KR102465550B1 - Uav controlling system for defens against gps spoofing and method thereof - Google Patents

Abstract

The present invention relates to a UAV control system for preventing GPS spoofing and a method thereof. The UAV control system for preventing GPS spoofing comprises: a plurality of IoT devices located on the ground and transmitting data including monitoring information and GPS information sensed at set periodic intervals at set time intervals; a plurality of UVAs flying individually along a flight path, calculating a distance difference from the corresponding IoT device using the data transmitted from the plurality of IoT devices, storing the calculated distance difference and the strength of a transmission signal of the corresponding data in a data set, determining whether or not a GPS spoofing attack is present using the data set, and generating a learning model that predicts a destination by learning the data set; and a control server for transmitting an integrated learning model integrated and generated by using a parameter set of learning models generated in each UAV to each of the plurality of UAVs. Accordingly, UAV resources can be used efficiently.

Description

UAV CONTROLLING SYSTEM FOR DEFENS AGAINST GPS SPOOFING AND METHOD THEREOF

The present invention relates to a UAV control system and method for preventing GPS spoofing, and more particularly, to detect whether or not a GPS spoofing attack of UAVs (Unmanned Aerial Vehicles) is present, and to use Federated Learning to deviate from a UAV's path. It relates to a UAV control system for preventing GPS spoofing and a method therefor.

Unmanned Aerial Vehicles (UAVs) are aircraft that fly automatically without user intervention, and have the advantage of being easy to deploy due to their high mobility. Therefore, it is recently used in various fields such as transportation, disaster monitoring, leisure, reconnaissance, and exploration covering a wide area. In particular, when the communication demand is less than the operating area of the system, communication using a UAV is more cost effective than installing a base station. is being done

Such a disaster monitoring system is a system that can prevent a disaster before it occurs by measuring environmental information such as water depth or ground temperature using a small Internet of Things (IoT) sensor. This is because the monitoring area is wide, but the device that collects information on the ground is located locally, and it is difficult to transmit data in real time because the sensor device takes time to measure the environment. It is more effective to collect local data.

Therefore, in the disaster monitoring system, the UAV periodically performs a mission to collect data transmitted by the IoT sensor through flight.

Since the UAV flies along the route according to the GPS location information, it is necessary to be able to check the GPS location in real time in order for the UAV to fly on the normal route. However, if the GPS information of the current location calculated by the UAV is incorrect, control over the flight path may be lost, resulting in inability to complete the mission. Due to these vulnerabilities, UAVs can be easily exposed to sabotage attacks on GPS radio waves.

In particular, a GPS spoofing attack in which an attacker manipulates the actual GPS value and transmits a disturbed signal may occur. Accidents caused by falls Therefore, a reliable technique to detect and respond to GPS spoofing attacks must be applied.

However, UAVs have limitations on storage devices and computing resources due to physical limitations. If the calculation process is complicated in the method of defending against an attack, there is a problem in that it is difficult to detect it within an appropriate time to respond to the attack.

Therefore, it is necessary to develop a technology capable of responding to GPS spoofing attacks and simplifying the calculation process so that UAV resources can be used efficiently.

The technology that is the background of the present invention is disclosed in Republic of Korea Patent Publication No. 10-2163698 (2020. 10. 08. Announcement).

The technical problem to be achieved by the present invention is to provide a UAV control system and a method for preventing GPS spoofing, in which the UAV detects whether a GPS spoofing attack is performed using the strength of a signal and GPS information transmitted by an IoT device located on the ground. will be.

Another object of the present invention is to provide a UAV control system and method for preventing GPS spoofing that prevent the attacked UAV from deviate from a path and move to an actual location using federated learning when detecting a GPS spoofing attack.

