Attack detection and security defense control method for cluster unmanned system under network attack

Abstract

The invention provides an attack detection and security defense control method for a cluster unmanned system under a network attack, an attack detection algorithm and a security defense strategy are respectively designed, and under the monitoring of the attack detection strategy, when the system is subjected to a hostile attack, a damaged control input signal is replaced by a reliable control input signal, so that the security defense of the cluster unmanned system is realized. The defense strategy is realized based on a virtual network layer, physical state variables are not exchanged in practice, and reliable control input is obtained by exchanging information with neighbor nodes only through virtual nodes in the virtual network layer, so that security tracking control of the cluster unmanned system under network attacks is realized; that is to say, the security defense strategy is based on a communication mechanism of a virtual variable in a virtual network layer and aims to ensure that a follower can safely and effectively track a leader under an unknown attack condition, and the strategy can provide a reliable control input signal in time when an attack is detected, so that the security of the leader is improved. Therefore, the anti-attack capability of the cluster unmanned system is obviously improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cluster unmanned systems, and in particular relates to an attack detection and security defense control method for a cluster unmanned system under network attacks. Background Art

[0002] In recent years, swarm unmanned systems have attracted widespread attention due to their efficient collaborative capabilities in complex environments. Through distributed information processing and collaborative control, these systems can effectively monitor and manage drones, unmanned vehicles and other equipment in large-scale distributed networks. However, as the application scenarios continue to expand, how to ensure the security and consistency of the system in the face of network attacks has become a core issue that needs to be solved urgently.

[0003] Traditional consensus protocols can achieve information sharing and decision consistency to a certain extent through network communication information between unmanned system neighbors, but the effectiveness and reliability of these protocols are seriously challenged in the face of network attacks, especially false data injection attacks. The vulnerability of network communication allows attackers to interfere with control inputs, causing system performance to degrade or even fail. Existing consensus control methods often have shortcomings in the face of such attacks. Therefore, how to design attack detection strategies based on linear function observers and design reliable security defense strategies to enhance the ability of cluster unmanned systems to resist network attacks is an urgent problem to be solved. Summary of the invention

[0004] To solve the above problems, the present invention provides an attack detection and security defense control method for a cluster unmanned system under network attacks, which can achieve safe tracking and control of the cluster unmanned system under network attacks, and improve the security and stability of the cluster unmanned system against network attacks.

[0005] An attack detection and security defense control method for a cluster unmanned system under network attack, which determines whether each follower is attacked according to the set attack detection criteria. For followers whose judgment result is no, the actual input control signal u is used. i (t) Track the leader's status and use control signals for followers whose judgment result is yes Track the status of the leader, where i = 1, 2, ..., N, N represents the number of followers in the swarm unmanned system, δ i (t) represents the false data injection attack signal on the i-th follower. At the same time, when the attacked followers track the status of the leader, they use virtual variables to exchange their respective status information in the virtual network layer. is the control signal of the ith follower obtained by exchanging respective status information according to the virtual network layer, and x i(t) represents the physical state of the ith follower, k 1,i , k 2,i With k 3,i Both represent coefficients, v i and ω i Both represent the dummy variables used by the i-th follower in the virtual network layer.

[0006] Furthermore, the method for determining whether any follower is attacked according to the set attack detection criteria is as follows:

[0007] Step 1: Define the initial values ​​of the variables of the attack detection criteria as follows:

[0008]

[0009] Among them, i represents the attack detection metric corresponding to the i-th follower, α i To represent T c Within the cycle i Reaching the measurement threshold Ψ i,max The cumulative variable of the number of times, s(t) is used to record and judge the attack detection metric Ψ i Whether the cumulative calculation time exceeds the first time period T o The clock function, r(t) is used to record and judge the cumulative variable α i Whether the cumulative calculation time exceeds the second time period T c The clock function of

[0010] Step 2: Continuously calculate the attack detection metric Ψ according to the set time step i as follows:

