Method and system for determining hidden states of a machine using privacy-preserving distributed data analytics and a semi-trusted server and a third-party

• US 9,471,810 B2
• Filed: 03/09/2015
• Issued: 10/18/2016
• Est. Priority Date: 03/09/2015
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Wang, Ye
• Primary Examiner(s): Nalven, Andrew
• Assistant Examiner(s): Li, Mary
• Application Number:
  US14/641,522
• Publication Number:
  US 20160267297A1
• Time in Patent Office: 589 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06F 21/6245 Protecting personal data, e.g. for financial or medical purposes
• CPC Class Codes: G06F 21/6263 during internet communication, e.g. revealing personal data from cookies
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A method classifies data to determine hidden states of a machine by first acquiring data from the machine in a client, which is permuting randomly, and then chaff is inserted before transmitting to server as private data. The server classifies the private data according to a hidden Markov model to obtain permuted noisy estimates of states of the machine and the chaff, which are transmitted to a third party. The third party removes the chaff and inverts noisy estimates using a permutation ordering to obtain unpermuted noisy estimates of the states of the machine.

Claims

1. A method for classifying data to determine hidden states of a machine, comprising:

acquiring, by a client, data from the machine, wherein the data include samples;

permuting randomly the data, according to a permutation, to generate permuted data;

inserting, by the client, chaff in the permuted data at locations to generate private data;

transmitting, by the client to the server, the private data;

transmitting, by the client to the third-party the locations of the chaff and a permutation ordering;

classifying, by the server, each sample of the private data independently according to a hidden Markov model (HMM) to obtain permuted noisy estimates of states of the machine and the chaff;

transmitting, by the server to the third-party, the permuted noisy estimates of the states and the chaff;

removing, by the third-party, the chaff using the locations;

inverting, by the third-party after removing the chaff, the permuted noisy estimates using the permutation ordering to obtain unpermuted noisy estimates of the states of the machine.

2. The method of claim 1, further comprising:

correcting errors, by the third-party, to obtain estimates of the hidden states.

3. The method of claim 1, wherein the server has either partial knowledge of a statistical distribution of the private data given each possible state, or a trained classifier that estimates the states, and the client has partial knowledge of a state transition distribution.

4. The method of claim 1, wherein parameters of the HMM include distributions of state transition statistics and statistics of the private data.

5. The method of claim 2, wherein the correcting uses temporal correlations to obtain the estimates of the hidden states.

6. The method of claim 1, wherein the client, the server and the third-party are connected by a communications network.

7. The method of claim 1, wherein information and computational resource capability between the three parties are asymmetric.

8. The method of claim 1, wherein the third-party applies a Viterbi reconstruction procedure during the classifying.