In order to achieve this technical problem, the UAV control system for preventing GPS spoofing according to an embodiment of the present invention is a plurality of transmitting data including monitoring information and GPS information sensed at a set period interval and located on the ground at a set time interval. IoT devices; It is provided in plurality and each flies according to the flight path, calculates the distance difference with the corresponding IoT device using the data transmitted from the plurality of IoT devices, and calculates the calculated distance difference and the strength of the transmitted signal of the data as data a UAV that stores in a set, determines whether a GPS spoofing attack is performed using the data set, and generates a learning model for predicting a destination by learning the data set; and a control server that transmits the federated learning model integratedly generated by using the parameter set of the learning models generated in each UAV to the plurality of UAVs, respectively.

At this time, if it is determined that the UAV has been subjected to a GPS spoofing attack, it may move to a destination location predicted by the federated learning model.

In addition, the UAV determines whether the GPS spoofing attack occurs using the distance to the IoT device that has transmitted data and the strength of the transmitted signal of the data, but the current GPS location value and the location value included in the GPS information of the data The distance difference with the corresponding IoT device is calculated using the count, and the number of IoT devices for which the strength of the transmitted signal of the data is determined to be outside the error range for the strength of the normal signal corresponding to the calculated distance difference is counted, If the number of IoT devices out of the above error range is more than half of the number of IoT devices that have transmitted data to the corresponding UAV, it may be determined that the corresponding UAV has been subjected to a GPS spoofing attack.

In addition, when the reception of the data transmitted from the IoT device is completed, the UAV calculates the task performance rate calculated using the size of data transmitted when the IoT device starts communication and the size of the actually received data in the data set. more can be stored in

In addition, the learning model can learn to predict a destination with a maximum data collection rate by iteratively calculating a local weight set capable of minimizing a loss function using the task performance rate stored in the data set.

In addition, the control server may generate the federated learning model by calculating global weights by the following equation using a local weight set that is a learning model parameter set of each UAV.

Here, GW is a global weight, u is a UAV, n and N are the number of UAVs, and LW _u is a set of local weights.

In addition, the UAV can fly along a normal flight path by resetting the current location to the destination location predicted by the federated learning model after moving to the destination location.

In addition, in the UAV control method for preventing GPS spoofing according to another embodiment of the present invention, data including monitoring information and GPS information sensed by a plurality of IoT devices located on the ground at a set period interval and data including GPS information are transmitted to the UAV at a set time interval. sending each to A plurality of UAVs, each flying according to a flight path, calculate a distance difference with a corresponding IoT device using the data transmitted from the plurality of IoT devices, and calculate the distance between the calculated distance difference and the strength of a transmission signal of the corresponding data. storing in a data set; generating, by the UAV, a learning model for predicting a destination by learning the data set; transmitting, by a control server, a federated learning model integratedly generated using a parameter set of learning models generated in each UAV to the plurality of UAVs, respectively; determining, by the UAV, whether a GPS spoofing attack is performed using the data set; and if it is determined that the UAV has been subjected to a GPS spoofing attack as a result of the determination, moving to a destination location predicted by the federated learning model.

In addition, the step of determining whether the GPS spoofing attack is performed is to determine whether the GPS spoofing attack has occurred using the distance to the IoT device that has transmitted the data and the strength of the transmitted signal of the data, but the current GPS location value and the GPS of the data IoT in which the distance difference with the corresponding IoT device is calculated using the location value included in the information, and the intensity of the transmitted signal of the data is determined to be outside the error range for the intensity of the normal signal corresponding to the calculated distance difference If the number of devices is counted, and the number of IoT devices out of the error range is more than half of the number of IoT devices that have transmitted data to the corresponding UAV, it may be determined that the corresponding UAV has been subjected to a GPS spoofing attack.

In addition, in the storing in the data set, when the reception of the data transmitted from the IoT device is completed, the task calculated using the size of the data transmitted when the IoT device starts communication and the size of the actually received data is performed. rate may be further stored in the data set.

In addition, the step of generating the learning model can be generated by learning to predict the destination where the data collection rate is the maximum by iteratively calculating the local weight set that can minimize the loss function using the task performance rate stored in the data set. can

In addition, the control server may generate the federated learning model by calculating global weights by the following equation using a local weight set that is a learning model parameter set of each UAV.

Here, GW is a global weight, u is a UAV, n and N are the number of UAVs, and LW _u is a set of local weights.

