API Reference

Tumult Analytics is a differentially private analytics library from Tumult Labs.

The library is broken up into a number of modules that enable complex differentially private queries to be defined and run. A typical workflow is for users to instantiate PySpark and read in their data; create a Session object with private datasets, their corresponding privacy protection, and a privacy budget; define differentially private queries using a QueryBuilder object, and then evaluating these queries with evaluate().

For specifying privacy guarantees:

• session defines the Session, an interactive interface for evaluating differentially private queries.

• privacy_budget contains types for representing privacy budgets.

• protected_change contains types for expressing what changes in input tables are protected by differential privacy.

For defining queries:

• query_builder provides an interface for constructing differentially private queries from basic query operations.

• keyset defines KeySet, a type for specifying the keys used in group-by operations.

• constraints contains types for representing constraints used when evaluating queries on data with the AddRowsWithID protected change.

• truncation_strategy and binning_spec provide types that are used to define certain types of queries.

Modules

binning_spec	A BinningSpec defines a binning operation on a column.
config	Configuration for Tumult Analytics.
constraints	Defines Constraint types.
keyset	A KeySet specifies a list of values for one or more columns.
privacy_budget	Classes for specifying privacy budgets.
protected_change	Types for programmatically specifying what changes in input tables are protected.
query_builder	An API for building differentially private queries from basic operations.
session	Interactive query evaluation using a differential privacy framework.
truncation_strategy	Defines strategies for performing truncation in private joins.
utils	Utility functions.

Exceptions

exception AnalyticsInternalError(message)

Bases: AssertionError

Generic error to raise for internal analytics errors.

Parameters:

message (str)

This diagram shows the basic workflow for most Analytics operations. (Click to see full-size image.)