System and method for managing security models through scenario generation and evaluation

Patent metadata

Field	Value
Publication number	US-12556587-B2
Application number	US-202318309273-A
Country	US
Kind code	B2
Filing date	Apr 28, 2023
Priority date	Apr 28, 2023
Publication date	Feb 17, 2026
Grant date	Feb 17, 2026

Abstract

Methods and systems for securing deployments are disclosed. The deployments may be secured by generating and deploying security models to components of the deployment. The security models may be obtained through simulation of the operation of the deployment. During the simulation, predictions of different types of attacks and the potential defenses to the attacks on its operation may be evaluated. Further, limits may be imposed on the different attacks and potential defenses to simulate various scenarios that may be encountered in real systems.

First claim

What is claimed is:

1 . A method for securing a deployment, the method comprising:
obtaining a digital twin model for the deployment, the digital twin model being adapted to replicate operation of the deployment in a digital environment; obtaining a first inference model adapted to select first parameters to disrupt operation of the digital twin model; obtaining a second inference model adapted to select second parameters to prevent the disruption of the operation of the digital twin model; obtaining a third inference model adapted to generate third parameters to regulate interaction between the digital twin model, the first inference model and the second inference model, the third parameters being generated as a scenario that the third inference model causes all of the digital twin model, the first inference model, and the second inference model to follow, the scenario setting limitations on actions that can be performed by the digital twin model, the first inference model, and the second inference model; obtaining a security model for the deployment using the first inference model, the second inference model, the third inference model, and the digital twin model; and deploying the security model to the deployment to secure the deployment.

2 . The method of claim 1 , wherein obtaining the third inference model comprises:
identifying a type of inference model to select the third parameters; and generating an instance of the type of the inference model.

3 . The method of claim 2 , wherein identifying the type of inference model comprises:
identifying a third set of manipulable operations of the digital twin model; identifying third operating metrics of the digital twin model that are monitorable; and using the third set of manipulable operations and the third operating metrics to discriminate the type of the inference model from other types of inference models.

4 . The method of claim 1 , wherein the third inference model is based on a type of inference model, and the type of the inference model is based on:
a third set of manipulable operations of the digital twin model, and third operating metrics of the digital twin model that are monitorable.

5 . The method of claim 1 , wherein obtaining the security model for deployment comprises:
performing iterative computations with the first inference model and the second inference model and third inference model to obtain a set of security models; ranking the security models based on performance criteria to obtain a rank ordering; and selecting the security model for deployment based on the rank ordering.

6 . The method of claim 5 , wherein performing iterative computations with the first, second, and third inference models to obtain the set of security models comprises:
running first training cycles with the first inference model with the digital twin model; running second training cycles with the second inference model with the digital twin model; running third training cycles with the third inference model with the digital twin model, the first inference model and the second inference model; and producing a security model optimized for performance through the first training cycles, the second training cycles, and the third training cycles.

7 . The method of claim 6 , wherein running the third training cycles comprises:
selecting a scenario using the third inference model; selecting the first parameters using the first inference model and the scenario; selecting the second parameters using the second inference model and the scenario; running the digital twin model using the scenario, the first parameters and the second parameters to identify an outcome; predicting the outcome using the third inference model; and updating operation of the third inference model based on an uncertainty level for the predicted outcome.

8 . The method of claim 7 , wherein selecting the first parameters comprises:
discriminating a portion of manipulable operations of the digital twin model based on the scenario; and setting the first parameters for the discriminated portion of manipulable operations.

9 . The method of claim 7 , wherein the outcome for the scenario indicates whether operation of the digital twin model during the running of the digital twin model was protected by the second parameters.

10 . The method of claim 7 , wherein updating operation of the third inference model comprises:
performing a reinforcement learning cycle based on the uncertainty level to incentivize generation of new scenarios for which predictions of corresponding outcomes have higher degrees of uncertainty.

11 . The method of claim 1 , further comprising:
obtaining a fourth inference model adapted to predict a result of interactions between the digital twin model, the first inference model, and the second inference model within the scenario, wherein the security model is further obtained using the fourth inference model.

12 . The method of claim 1 , wherein the scenario is generated by the third inference model based on a reward obtained by the third inference model for previous ones of the scenario generated by the third inference model.

13 . A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations for securing a deployment, the operations comprising:
obtaining a digital twin model for the deployment, the digital twin model being adapted to replicate operation of the deployment in a digital environment; obtaining a first inference model adapted to select first parameters to disrupt operation of the digital twin model; obtaining a second inference model adapted to select second parameters to prevent the disruption of the operation of the digital twin model; obtaining a third inference model adapted to generate third parameters to regulate interaction between the digital twin model, the first inference model and the second inference model, the third parameters being generated as a scenario that the third inference model causes all of the digital twin model, the first inference model, and the second inference model to follow, the scenario setting limitations on actions that can be performed by the digital twin model, the first inference model, and the second inference model; obtaining a security model for the deployment using the first inference model, the second inference model, the third inference model, and the digital twin model; and deploying the security model to the deployment to secure the deployment.

14 . The non-transitory machine-readable medium of claim 13 , wherein obtaining the third inference model comprises:
identifying a type of inference model to select the third parameters; and generating an instance of the type of the inference model.

15 . The non-transitory machine-readable medium of claim 14 , wherein identifying the type of inference model comprises:
identifying a third set of manipulable operations of the digital twin model; identifying third operating metrics of the digital twin model that are monitorable; and using the third set of manipulable operations and the third operating metrics to discriminate the type of the inference model from other types of inference models.

16 . The non-transitory machine-readable medium of claim 13 , wherein the third inference model is based on a type of inference model, and the type of the inference model is based on:
a third set of manipulable operations of the digital twin model, and third operating metrics of the digital twin model that ar

Assignees

• Dell Products Lp

Inventors

• Ezrielev Ofir
• Kushnir Tomer
• Balin Maxim

Classifications

• H04L63/205 Primary

  involving negotiation or determination of the one or more network security mechanisms to be used, e.g. by negotiation between the client and the server or between peers or by selection according to the capabilities of the entities involved (negotiation of communication capabilities H04L69/24) · CPC title