KR102515790B1 - 6g cellular networks system based on federated learning and method of performing thereof - Google Patents

Abstract

A federated learning-based 6G architecture network system according to an embodiment of the present invention calculates a local learning model using a local data set, and then provides update information for the local learning model to a pre-assigned UAV-based wireless mobile miner module. Upon receiving update information on the local learning model from the user device module and user device module, update information on the local learning model is exchanged and authenticated through communication between other UAV-based wireless mobile miner modules, and then local training is performed according to the authentication result. A data side block including a UAV-based wireless mobile miner module that updates update information on the learning model to blocks on the block chain, a global controller module that controls a global switch installed in the core network module, and a UAV-based wireless mobile miner module When an update to the local learning model is executed by the control side block including the global controller module including the local controller module that controls the installed local switch and the user device module, assigning the user device module and the UAV-based wireless mobile minor module and a management aspect block including a global controller management function module and a local controller management function module providing information about management of a local learning model to the local controller module.

Description

6G architecture network system based on federated learning and its execution method

The present invention relates to a federated learning-based 6G architecture network system and a method for executing the same, and more particularly, to a federated learning-based 6G architecture network system enabling federated learning for different smart applications using a block chain and UAV, and a method thereof It's about how to run it.

6G cellular networks aim to provide new applications by overcoming the inherent limitations of 5G networks. Although 5G cellular networks have been targeted as promising candidates for IoE (IoE).

However, emerging IoE services such as brain-computer interfaces, haptics, flying vehicles, and extended reality (XR) services are disrupting the 5G vision of using short packet-based ultra-reliable low-latency (URLC).

Therefore, it is essential to propose a new architecture for cellular networks capable of supporting new smart services.

Among the many promising features of 6G, the preservation of users' privacy is very important.

To enable 6G services with privacy preservation, it is essential to use federated learning to provide global model training without sending user data to a centralized edge/cloud server.

Federated learning offers significant benefits, but presents several challenges:

First, there is a problem that malicious users can cause errors in the learning process by accessing the centralized server and changing federated learning model parameters.

Second, there is a problem that physical damage to the centralized edge/cloud server also leads to interruption of the learning process.

Third, learning model parameters are vulnerable to security attacks while being transmitted through a wireless channel between a user device module and a centralized edge/cloud server.

Fourth, the number of devices that can impose constraints on computational and communication resources required for federated learning is expected to increase significantly in the near future. Therefore, it is necessary to optimally manage computational and communication resources for federated learning.

To address these challenges, a software-defined networking (SDN) and network function visualization (NFV)-based architecture for 6G networks that enables federated learning-based smart applications is needed.

An object of the present invention is to provide a federated learning-based 6G architecture network system using a block chain for exchanging learning model parameters between user device modules to provide improved security and robustness, and an execution method thereof.

In addition, the present invention aims to provide a federated learning-based 6G architecture network system using a hybrid control plane for SDN that provides scalable operation, which is one of the main goals beyond 6G and networks, and a method for implementing the same.

In addition, an object of the present invention is to provide a federated learning-based 6G architecture network system and an execution method thereof that enable federated learning for different smart applications using a block chain and a UAV.

In addition, the present invention aims to provide a federated learning-based 6G architecture network system and method for implementing the same that enable federated learning over a wireless network for different smart applications in a robust, secure, and scalable manner. .

In addition, the present invention is a federated learning-based 6G architecture network system that enables low-latency operation because the signaling overhead is distributed among multiple controllers compared to a conventional single SDN controller when multiple controllers are used in a hybrid SDN controller, and a method for executing the same is intended to provide

In addition, an object of the present invention is to provide a federated learning-based 6G architecture network system and an execution method thereof that allow a user to adaptively add a new resource management system using a management reference plane.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention not mentioned above can be understood by the following description and will be more clearly understood by the examples of the present invention. It will also be readily apparent that the objects and advantages of the present invention may be realized by means of the instrumentalities and combinations indicated in the claims.

A federated learning-based 6G architecture network system for achieving this purpose is a user device that calculates a local learning model using a local data set and then provides update information for the local learning model to a pre-assigned UAV-based wireless mobile miner module. Upon receiving update information on the local learning model from the module and the user device module, update information on the local learning model is exchanged and authenticated through communication between other UAV-based wireless mobile miner modules, and then the local learning model according to the authentication result A data-side block including a UAV-based wireless mobile miner module that updates update information on a block on a block chain, a global controller module that controls a global switch installed in the core network module, and a local installed in the UAV-based wireless mobile miner module. A control aspect block including a global controller module including a local controller module that controls the switch and a global controller that allocates the user device module and the UAV-based wireless mobile minor module when an update to the local learning model is executed by the user device module and a management aspect block including a management function module and a local controller management function module providing information about management of the local learning model to the local controller module.