In addition, after the UAV moves to the destination location, resetting the current location to the destination location predicted by the federated learning model may further include the step of flying along a normal flight path.

As described above, according to the present invention, the UAV detects whether a GPS spoofing attack exists by calculating the difference between the signal strength according to the distance transmitted by the IoT device located on the ground and the position according to the GPS signal. Since the process can be simplified, there is an effect that UAV resources can be used efficiently.

In addition, according to the present invention, when a GPS spoofing attack is detected, the UAV's mission performance rate can be improved by calculating the actual location and resetting the UAV's route so that the UAV's disaster monitoring data collection mission can be smoothly performed. .

In addition, according to the present invention, each UAV learns as a learning technique for calculating the actual location during a GPS spoofing attack, and then learns the learning contents for each intermediate base station by applying the federated learning that exchanges the learning contents at the management institution located at the departure and arrival points. Compared to the exchange method, communication cost can be reduced, and UAV resources with limited storage space can be efficiently used, and accuracy is improved to prevent the UAV from straying from the path.

In addition, according to the present invention, it is possible to solve the security problem occurring in the UAV-based system, so that a more stable system can be built, so it is expected that it can be applied to various UAV-based systems as well as the disaster and disaster monitoring system. .

1 is a block diagram illustrating a UAV control system for preventing GPS spoofing according to an embodiment of the present invention.
2 is a diagram illustrating a UAV control system for preventing GPS spoofing according to an embodiment of the present invention.
3 is a flowchart illustrating an operation flow of a UAV control method for preventing GPS spoofing according to an embodiment of the present invention.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In this process, the thickness of the lines or the size of the components shown in the drawings may be exaggerated for clarity and convenience of explanation.

In addition, the terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to intentions or customs of users and operators. Therefore, definitions of these terms should be made based on the content throughout this specification.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

First, a UAV control system for preventing GPS spoofing according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2 .

1 is a block diagram illustrating a UAV control system for preventing GPS spoofing according to an embodiment of the present invention.

As shown in FIG. 1 , the UAV control system for preventing GPS spoofing according to an embodiment of the present invention includes an IoT device 100 , a UAV 200 , and a control server 300 .

First, the IoT device 100 is provided with a plurality of sensors 110 that can send and receive data in real time, and is located on the ground and transmits data including monitoring information and GPS information sensed at a set interval interval at a set time interval. to the UAV 200 .

In this case, the IoT device 100 may be a user's tablet, smart phone, smart watch, notebook, etc., and may include a device, an application, a terminal, a user equipment (UE), a mobile station (MS), It may also be called by other terms such as a wireless device, a handheld device, and the like.

IoT refers to a technology or environment in which data can be exchanged and used in real time through objects and sensors attached to them. IoT services based on the IoT environment do not require human manipulation and have the characteristics of enabling Internet communication and information exchange between objects.

In addition, a plurality of UAVs 200 are provided to each fly according to a flight path, and by using data transmitted from a plurality of IoT devices 100 to calculate a distance difference with the corresponding IoT device 100, the calculated distance difference and The strength of the transmission signal of the corresponding data is stored in a data set.

At this time, when the reception of the data transmitted from the IoT device 100 is completed, the UAV 200 calculates using the size of the data transmitted when the corresponding IoT device 100 starts communication and the size of the actually received data. It is possible to further store the mission performance rate in the data set.

In addition, the UAV 200 determines whether a GPS spoofing attack is performed using the data set, and generates a learning model for predicting the destination by learning the data set.

In detail, the UAV 200 determines whether a GPS spoofing attack occurs by using a distance from the IoT device 100 that has transmitted data and the strength of a transmission signal of the corresponding data.

In more detail, the UAV 200 uses the current GPS position value sensed by the built-in GPS module 210 and the position value included in the GPS information of the data transmitted from the IoT device 100 to the corresponding IoT device 100 . ) and the number of IoT devices 100 determined that the strength of the transmission signal of the data transmitted from the IoT device 100 is outside the error range for the strength of the normal signal corresponding to the calculated distance difference , and if the number of IoT devices 100 out of the error range is greater than or equal to a majority of the number of IoT devices 100 that have transmitted data to the UAV 200, it is determined that the UAV 200 has been subjected to a GPS spoofing attack.

