IP Library Granted Patent US 11,423,157
Granted Patent B2
US 11,423,157 · App. 16/864,826 · Granted Aug 23, 2022

Adversarial reinforcement learning system for simulating security checkpoint environments

Inventors: Brian Jacob Lewis (Washington, DC); Jason Adam Deich (Arlington, VA); Stephen John Melsom (Reston, VA); Kara Jean Dodenhoff (Reston, VA); William Tyler Niggel (Wilmington, NC)
Assignee: NOBLIS, INC.
G06F21/577G06F21/566G06F2221/034
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 11,423,157
App. No.
16/864,826
Granted
Aug 23, 2022
Kind
B2
Abstract

An adversarial reinforcement learning system is used to simulate a security checkpoint. The system includes a simulation engine configured to simulate a security checkpoint and various threat objects and threat-mitigation objects therein. The system further includes an attack model configured to control threat objects in the simulation and a defense model configured to control threat-mitigation objects in the simulation. A first portion of the simulation is executed by the simulation engine in order to generate an outcome of the first portion of the simulation. The defense model then generates a threat-mitigation input to control threat-mitigation objects in a subsequent portion of the simulation, and the attack model then generates a threat input to control threat objects in the subsequent portion of the simulation, wherein the inputs are based in part on the outcome of the first portion of the simulation.

Claims (41)

1. A method for simulating a security checkpoint, the method performed by an adversarial reinforcement learning system comprising one or more processors, the method comprising:

executing a first portion of a simulation of a security checkpoint to generate an outcome of the first portion of the simulation;

generating, by a first model, a threat mitigation input, wherein the threat mitigation input comprises instructions for controlling one or more of a plurality of simulated threat-mitigation objects in a subsequent portion of the simulation, wherein the threat mitigation input is generated based at least in part on the outcome of the first portion; and

generating, by a second model, a threat input, wherein the threat input comprises instructions for controlling one or more of a plurality of simulated threat objects in the subsequent portion of the simulation, wherein the threat input is generated based at least in part on the outcome of the first portion; and

executing the subsequent portion of the simulation based at least in part of the threat mitigation input and the threat input.

2. The method of claim 1 , comprising:

receiving, by the first model, a first set of information regarding the simulated security checkpoint, wherein generating the threat mitigation inputs is based on the first set of information; and

receiving, by the second model, a second set of information regarding the simulated security checkpoint, wherein generating the threat inputs is based on the second set of information.

3. The method of claim 1 , wherein a threat-mitigation object of the plurality of threat-mitigation objects is simulated in accordance with an associated effectiveness at mitigating an outcome associated with one or more threat objects of the plurality of threat objects.

4. The method of claim 3 , wherein the effectiveness comprises a likelihood of detecting the one or more threat objects.

5. The method of claim 3 , wherein the effectiveness comprises a capability to decrease severity of the outcome associated with the one or more threat objects.

6. The method of claim 1 , wherein a first threat object of the plurality of threat objects is simulated in accordance with an outcome associated with the first threat object.

7. The method of claim 6 , wherein the outcome associated with the first threat object is based on a location of the first threat object within the simulated security checkpoint.

8. The method of claim 6 , wherein the outcome associated with the first threat object is associated with a combination of the first threat object and a second threat object.

9. The method of claim 1 , wherein:

a threat object of the plurality of threat objects is simulated in accordance with synthetic data generated by a synthetic data generator;

a threat-mitigation object of the plurality of threat-mitigation objects is simulated in accordance with an automated threat-detection algorithm; and

executing the simulation comprises applying the automated threat-detection algorithm to the synthetic data.

10. The method of claim 1 , comprising automatically receiving environment data regarding the security checkpoint to be simulated, wherein executing the simulation of the security checkpoint is based on the environment data received.

11. The method of claim 1 , comprising automatically receiving object data regarding one or more of the threat-mitigation objects or one or more of the threat objects to be simulated, wherein executing the simulation of the security checkpoint is based on the object data received.

12. The method of claim 1 , wherein generating the threat mitigation input is performed in accordance with a first set of asset limitations that limits the threat-mitigation objects that may be deployed in the simulation.

13. The method of claim 1 , wherein generating the threat input is performed in accordance with a second set of asset limitations that limits the threat objects that may be deployed in the simulation.

14. The method of claim 1 , further comprising ceasing executing one or more of the portions of the simulated environment in accordance with determining that one or more stability criteria have been satisfied.

15. The method of claim 1 , further comprising generating output data representing a configuration of the plurality of threat mitigations represented by the threat mitigation inputs for one of the portions of the simulation.

16. The method of claim 1 , further comprising generating output data representing a plurality of threats represented by the threat inputs for one of the portions of the simulation.

17. The method of claim 1 , wherein:

executing the first portion of the simulation comprises executing a first time-step of the simulation; and

executing the subsequent portion of the simulation comprises executing a subsequent time-step of the simulation occurring, in the simulated environment, after the first time-step of the simulation.

18. The method of claim 1 , wherein

executing the first portion of the simulation comprises executing a first iteration of the simulation; and

executing the subsequent portion of the simulation comprises executing a subsequent iteration of the simulation executed after the first iteration of the simulation.

19. An adversarial reinforcement learning system for simulating a security checkpoint, the system comprising one or more hardware processors configured to:

execute a first portion of a simulation of a security checkpoint to generate an outcome of the first portion of the simulation;

generate, by a first model, a threat mitigation input, wherein the threat mitigation input comprises instructions for controlling one or more of a plurality of simulated threat-mitigation objects in a subsequent portion of the simulation, wherein the threat mitigation input is generated based at least in part on the outcome of the first portion; and

generate, by a second model, a threat input, wherein the threat input comprises instructions for controlling one or more of a plurality of simulated threat objects in the subsequent portion of the simulation, wherein the threat input is generated based at least in part on the outcome of the first portion; and

execute the subsequent portion of the simulation based at least in part of the threat mitigation input and the threat input.

20. A non-transitory computer-readable storage medium storing instructions configured to be executed by one or more processors of an adversarial reinforcement learning system the instructions configured to cause the system to:

execute a first portion of a simulation of a security checkpoint to generate an outcome of the first portion of the simulation;

generate, by a first model, a threat mitigation input, wherein the threat mitigation input comprises instructions for controlling one or more of a plurality of simulated threat-mitigation objects in a subsequent portion of the simulation, wherein the threat mitigation input is generated based at least in part on the outcome of the first portion; and

generate, by a second model, a threat input, wherein the threat input comprises instructions for controlling one or more of a plurality of simulated threat objects in the subsequent portion of the simulation, wherein the threat input is generated based at least in part on the outcome of the first portion; and

execute the subsequent portion of the simulation based at least in part of the threat mitigation input and the threat input.

Assignments (2)
SECURITY INTEREST Recorded May 27, 2025
From: NOBLIS, INC.
To: PNC BANK, NATIONAL ASSOCIATION
Reel/Frame 071415/0887 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 20, 2020
From: LEWIS, BRIAN JACOB; DEICH, JASON ADAM; MELSOM, STEPHEN JOHN; DODENHOFF, KARA JEAN; NIGGEL, WILLIAM TYLER
To: NOBLIS, INC.
Reel/Frame 052712/0714 →
Continuity (2)
Provisional Application 62847592 · May 14, 2019
Related Publication 20200364347A1 · Nov 19, 2020
Cited By (2)
US 12,499,241 US 12,517,812