[0011]

[0012] in, is the actual input control signal u i The estimated value of (t), μ i is the integral weight corresponding to the ith follower, τ is the integral variable;

[0013] Step 3: Determine the currently calculated attack detection metric Ψ i Is it greater than the measurement threshold Ψ i,max And whether s(t) exceeds the first time period T o If the judgment results are all negative, then directly continue to step 2 to step 3. i Not greater than the measurement threshold Ψ i,max And s(t) exceeds the first time period T o , will i and s(t) are reset to 0 and then steps 2 to 3 are executed again. If Ψi Greater than the measurement threshold Ψ i,max , then α i Add 1 and go to step S4;

[0014] Step 4: Determine α i Whether it exceeds the cumulative threshold α i,max And whether r(t) exceeds the second time period T c If the judgment results are all negative, then directly continue to step 2 to step 4. If α i Does not exceed the cumulative threshold α i,max When r(t) exceeds the second time period T c , then i , s(t), r(t) and α i After resetting to 0, execute steps 2 to 4 again. If α i Exceeding the cumulative threshold α i,max And r(t) does not exceed the second time period T c , it means the current follower is under attack.

[0015] Furthermore, the dynamic model of the leader of the swarm unmanned system is as follows:

[0016]

[0017] Among them, x 0 (t) represents the physical state of the leader, y 0 (t) represents the output of the leader’s dynamic model, Represents x 0 (t), A represents the state coefficient matrix, and C represents the output coefficient matrix.

[0018] Furthermore, the dynamic model of each follower in the swarm unmanned system is as follows:

[0019]

[0020] Among them, x i (t) represents the physical state of the ith follower, u i (t) represents the control input signal actually input by the ith follower, y i (t) represents the output of the dynamics model of the ith follower, Represents x i (t), A represents the state coefficient matrix of the follower dynamics model, B represents the input coefficient matrix of the follower dynamics model, C represents the output coefficient matrix of the follower dynamics model, and matrices A, B and C are known.

[0021] Furthermore, the actual input control signal u of any followeri Estimated value of (t) The method to obtain is as follows:

[0022] The linear function observer system for estimating the control input signal is constructed for each follower of the cluster unmanned system as follows:

[0023]

[0024] in, Represents the estimated value of the actual input control signal of each follower The compact form of (·) T represents the transpose of , The actual input control signal u of each follower 1 (t)~u N The compact form of (t), represents the observer state w of each follower 1 (t)~w N The compact form of (t), represents the first-order derivative of w(t), T, F, J and P represent the linear function observer system matrix selected according to the stability of the linear function observer system, the matrix S is the Hurwitz matrix, and the linear function observer system matrix satisfies the following equation:

[0025]

[0026] in, A′ represents the compact form of the state coefficient matrix A, and the symbol represents the Kronecker product, I N represents the N-order identity matrix, B′ represents the compact form of the input coefficient matrix B, C′ represents the compact form of the output coefficient matrix C, K′ represents the compact form of the first feedback gain matrix K, H′ represents the compact form of the second feedback gain matrix H, is the Laplace matrix, Δ is the diagonal matrix, represents the estimate of the observer state in compact form w(t), L is the error coefficient matrix, represents the tracking error e of each follower 1 (t)~e N (t), and the tracking error corresponding to any follower is expressed as e i (t) = x i (t)-x 0 (t);

[0027] The linear function observer system is used to obtain the actual input control signal u of each follower i Estimated value of (t)

[0028] Furthermore, the Laplace matrix The weight coefficient a of the communication network ij ∈{0,1}, and a ij =1 indicates that there is a communication flow between the ith follower and the jth follower, otherwise there is no communication flow, a i0 =1 indicates that there is a communication flow between the ith follower and the leader, otherwise there is no communication flow, and the diagonal matrix Δ = diag{a 10 ,...,a N0}.