In an embodiment of the present invention, the UAV 200 determines whether a GPS spoofing attack occurs by using the characteristic that the shorter the distance between communication targets, the higher the signal strength.

That is, by comparing signal strengths according to distances based on the GPS position values of the IoT device 100 and the UAV 200 , it can be determined whether the UAV 200 has become a target for a spoofing attack.

Here, the GPS spoofing attack means an attack that transmits a false GPS location signal to a GPS signal receiver and causes the receiver to lose an actual GPS location value. Specifically, an attacker steals a real GPS signal and creates a forged copy signal, and the GPS signal becomes weaker as the distance between the satellite transmitting the GPS signal and the UAV 200 receiving it increases.

Therefore, when a false signal manipulated by an attacker is transmitted with a higher strength than the normal signal of the GPS satellite, the frequency is forged and modulated, and the UAV 200 generates the error-applied location information. As a result, as the location information calculated by the UAV 200 is changed, the attacker may be guided to a desired route or may fall due to a collision with an obstacle.

In addition, the UAV 200 learns a learning model to predict a destination with a maximum data collection rate by iteratively calculating a local weight set capable of minimizing a loss function using the mission performance rate stored in the data set.

Finally, the control server 300 transmits the federated learning model integratedly generated by using the parameter set of the learning models generated by each UAV 200 to the plurality of UAVs 200 , respectively. In this case, the control server 300 may be a base station located at a point where the UAV 200 takes off and land, that is, at an UAV airfield and communicates with a plurality of UAVs 200 .

Here, federated learning is a type of distributed learning, and is a technique for learning the same learning model using data generated from each participating node without exchanging data sets. In deep learning, the prediction accuracy of the learning model increases as it learns using a variety of data, so federated learning is a process of re-learning by integrating and redistributing the model so that the data of all nodes can be reflected in the learning model to improve learning performance. do. At this time, since only the parameters of the model are exchanged, not the data set, in order to integrate the models into one, there is an advantage in that it is possible to not only increase the security of the data but also to improve the learning accuracy. In addition, it is a learning method suitable for the IoT device 100, which has a limited storage device and computational cost because it has relatively less computational burden than the centralized model learning method.

In an embodiment of the present invention, the control server 300 generates a federated learning model by calculating global weights according to the following Equation 1 using a local weight set that is a learning model parameter set of each of the UAVs 200 .

Here, GW is a global weight, u is a UAV, n and N are the number of UAVs, and LW _u is a set of local weights.

Therefore, when it is determined that the UAV 200 has been subjected to a GPS spoofing attack, it moves to a destination location predicted by the federated learning model transmitted from the control server 300 .

In addition, after moving to the destination location, the UAV 200 flies along the normal flight path by resetting the current location to the destination location predicted by the federated learning model.

Hereinafter, a UAV control system for preventing GPS spoofing will be described in more detail with reference to FIG. 2 .

2 is a diagram illustrating a UAV control system for preventing GPS spoofing according to an embodiment of the present invention.

이며, u_n의 현재 위치(Current Position)는

로 나타낼 수 있다.As shown in FIG. 2 , the UAV control system for preventing GPS spoofing consists of a UAV 200 group U that travels in air, and an IoT device 100 group I that senses the environment on the ground. At this time, U is when the number of UAVs 200 is n.

, and the current position of u _n is

can be expressed as

In the disaster monitoring system, I measures data during a set period, so I, which transmits the measured data every set time (t), is fixed. Therefore, the control server 300 located at the starting point (Start _Position , _S ) has the position, speed, Set the optimal flight path (Path, P) based on navigation information such as direction.