In addition, a method of executing a federated learning-based 6G architecture network system for achieving this purpose includes calculating a local learning model by a user device module using a local data set and receiving a request for updating the local learning model, the user device Allocating a module and a UAV-based wireless mobile miner module, providing update information on the local learning model updated by the user device module to the UAV-based wireless mobile miner module, Upon receiving update information on the local learning model from the device module, exchanging and authenticating update information on the local learning model through communication between different UAV-based wireless mobile miner modules, wherein the UAV-based wireless mobile minor module After exchanging and authenticating update information on the learning model, updating the update information on the local learning model to a block on the block chain, and after the user device module receives the block on the block chain, through the block It includes performing model aggregation and checking global federated learning model accuracy.

According to the present invention as described above, there is an advantage of using a block chain for exchanging learning model parameters between user device modules to provide improved security and robustness.

In addition, according to the present invention, it is possible to use a hybrid control plane for SDN that provides scalable operation, which is one of the main goals beyond 6G and networks.

In addition, according to the present invention, there is an advantage that federated learning for different smart applications is possible using a block chain and a UAV.

In addition, according to the present invention, there is an advantage of enabling federated learning over a wireless network for different smart applications in a robust, secure and scalable manner.

In addition, according to the present invention, when multiple controllers are used in a hybrid SDN controller, compared to a conventional single SDN controller, signal overhead is distributed among multiple controllers, thereby enabling low-latency operation.

In addition, the present invention has an advantage in that a user can adaptively add a new resource management system using the management reference plane.

1 is a diagram for explaining a 6G architecture network system based on federated learning according to an embodiment of the present invention.
2 is a flowchart illustrating an execution process of a 6G architecture network system based on federated learning according to the present invention.

The above objects, features and advantages will be described later in detail with reference to the accompanying drawings, and accordingly, those skilled in the art to which the present invention belongs will be able to easily implement the technical spirit of the present invention. In describing the present invention, if it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.

The present invention relates to a Software Defined Networking (SDN) and Network Functions Visualization (NFV) based architecture for 6G that enables federated learning for different smart applications using blockchain and UAVs.

Software-Defined Networking (SDN) provides separation of the control plane from the data plane to enable efficient network management, while Network Functions Visualization (NFV) can use generic hardware to implement various network functions using virtual machines.

Federated learning preserves user privacy as it provides training on a global federated learning model without migrating data from user devices to centralized edge/cloud servers.

Besides that, SDN and NFV-based architectures that use UAVs and blockchain to enable federated learning provide the salient features of distributed learning with privacy, enhanced security, robustness, and efficient management.

Also, the federated learning based 6G architecture network system according to the present invention can be used in 6G and telecommunication networks to enable other smart applications.

Common federated learning uses centralized edge/cloud servers for global model aggregation. In contrast to existing federated learning models, the federated learning-based 6G architecture network system according to the present invention enables federated learning without using a centralized edge/cloud server and thus provides robustness.

That is, the UAV-based wireless mobile miner according to the present invention can train a global federated learning model in a secure and distributed manner without using a centralized edge/cloud server.

In addition, the federated learning-based 6G architecture network system according to the present invention can be used to train a federated learning model used for UAVs to provide on-demand BS services.

1 is a diagram for explaining a 6G architecture network system based on federated learning according to an embodiment of the present invention.

Referring to FIG. 1 , the federated learning-based 6G architecture network system includes a management side block 100, a control side block 200 and a data side block 300.

The management aspect block 100 is used to implement management functions.

The management aspect block 100 includes a local controller management function module 110 and a global controller management function module 120 . The local controller management function module 110 and the global controller management function module 120 communicate with the control side block 200 through the south interface module.

The local controller management function module 110 controls functions that require local network knowledge rather than a global network view of the entire network.

The local controller management function module 110 through the southern interface module to instruct the local controller modules 220_1 to 200_N to enable various tasks such as local learning model management, edge server management, and global model integration of user device modules. Communicates with the control aspect block (200).

The global controller management function module 120 communicates with the control side block 200 through the south interface module like the local controller management function module 110.

Several functions directed by the global controller management function module 120 are user device modules (320_1 to 320_N) based on user-defined criteria, radio access network resource management, core network management, overall quality of service (QoS) management, and UAV-based wireless mobile Allocate minor modules (310_1 to 310_N).

The south interface module is used for communication between the control side block (200) and the data side block (300).

The northbound interface module is used for communication between the control side block 200 and the management side block 100.

The control side block 200 includes global controller modules 210_1 to 210_N and local controller modules 220_1 to 200_N. Global controller modules (210_1 to 210_N) and local controller modules (220_1 to 200_N) interact with local and global switches to efficiently control a 6G network capable of federated learning.

