US 11,895,034 B1Granted

Training and implementing a machine learning model to selectively restrict access to traffic

AssigneeJoinesty, Inc.

InventorsRobert Jeffrey Yoskowitz, Stephen Michael Yoskowitz, Elder Donizetti Dos Santos

Life of this patent

Estimating the term…

Loading patent…

Abstract

The disclosed systems and techniques enable selectively restricting access of an end-user device to traffic. An example method may include training a machine learning (ML) model using a plurality of instances of historical traffic to classify fields included in the plurality of instances into a plurality of categories of information. The method may further include implementing the ML model by: (i) receiving traffic from a server, the traffic responsive to a request from an end-user device; (ii) identifying a field included in the traffic that corresponds to a category of the plurality of categories by applying the ML model to the traffic; (iii) in response to identifying that the field corresponds to the category, selecting a rule corresponding to the category; (iv) modifying the traffic in accordance with the rule to restrict access to information in the field; and (v) transmitting the modified traffic to the end-user device.

Claims16

Dependent claims

A method for selectively restricting access to traffic, the method comprising:

A) acquiring a plurality of instances of historical traffic (i) originating from one or more servers each implementing a different service, and (ii) destined for one or more end-user devices, each instance of historical traffic including one or more fields;
B) training a machine learning (ML) model using the plurality of instances to classify fields included in the plurality of instances of historical traffic into a plurality of categories of information;
and
C) implementing the ML model, including:
(i) receiving traffic from a server implementing a service, the traffic responsive to a request from an end-user device;
(ii) identifying a field included in the traffic that corresponds to a category of the plurality of categories of information by applying the ML model to the traffic;
(iii) in response to identifying that the field corresponds to the category, selecting a rule corresponding to the category;
(iv) modifying the traffic in accordance with the rule to restrict access to information in the field without restricting access to at least a portion of the traffic;
and
(v) transmitting the modified traffic to the end-user device;
wherein acquiring the plurality of instances of historical traffic include:
acquiring a first set of instances of historical traffic originating from a first server, the first set of instances having a first format;
and
acquiring a second set of instances of historical traffic originating from a second server, the second set of instances having a second format;
and
wherein training the ML model includes:
training a first version of the ML model using the first set of instances to classify fields included in the first set of instances into the plurality of categories;
and
training a second version of the ML model using the second set of instances to classify fields included in the second set of instances into the plurality of categories;
and
wherein applying the ML model includes:
determining a format of the traffic;
responsive to the format being the first format, applying the first version of the ML model;
and
responsive to the format being the second format, applying the second version of the ML model.

22 more elements

A method for selectively restricting access to traffic, the method comprising:

A) acquiring a plurality of instances of historical traffic (i) originating from one or more servers each implementing a different service, and (ii) destined for one or more end-user devices, each instance of historical traffic including one or more fields;
B) training a machine learning (ML) model using the plurality of instances to classify fields included in the plurality of instances of historical traffic into a plurality of categories of information;
and
C) implementing the ML model, including:
(i) receiving traffic from a server implementing a service, the traffic responsive to a request from an end-user device;
(ii) identifying a field included in the traffic that corresponds to a category of the plurality of categories of information by applying the ML model to the traffic;
(iii) in response to identifying that the field corresponds to the category, selecting a rule corresponding to the category;
(iv) modifying the traffic in accordance with the rule to restrict access to information in the field without restricting access to at least a portion of the traffic;
and
(v) transmitting the modified traffic to the end-user device;
wherein acquiring the plurality of instances of historical traffic include:
acquiring a first set of instances of historical traffic originating from a first server, the first set of instances having a first format;
and
acquiring a second set of instances of historical traffic originating from a second server, the second set of instances having a second format;
and
wherein training the ML model further includes:
training the ML model to identify the first set of instances having the first format and the second set of instances having the second format;
training the ML model to classify fields included in the first set of instances into the plurality of categories;
and
training the ML model to classify fields included in the second set of instances into the plurality of categories.