)을 포함하고 있다. 목적지 위치(DP)는 k-means 알고리즘을 이용하여 설정 시간(t)에 데이터를 송신하는 l의 데이터를 효율적으로 수집할 수 있도록 클러스터링하여 UAV(200)들의 비행경로(P)에 포함한다. At this time, P is a set of l destination positions (Destination Position, DP) to which the UAV 200 must fly (

) is included. The destination location (DP) is included in the flight path (P) of the UAVs 200 by clustering so that data of l transmitting data at the set time (t) can be efficiently collected using the k-means algorithm.

로 나타낼 수 있으며, 이때

의 위치는

으로 표현한다. 또한 i가 전송하는 신호의 세기(Signal Strength)는

로 표현한다.At this time, when the number of IoT devices 100 clustered to collect data by the n-th UAV (u _n ) at the l-th destination location (DP) of the flight path (P) is m,

can be expressed as, where

the location of

to express In addition, the signal strength (Signal Strength) transmitted by i is

expressed as

Table 1 below describes main symbols according to an embodiment of the present invention.

Hereinafter, a UAV control method for preventing GPS spoofing according to an embodiment of the present invention will be described with reference to FIG. 3 .

3 is a flowchart illustrating an operation flow of a UAV control method for preventing GPS spoofing according to an embodiment of the present invention. A detailed operation of the present invention will be described with reference to this.

According to an embodiment of the present invention, first, a plurality of IoT devices 100 located on the ground transmit data including monitoring information and GPS information sensed at a set period interval to the UAV 200 at a set time interval, respectively ( S10).

Then, a plurality of UAVs 200 each flying according to the flight path calculate the distance difference with the IoT device 100 using the data transmitted in step S10, and transmit the calculated distance difference and the corresponding data. The signal strength is stored in the data set (S20).

To describe step S20 in detail, the UAV 20 stores data for understanding the relationship between the distance between the IoT devices 100 and the strength of the transmission signal while flying for attack detection and defense. At this time, the UAV (200, u _n ) calculates the distance between the current location (CP _u ) of the UAV 200 and the GPS location information set (IP ^l _nm ) of the IoT devices 100 that transmit data to the UAV 200 . do. And the calculated distance is stored in the data set (DS _u ) together with the intensity of the transmission signal (SS ^l _nm ).

In addition, when the reception of data transmitted from the IoT device 100 is completed, the UAV 200 performs a task calculated using the size of the data transmitted when the IoT device 100 starts communication and the size of the actually received data. It is also possible to store the performance rate further in the data set.

에 대한 데이터를 저장한다.That is, in order to predict DP ¹ , it is necessary to calculate the complete rate of the UAV 200 in order to determine whether the strength of the transmission signal is within a normal range. Therefore, when the data reception from I ^l _u is finished, the mission performance rate is calculated using the size of the data to be transmitted and the size of the data received when the I ^l _u starts communication with the UAV 200, and the data set ( DS _u ). As a result, in order to detect attacks while the UAV 200 is in flight,

save data for

Then, the UAV 200 generates a learning model for predicting the destination by learning the data set stored in step S20 (S30).

In detail, by iteratively calculating a local weight set (LW _u ) that can minimize the loss function (L) using the mission performance rate stored in the data set, the destination where the data collection rate is maximized is determined. It learns to predict and creates a learning model.

Then, the UAV 200 transmits the learning model generated in step S30 to the control server 300 (S40).

Next, the control server 300 generates a federated learning model by using the parameter set of the learning models generated in each UAV 200 (S50).

In detail, the control server 300 calculates the global weight (Global Weight, GW) by Equation 1 above using the local weight set (LW _u ), which is the learning model parameter set of each of the UAVs 200, and performs federated learning. create a model

Next, the control server 300 transmits the federated learning model generated in step S50 to each of the plurality of UAVs 200 (S60).

Next, the UAV 200 determines whether a GPS spoofing attack is performed using the data set stored in step S20 (S70).