[0029] Furthermore, the actual control signal u input by any follower i The method for obtaining (t) is:

[0030]

[0031] u i (t) = Ku' i (t)

[0032] Among them, u' i (t) is the intermediate control signal.

[0033] Furthermore, the variable dynamics equation in the virtual network layer corresponding to any follower is as follows:

[0034]

[0035] Among them, v i ,ω i and θ i All represent the three dummy variables corresponding to the ith follower, v j ,ω j and θ j All represent the three dummy variables corresponding to the ith follower, T vi is the dummy variable v i The corresponding coefficient, T θi is the dummy variable θ i The corresponding coefficient, T ωi is the dummy variable ω i The corresponding coefficient, η i represents the weight coefficient, γ i,j ∈{0, 1},γ i,j With γ j,i is the weight coefficient of the topological structure between any two followers in the virtual network layer, γ i0is the proportional coefficient of the topological structure between followers and leaders in the virtual network layer, k is the sum weight of the topological structure between any two followers, and β is the proportional coefficient of the topological structure between followers and leaders.

[0036] Beneficial effects:

[0037] 1. The present invention provides an attack detection and security defense control method for a cluster unmanned system under a network attack, and designs an attack detection algorithm and a security defense strategy respectively. Under the monitoring of the attack detection strategy, when the system suffers a malicious attack, the destroyed control input signal will be replaced with a reliable control input signal. The defense strategy is implemented based on a virtual network layer. The physical state variables are not actually exchanged. Only by exchanging information between virtual nodes and neighboring nodes in the virtual network layer, reliable control input is obtained to achieve safe tracking and control of the cluster unmanned system under a network attack. In other words, the security defense strategy of the present invention is based on the communication mechanism of virtual variables in the virtual network layer, and aims to ensure that under unknown attack conditions, followers can safely and effectively track leaders. The strategy can provide reliable control input signals in a timely manner when an attack is detected, thereby significantly improving the anti-attack capability of the cluster unmanned system.

[0038] 2. The present invention provides an attack detection and security defense control method for a cluster unmanned system under a network attack, and constructs a control input signal u for each follower of the cluster unmanned system. i (t) and then construct the attack detection criterion Ψ i , and according to i With its threshold Ψ i,max For comparison, use the variable α i Indicates the number of times the cumulative value exceeds the threshold, and finally according to α i Reaching its threshold α i,max At this moment, it means that the system is under attack, which can avoid false alarms, and then the control input signal is changed from u i (t) Switch to a reliable input signal So that each follower can achieve x 0 (t) tracking; it can be seen that the present invention can achieve safe tracking and control of cluster unmanned systems under network attacks, and improve the security and anti-destruction capabilities of the system against network attacks.

[0039] 3. The present invention provides an attack detection and security defense control method for a cluster unmanned system under network attack, and proposes a virtual network layer control framework, wherein Designed in the virtual network layer, v i ,ω i and θ iRepresents virtual variables, which have no physical meaning. They exchange information with neighboring virtual nodes in the virtual network layer and the i-th unmanned system in the physical layer. Through this design, each unmanned system can switch the damaged control input signal in time, thereby achieving the system's resilience and anti-attack capabilities.

[0040] 4. The present invention provides an attack detection and security defense control method for a cluster unmanned system under network attacks. The attack detection strategy is constructed based on a linear function observer system and has a simple observer structure. It can detect the existence of attack signals in a timely manner by comparing the estimated control input signal with the actual input signal, and can enhance the security of the cluster unmanned system when encountering false data injection attacks. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of the construction process of an attack detection and security defense control method for a cluster unmanned system under a network attack provided by the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0043] In order to avoid the unmanned system from being unable to work properly due to the inelasticity of traditional consensus control to attacks when unknown attacks occur, a control input signal u is estimated for each follower of the cluster unmanned system. i (t), and then construct the attack detection criterion Ψ i , through the attack detection algorithm, to avoid false alarms, and then the control input signal is u i (t) Switch to in Designed in the virtual network layer, so that each follower can communicate with the leader x 0 (t) tracking can achieve safe tracking and control of cluster unmanned systems under network attacks, and improve the security and anti-destruction capability of the system against network attacks.

