US 11,687,769 B2Granted

Advanced techniques for machine learning using sample comparisons

AssigneePaypal, Inc.

InventorsDavid Tolpin, Benjamin Hillel Myara, Michael Dymshits

Life of this patent

Estimating the term…

Loading patent…

Abstract

Machine learning techniques can be used to train a classifier, in some embodiments, to accurately detect similarities between different records of user activity for a same user. When more recent data is received, newer data can be analyzed by selectively removing particular sub-groups of data to see if there is any particular data that accounts for a large difference (e.g. when run through a classifier that has been trained to produce similar results for known activity data from a same user). If a sub-group of data is identified as being significantly different from other user data, this may indicate an account breach. Advanced machine learning techniques described herein may be applicable to a variety of different environments.

Claims20

Dependent claims

A method, comprising:

training, by one or more hardware processors, a classifier comprising a neural network using records associated with transactions conducted by different users of an electronic service provider (ESP), wherein the classifier is trained to minimize outcome differences between samples from the records associated with a same user;
accessing, by the one or more hardware processors, a first set of records associated with a first plurality of transactions conducted through a particular user account with the ESP during a first time period;
generating, by the one or more hardware processors, a plurality of modified record sets based on selectively removing different subsets of records from the first set of records;
determining, by the one or more hardware processors using the trained classifier, a plurality of outcomes for the plurality of modified record sets;
determining, by the one or more hardware processors, differences between each outcome in the plurality of outcomes and a particular outcome corresponding to a second set of records associated with a second plurality of transactions conducted through the particular user account during a second time period;
and
determining whether the particular user account appears to have been used by different users based on the differences.

4 more elements

A non-transitory computer-readable medium having stored thereon machine-readable instructions executable to cause a computer system to perform operations comprising:

training a classifier comprising a neural network using records associated with transactions conducted by different users of an electronic service provider (ESP), wherein the classifier is trained to minimize outcome differences between samples of the records associated with a same user;
accessing a first set of records associated with a first plurality of transactions conducted through a particular user account with the ESP during a first time period;
generating a plurality of modified record sets based on selectively removing different subsets of records from the first set of records;
calculating, using the trained classifier and for each modified record set in the plurality of modified record set, a degree of similarity between the modified record set and a second set of records associated with a second plurality of transactions conducted through the particular user account during a second time period;
and
determining whether the particular user account appears to have been used by an unauthorized user based on a deviation between a first degree of similarity calculated for a first modified record set in the plurality of modified record set and a second degree of similarity calculated for a second modified record set in the plurality of modified record set.

3 more elements

A system, comprising:

a non-transitory memory;
and
one or more hardware processors coupled to the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising:
accessing a classifier comprising a neural network using records associated with different users, wherein the classifier is trained to minimize outcome differences between samples from the records that are associated with a same user;
accessing a first set of records associated with a first plurality of transactions conducted through a particular user account with an electronic service provider (ESP) during a first time period;
generating a plurality of modified record sets based on selectively removing different subsets of records from the first set of records;
determining, using the classifier and for each modified record set in the plurality of modified record sets, a degree of similarity between the modified record set and a second set of records associated with a second plurality of transactions conducted through the particular user account during a second time period;
and
determining whether the particular user account appears to have been used by different users based on a deviation between a first degree of similarity determined for a first modified record set in the plurality of modified record sets and a second degree of similarity determined for a second modified record set in the plurality of modified record sets.

6 more elements

Dependencies are inferred from the claim text.