value pair.
`entity_collapse`	BOOLEAN	`false`	Operator-level opt-in for full identity collapse. When request-level `entity_collapse` is enabled, aliases in the same collapsed group share one hash.

Example

SELECT app_public.mask(
  'Contact: john.smith@acme.com',
  OBJECT_CONSTRUCT(
    'entity_config', OBJECT_CONSTRUCT(
      'EMAIL_ADDRESS', OBJECT_CONSTRUCT(
        'operator', 'hash',
        'operator_params', OBJECT_CONSTRUCT('hash_type', 'sha256')
      )
    )
  )
);
-- Contact: 83f1c2b4e0a9d8f7...  (64-character SHA-256 hex digest)

Example — salted

OBJECT_CONSTRUCT(
  'entity_config', OBJECT_CONSTRUCT(
    'EMAIL_ADDRESS', OBJECT_CONSTRUCT(
      'operator', 'hash',
      'operator_params', OBJECT_CONSTRUCT(
        'hash_type', 'sha256',
        'salt', 'my-dataset-salt-2024'
      )
    )
  )
)

Example — collapse aliases before hashing

By default, hash normalizes small formatting differences before hashing. To hash aliases such as "Sarah Chen", "Chen", and "Dr. Chen" to the same digest, enable request-level entity_collapse and opt the operator into full collapse:

OBJECT_CONSTRUCT(
  'entity_collapse', 0.5,
  'entity_config', OBJECT_CONSTRUCT(
    'PERSON', OBJECT_CONSTRUCT(
      'operator', 'hash',
      'operator_params', OBJECT_CONSTRUCT(
        'hash_type', 'sha256',
        'entity_collapse', TRUE
      )
    )
  )
)

When to use

Analytics on de-identified data: count distinct users, join across datasets, and aggregate by customer with less direct exposure of raw identifiers.
Cross-dataset linking: the same email gets the same hash when the value, hash type, salt, and collapse behavior match, enabling joins without sharing the raw value.
Pseudonymous identifiers: include salted hashes in controlled operational datasets when support or debugging workflows need a stable reference to the same entity.

Security notes

SHA hashing is not reversible through Agent Mask. However, deterministic hashes can still be linkable and guessable when the source value is common or known.
Hashes are pseudonymous identifiers, not anonymous data. They may remain sensitive under your organization’s policies or applicable privacy regulations, especially when they can be joined back to people or records.
If an attacker can guess the original value, such as a common email address, they can hash the guess and compare it. Hash is not protection against a known-plaintext attack by itself.
Use salt to reduce rainbow-table and common-value matching risk. Keep the salt secret, restrict access to it, and keep it consistent only across datasets that intentionally need joinability.
If operator_params.entity_collapse=true, all aliases in a collapsed group hash from the canonical value chosen by the collapse step.
For reversible re-identification, use encrypt instead.