The global controller modules 210_1 to 210_N are located in the control module and control network entity involved in performing the management function of the global controller management function module 120, and control the global switch installed in the core network module 330. At this time, the global switch can control several local switches and communicate with each other.

There are two methods for controlling the functions of these global controller modules (210_1 to 210_N). That is, the global controller modules 210_1 to 210_N control global control switches disposed in the core network module and control core network components supporting the global control controller.

Enabling core network components with global controller functionality increases complexity compared to a separate global network switch. However, a separate global network switch costs more. Thus, a trade-off must be made between system complexity and cost.

For this reason, the present invention controls the local switch installed in the UAV-based wireless mobile minor modules (310_1 to 310_N) and the global switch installed in the core network module 330, respectively, in contrast to the existing SDN control surface. It has the advantage of reducing overhead.

The local controller modules 220_1 to 200_N control local switches installed in the UAV-based wireless mobile miner modules 310_1 to 310_N.

The data aspect block 300 includes several modules responsible for actual physical communication between different modules. The data side block 300 includes UAV-based wireless mobile miner modules 310_1 to 310_N, user device modules 320_1 to 320_N, a core network module 330, and a cloud module 340.

UAV-based wireless mobile miner modules (310_1 to 310_N) are composed of unmanned aerial vehicles with SDN functions activated and sufficient computing power and backup power.

The UAV-based wireless mobile miner modules 310_1 to 310_N communicate with the user device modules 320_1 to 320_N, check update information on the local learning model, and exchange with other UAV-based wireless mobile miner modules 310_1 to 310_N.

That is, when the UAV-based wireless mobile miner modules 310_1 to 310_N receive update information on the local learning model from the user device modules 320_1 to 320_N, communication between other UAV-based wireless mobile miner modules 310_1 to 310_N is established. This is to prove through the process of exchanging update information for the local learning model.

Then, the UAV-based wireless mobile miner modules (310_1 to 310_N) communicate with other UAV-based wireless mobile miner modules (310_1 to 310_N) to exchange and prove the update information for the local learning model, and then to the block on the blockchain. Updates the update information for the local learning model.

In this way, the UAV-based wireless mobile miner modules (310_1 to 310_N) update the update information on the local learning model in the block on the block chain, so that the block has update information on the local learning model of all user device modules (320_1 to 320_N). will be recorded

Therefore, later, after the user device modules 320_1 to 320_N receive blocks on the block chain, a global federation model aggregation that aggregates update information on the local learning models of all user device modules 320_1 to 320_N recorded in the block is aggregated. that can be run

The user device modules 320_1 to 320_N are based on SDN functions and implement local learning model management.

These user device modules 320_1 to 320_N calculate a local learning model using a local data set, and then provide update information on the local learning model to pre-allocated UAV-based wireless mobile miner modules 310_1 to 310_N.

To this end, the user device modules 320_1 to 320_N provide an update request signal for the local learning model to the global controller management function module 120, so that the global controller management function module 120 uses user-defined criteria, radio access network resources Based on management, core network management, and overall quality of service (QoS) management, user device modules (320_1 to 320_N) and UAV-based wireless mobile minor modules (310_1 to 310_N) can be allocated.

As described above, after the user device modules 320_1 to 320_N and the UAV-based wireless mobile minor modules 310_1 to 310_N are allocated by the global controller management function module 120, the user device modules 320_1 to 320_N perform local learning Update information on the model is provided to the UAV-based wireless mobile miner modules 310_1 to 310_N pre-allocated by the global controller management function module 120.

Then, after the update information for the local learning model is authenticated between the UAV-based wireless mobile miner module (310_1 to 310_N) and other UAV-based wireless mobile miner modules, the update information for the local learning model is updated in the block on the block chain. If so, the user device modules 320_1 to 320_N perform global model aggregation and check global federated learning model accuracy.

Therefore, the global federation model aggregation receives blocks for updates to the local learning models of all user device modules (320_1 to 320_N) from the UAV-based wireless mobile miner modules (310_1 to 310_N), and then the user device modules (320_1 to 320_N) runs on

If, the user device modules 320_1 to 320_N stop the federated learning process when the desired global federated learning model accuracy is achieved. That is, the global federation model aggregation is executed in the user device modules 320_1 to 320_N after receiving a block for updating the local learning model of all devices from the UAV-based wireless mobile miner modules 310_1 to 310_N.

The core network module 330 is based on SDN functions and is used for management of core network functions during federated learning.

The cloud module 340 is based on SDN functionality and is used to provide on-demand computing and storage capabilities.

2 is a flowchart illustrating an execution process of a 6G architecture network system based on federated learning according to the present invention.

Referring to FIG. 2, the user device modules 320_1 to 320_N calculate a local learning model using a local data set (step S210), and then send update information about the local learning model to a pre-assigned UAV-based wireless mobile miner module ( 310_1 to 310_N) (step S220).