17 more elements

A system for selectively restricting access to traffic, the system comprising:

a first one or more processors;
a training engine configured to execute on the first one or more processors, the training engine further configured to generate a machine learning (ML) model by:
A) acquiring a plurality of instances of historical traffic (i) originating from one or more servers each implementing a different service, and (ii) destined for one or more end-user devices, each instance of historical traffic including one or more fields;
and
B) training the ML model using the plurality of instances to classify fields included in the plurality of instances into a plurality of categories of information;
a second one or more processors configured to receive the ML model from the first one or more processors via a communication network;
and
an obfuscation engine configured to execute on the second one or more processors, the obfuscation engine further configured to implement the ML model by:
(i) receiving traffic from a server implementing a service, the traffic responsive to a request from an end-user device;
(ii) identifying a field included in the traffic that corresponds to a category of the plurality of categories of information by applying the ML model to the traffic;
(iii) in response to identifying that the field corresponds to the category, selecting a rule corresponding to the category;
(iv) modifying the traffic in accordance with the rule to restrict access to information in the field without restricting access to at least a portion of the traffic;
and
(v) transmitting the modified traffic to the end-user device;
wherein the training engine is configured to acquire the plurality of instances of historical traffic by:
acquiring a first set of instances of historical traffic originating from a first server, the first set of instances having a first format;
and
acquiring a second set of instances of historical traffic originating from a second server, the second set of instances having a second format;
and
wherein the training engine is further configured to train the ML model by:
training the ML model to identify the first set of instances having the first format and the second set of instances having the second format;
training the ML model to classify fields included in the first set of instances into the plurality of categories;
and
training the ML model to classify fields included in the second set of instances into the plurality of categories.

21 more elements

A system for selectively restricting access to traffic, the system comprising:

a first one or more processors;
a training engine configured to execute on the first one or more processors, the training engine further configured to generate a machine learning (ML) model by:
A) acquiring a plurality of instances of historical traffic (i) originating from one or more servers each implementing a different service, and (ii) destined for one or more end-user devices, each instance of historical traffic including one or more fields;
and
B) training the ML model using the plurality of instances to classify fields included in the plurality of instances into a plurality of categories of information;
a second one or more processors configured to receive the ML model from the first one or more processors via a communication network;
and
an obfuscation engine configured to execute on the second one or more processors, the obfuscation engine further configured to implement the ML model by:
(i) receiving traffic from a server implementing a service, the traffic responsive to a request from an end-user device;
(ii) identifying a field included in the traffic that corresponds to a category of the plurality of categories of information by applying the ML model to the traffic;
(iii) in response to identifying that the field corresponds to the category, selecting a rule corresponding to the category;
(iv) modifying the traffic in accordance with the rule to restrict access to information in the field without restricting access to at least a portion of the traffic;
and
(v) transmitting the modified traffic to the end-user device;
wherein the training engine is configured to acquire the plurality of instances of historical traffic by:
acquiring a first set of instances of historical traffic originating from a first server, the first set of instances having a first format;
and
acquiring a second set of instances of historical traffic originating from a second server, the second set of instances having a second format;
and
wherein:
the training engine is configured to train the ML model by:
training a first version of the ML model using the first set of instances to classify fields included in the first set of instances into the plurality of categories;
and
training a second version of the ML model using the second set of instances to classify fields included in the second set of instances into the plurality of categories;
and
the obfuscation is configured to apply the ML model by:
determining a format of the traffic;
responsive to the format being the first format, applying the first version of the ML model;
and
responsive to the format being the second format, applying the second version of the ML model.

27 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/333,450Filed 28 May 2021 · us provisional application
Priority: 29 Jan 2021
Granted: 6 Feb 2024
Term ends: Not estimatedThe source records don’t support an estimate.

Classification14 codes

H04L 47/22
Flow control; Congestion control: traffic shaping
H04L 41/16
Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks: using machine learning…
H04L 43/028
Capturing of monitoring data: by filtering
H04L 47/125
Avoiding congestion; Recovering from congestion: by balancing the load, e.g. traffic engineering