Cryptographically secure machine learning

Abstract

Embodiments are directed towards classifying data. A machine learning (ML) engine may select an ML model that may employ a cryptographic multi-party computation (MPC) protocol based on model preferences, including a parameter model, provided by a client. A randomness engine may be employed to provide random values and other random values based on the MPC protocol such that the random values may be provided to the client and the other random values may be provided to an answer engine. Input values that correspond to fields in the parameter model may be provided by the client such that the input values may be based on the MPC protocol and the random values. The answer engine may be employed to provide partial results to the question based on the ML model, the input values, and the MPC protocol that may be provided to the client.

What is claimed as new and desired to be protected by Letters Patent of the United States is: 
     
         1 . A method for classifying data over a network using one or more processors, included in one or more network computers, to perform actions, comprising:
 employing a machine learning (ML) engine to perform actions, including:
 selecting an ML model that employs a cryptographic multi-party computation (MPC) protocol based on model preferences provided by a client, wherein the provided model preferences include both a question and a parameter model, and wherein the parameter model includes one or more model objects of the ML model, and wherein the ML engine uses the parameter model to define one or more input values that are compatible with the ML model; 
   employing a randomness engine to perform actions, including:
 providing one or more random values and one or more other random values based on the cryptographic MPC protocol, wherein the one or more random values are provided to the client and the one or more other random values are provided to an answer engine; 
 distributing a first instance of the randomness engine and a first random information datastore to the client, wherein the one or more random values are provided from the first random information datastore; and 
 distributing a second instance of the randomness engine and a second random information datastore to the answer engine, wherein the one or more other random values are provided from the second random information datastore; and 
   employing the answer engine to perform further actions, including:
 synchronizing the first random information datastore and second random information datastore to maintain a correlation between the one or more random values and the one or more other random values; 
 receiving, from the client, a data model having model objects that include the one or more input values that correspond to one or more fields of the one or more model objects in the parameter model, wherein the one or more input values are based on the cryptographic MPC protocol and the one or more random values; 
 determining compliance of the data model with one or more requirements of the ML model based on a comparison of the data model to the parameter model; 
 in response to the data model complying with the one or more requirements of the ML model, providing one or more partial results to the question based on the ML model, the one or more input values, and the cryptographic MPC protocol; and 
 providing the one or more partial results to the client, wherein a ML client engine provides one or more answers to the question based on the one or more partial results.