Specifically, in step S70 , the UAV 200 uses the current GPS position value sensed by the built-in GPS module 210 and the position value included in the GPS information of the data transmitted from the IoT device 100 to the corresponding IoT device. The IoT device 100, which calculates the distance difference with 100, and determines that the intensity of the transmission signal of the data transmitted from the IoT device 100 is out of the error range for the intensity of the normal signal corresponding to the calculated distance difference (100) If the number of IoT devices 100 out of the error range is more than a majority of the number of IoT devices 100 that have transmitted data to the UAV 200, it is determined that the UAV 200 has been subjected to a GPS spoofing attack. do.

와

사이의 거리에 기대되는 신호의 세기보다

가 너무 크거나 작은 이상치를 가지면

가 잘못되었을 수 있다. 그러나 신호의 세기는 전파의 잡음 등 영향을 받는 다른 요소들이 존재하므로 하나의 값으로

값이 오류라고 단정 지을 수는 없다. 따라서 거리에 따른 신호 세기의 오차 범위를 벗어나는 IoT 기기(100)가 과반수 이상이라면 UAV(200)는

값의 오류로 판단하여 GPS 스푸핑 공격을 당한 것으로 판단한다.in other words,

Wow

than the expected signal strength for the distance between

has outliers that are too large or too small

may be wrong However, the signal strength is a single value because there are other factors that are affected, such as radio wave noise.

It cannot be concluded that the value is an error. Therefore, if more than half of the IoT devices 100 out of the error range of the signal strength according to the distance, the UAV 200

It is judged as an error in the value and judged to have been subjected to a GPS spoofing attack.

If it is determined that the UAV 200 has been subjected to a GPS spoofing attack as a result of the determination in step S70, it moves to the destination location predicted by the federated learning model received in step S60 (S80).

Finally, after the UAV 200 moves to the destination location predicted by the federated learning model, the current location is reset to the destination location predicted by the federated learning model and flies along the normal flight path (S90).

는 거짓 정보이므로

의 상대적인 위치를 계산한

또한 이상 수치이다. 따라서 UAV(200)는 이동 후

를 원래의

로 재설정한 후 정상 경로(P)를 따라 비행을 진행한다.i.e. current location

is false information

calculated the relative position of

It is also an abnormal number. Therefore, the UAV 200 is moved after

to the original

After resetting to , the flight proceeds along the normal path (P).

As described above, the UAV control system and method for preventing GPS spoofing according to an embodiment of the present invention calculates the difference between the signal strength according to the distance transmitted by the IoT device located on the ground and the position according to the GPS signal of the UAV. By detecting the presence of GPS spoofing attacks in this way, the calculation process for detecting GPS spoofing attacks can be simplified, so that UAV resources can be used efficiently.

In addition, according to an embodiment of the present invention, when a GPS spoofing attack is detected, the actual location is calculated and the UAV's route is reset, so that the UAV's disaster monitoring data collection mission can be smoothly performed, thereby improving the mission performance of the UAV. can do it

In addition, according to an embodiment of the present invention, each UAV learns as a learning technique for calculating the actual location in the case of a GPS spoofing attack, and by applying federated learning in which the learning contents are exchanged at the management institution located at the departure and arrival points, each intermediate base station Compared to the method of exchanging learning contents, communication cost can be reduced, and UAV resources, which have limited storage space, can be used efficiently and accuracy can be improved to prevent the UAV from straying from its path.

Although the present invention has been described with reference to the embodiment shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible therefrom. will be. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the following claims.

100: IoT device 110: sensor
200: UAV 210: GPS module
300: control server

Claims (12)