To this end, the user device modules 320_1 to 320_N and the UAV-based wireless mobile miner modules 310_1 to 310_N are allocated according to the update request for the local learning model by the user device modules 320_1 to 320_N.

At this time, the user device modules 320_1 to 320_N and the UAV-based wireless mobile minor modules 310_1 to 310_N are allocated based on user-defined criteria, radio access network resource management, core network management, and overall quality of service (QoS) management. .

Therefore, the user device modules 320_1 to 320_N calculate a local learning model using the local data set, and then provide update information on the local learning model to the pre-assigned UAV-based wireless mobile miner modules 310_1 to 310_N. It can.

When the UAV-based wireless mobile miner modules 310_1 to 310_N receive update information on the local learning model from the user device modules 320_1 to 320_N, the local UAV-based wireless mobile miner modules 310_1 to 310_N communicate with each other. Update information on the learning model is exchanged and verified (step S230).

UAV-based wireless mobile miner modules (310_1 to 310_N) communicate with other UAV-based wireless mobile miner modules (310_1 to 310_N) to exchange and prove update information on the local learning model, and then transfer the local learning model to a block on the blockchain. Updates update information for (step S240).

The user device modules 320_1 to 320_N perform global model aggregation and check global federated learning model accuracy. That is, the user device modules 320_1 to 320_N may perform global model aggregation and check global federated learning model accuracy when update information on the local learning model is updated in a block on the block chain.

Therefore, the global federation model aggregation receives blocks for updates to the local learning models of all user device modules (320_1 to 320_N) from the UAV-based wireless mobile miner modules (310_1 to 310_N), and then the user device modules (320_1 to 320_N) runs on

If, the user device modules 320_1 to 320_N stop the federated learning process when the desired global federated learning model accuracy is achieved.

That is, the global federation model aggregation is executed in the user device modules 320_1 to 320_N after receiving a block for updating the local learning model of all devices from the UAV-based wireless mobile miner modules 310_1 to 310_N.

Although described by the limited embodiments and drawings, the present invention is not limited to the above embodiments, and those skilled in the art can make various modifications and variations from these descriptions. Therefore, the spirit of the present invention should be grasped only by the claims described below, and all equivalent or equivalent modifications thereof will be said to belong to the scope of the spirit of the present invention.

100: management aspect block;
110: local controller management function module;
120: global controller management function module;
200: control side block,
210: global controller module,
220: local controller module,
300: data side block,
310: UAV-based wireless mobile miner module,
320: local controller module,
330: core network module,
340: cloud module

Claims (5)

A user device module that calculates a local learning model using a local data set and then provides update information for the local learning model to a pre-assigned UAV-based wireless mobile miner module, and update information for the local learning model from the user device module. Upon receiving, UAV-based that exchanges and authenticates update information on the local learning model through communication between other UAV-based wireless mobile miner modules, and then updates the update information on the local learning model to a block on the block chain according to the authentication result. a data side block containing a wireless mobile miner module;
A control side block including a global controller module for controlling the global switch installed in the core network module and a local controller module for controlling the local switch installed in the UAV-based wireless mobile minor module; and
A global controller management function module that allocates a user device module and a UAV-based wireless mobile minor module when an update to the local learning model is executed by the user device module and provides information about management of the local learning model to the local controller module Characterized in that it comprises a management aspect block comprising a local controller management function module
Federated learning based 6G architecture network system.
According to claim 1,
The user device module
Characterized in that after receiving a block on the block chain, global model aggregation is performed through the block and global federated learning model accuracy is checked.
Federated learning based 6G architecture network system.
According to claim 1,
The global controller management function module
Allocating the user device module and the UAV-based wireless mobile minor module based on user-defined criteria, radio access network resource management, core network management, and overall quality of service (QoS) management
Federated learning based 6G architecture network system.
calculating, by a user device module, a local trained model using a local data set, and receiving an update request for the local trained model;
Allocating the user device module and the UAV-based wireless mobile minor module;
providing update information on the local learning model updated by the user device module to a UAV-based wireless mobile miner module;
When the UAV-based wireless mobile miner module receives update information on a local learning model from a user device module, exchanging and authenticating update information on a local learning model through communication between other UAV-based wireless mobile minor modules;
After the UAV-based wireless mobile miner module exchanges and authenticates update information for the local learning model, updating the update information for the local learning model to a block on a block chain;
Characterized in that, after the user device module receives a block on the block chain, performing global model aggregation through the block and checking global federation learning model accuracy.
Implementation method of federated learning based 6G architecture network system.
According to claim 4,
Allocating the user device module and the UAV-based wireless mobile minor module
Allocating the user device module and the UAV-based wireless mobile minor module based on user-defined criteria, radio access network resource management, core network management, and overall quality of service (QoS) management.
Implementation method of federated learning based 6G architecture network system.

Images