In the UAV control system for preventing GPS spoofing,
a plurality of IoT devices located on the ground and transmitting data including monitoring information and GPS information sensed at a set period interval at a set time interval;
It is provided in plurality and each flies according to the flight path, calculates the distance difference with the corresponding IoT device using the data transmitted from the plurality of IoT devices, and calculates the calculated distance difference and the strength of the transmitted signal of the data as data a UAV that stores in a set, determines whether a GPS spoofing attack is performed using the data set, and generates a learning model for predicting a destination by learning the data set; and
A control server that transmits the federated learning model integratedly generated by using the parameter set of the learning models generated in each UAV to the plurality of UAVs, respectively,
The UAV is
When it is determined that a GPS spoofing attack has occurred, the UAV control system moves to the destination location predicted by the federated learning model. According to claim 1,
The UAV is
The GPS spoofing attack is determined using the distance to the IoT device that has transmitted data and the strength of the transmitted signal of the data,
The distance difference with the corresponding IoT device is calculated using the current GPS position value and the position value included in the GPS information of the data, and the strength of the transmitted signal of the data is the strength of a normal signal corresponding to the calculated distance difference. By counting the number of IoT devices that are determined to be out of the error range,
A UAV control system that determines that the UAV has been subjected to a GPS spoofing attack if the number of IoT devices out of the above error range is more than half of the number of IoT devices that have transmitted data to the UAV. According to claim 1,
The UAV is
When the reception of the data transmitted from the IoT device is completed, the UAV further stores the task performance rate calculated using the size of the data transmitted when the IoT device starts communication and the size of the actually received data in the data set control system. 4. The method of claim 3,
The learning model is
A UAV control system that learns to predict a destination with a maximum data collection rate by iteratively calculating a local weight set capable of minimizing a loss function using the mission performance rate stored in the data set. According to claim 1,
The control server is
A UAV control system that generates the federated learning model by calculating global weights by the following equation using a local weight set, which is a set of learning model parameters of each UAV:

Here, GW is a global weight, u is a UAV, n and N are the number of UAVs, and LW _u is a set of local weights. According to claim 1,
The UAV is
After moving to the destination location, the current location is reset to the destination location predicted by the federated learning model, and the UAV control system flies along the normal flight path. In the UAV control method performed by the UAV control system for preventing GPS spoofing,
transmitting data including monitoring information and GPS information sensed by a plurality of IoT devices located on the ground at a set period interval to the UAV at set time intervals, respectively;
A plurality of UAVs, each flying according to a flight path, calculate a distance difference with a corresponding IoT device using the data transmitted from the plurality of IoT devices, and calculate the distance between the calculated distance difference and the strength of a transmission signal of the corresponding data. storing in a data set;
generating, by the UAV, a learning model for predicting a destination by learning the data set;
transmitting, by a control server, a federated learning model integratedly generated using a parameter set of learning models generated in each UAV to the plurality of UAVs, respectively;
determining, by the UAV, whether a GPS spoofing attack is performed using the data set; and
When it is determined that the UAV has been subjected to a GPS spoofing attack as a result of the determination, the UAV control method comprising the step of moving to a destination location predicted by the joint learning model. 8. The method of claim 7,
The step of determining whether the GPS spoofing attack is,
The GPS spoofing attack is determined using the distance to the IoT device that has transmitted data and the strength of the transmitted signal of the data,
The distance difference with the corresponding IoT device is calculated using the current GPS position value and the position value included in the GPS information of the data, and the strength of the transmitted signal of the data is the strength of a normal signal corresponding to the calculated distance difference. By counting the number of IoT devices that are determined to be out of the error range,
A UAV control method for determining that the UAV has been subjected to a GPS spoofing attack when the number of IoT devices out of the above error range is more than a majority of the number of IoT devices that have transmitted data to the UAV. 8. The method of claim 7,
Storing the data set includes:
When the reception of the data transmitted from the IoT device is completed, the UAV further stores the task performance rate calculated using the size of the data transmitted when the IoT device starts communication and the size of the actually received data in the data set control method. 10. The method of claim 9,
Creating the learning model comprises:
A UAV control method for generating and learning to predict a destination with a maximum data collection rate by iteratively calculating a regional weighting set capable of minimizing a loss function using the mission performance rate stored in the data set. 8. The method of claim 7,
The control server is
A UAV control method for generating the federated learning model by calculating global weights by the following equation using a local weight set, which is a set of learning model parameters for each UAV:

Here, GW is a global weight, u is a UAV, n and N are the number of UAVs, and LW _u is a set of local weights. 8. The method of claim 7,
After the UAV moves to the destination location, resetting the current location to the destination location predicted by the federated learning model, further comprising the step of flying along a normal flight path.

Images