[0044] To achieve the above object, the present invention provides an attack detection and security defense control method for a cluster unmanned system under network attack, wherein the cluster unmanned system consists of a leader and N follower systems, which communicate with each other through a directed topology graph, and the topology graph includes a spanning tree with the leader as the root node. Based on this, the overall process of the present invention is:

[0045] According to the set attack detection criteria, each follower is judged whether it is attacked. For followers whose judgment result is no, the actual input control signal u is used. i(t) Track the leader's status and use control signals for followers whose judgment result is yes Track the status of the leader, where i = 1, 2, ..., N, N represents the number of followers in the swarm unmanned system, δ i (t) represents the false data injection attack signal on the i-th follower. At the same time, when the attacked followers track the status of the leader, they use virtual variables to exchange their respective status information in the virtual network layer. is the control signal of the ith follower obtained by exchanging respective status information according to the virtual network layer, and x i (t) represents the physical state of the ith follower, k 1,i , k 2,i With k 3,i All represent coefficients and meet the system stability requirements. It is required that k 1,i >0, k 3,i >0,v i and ω i Both represent the dummy variables used by the i-th follower in the virtual network layer.

[0046] The construction process of the attack detection and security defense control method of the cluster unmanned system under network attack provided by the present invention is deduced in detail below. Figure 1 As shown, the construction process includes the following steps:

[0047] S1: The dynamic model of the leader of the swarm unmanned system is established as follows:

[0048]

[0049] in represents the state of the leader dynamics model, represents the output of the leader dynamics model, Represents x 0 The first-order derivative of (t), A represents the state coefficient matrix of the leader dynamics model, and C represents the output coefficient matrix of the leader dynamics model;

[0050] S2: The dynamic model of each follower of the swarm unmanned system is established as follows:

[0051]

[0052] Where i = 1, 2, ..., N, indicating that the cluster unmanned system has N followers. represents the physical state of the ith follower dynamics model, represents the control input signal of the ith follower dynamics model, represents the output of the ith follower dynamics model, Represents x i (t), A represents the state coefficient matrix of the follower dynamics model, B represents the input coefficient matrix of the follower dynamics model, and C represents the output coefficient matrix of the follower dynamics model. The matrices A, B and C are known.

[0053] S3: Establish the false data injection attack model as follows:

[0054]

[0055] Among them, δ i (t) represents the false data injection attack signal on the i-th follower. The attack signal acts on the actuator and its amplitude is limited to a predetermined range. within, that is represents the control input signal of the damaged i-th follower;

[0056] S4: A linear function observer system for estimating the control input signal is constructed for each follower of the cluster unmanned system as follows:

[0057]

[0058] in, Represents the estimated value of the actual input control signal of each follower The compact form of (·) T represents the transpose of , The actual input control signal u of each follower 1 (t)~u N The compact form of (t), represents the observer state w of each follower 1 (t)~w N The compact form of (t), represents the first-order derivative of w(t), T, F, J and P represent the linear function observer system matrix selected according to the stability of the linear function observer system, the matrix S is the Hurwitz matrix, and the linear function observer system matrix satisfies the following equation:

[0059]

[0060] in, A′ represents the compact form of the state coefficient matrix A, and the symbol represents the Kronecker product, I N represents the N-order identity matrix, B′ represents the compact form of the input coefficient matrix B, C′ represents the compact form of the output coefficient matrix C, K' and H' are known, K′ represents the compact form of the known first feedback gain matrix K, H′ represents the compact form of the second feedback gain matrix H, is the Laplace matrix, Δ is the diagonal matrix, represents the estimate of the observer state in compact form w(t), L is the error coefficient matrix, represents the tracking error e of each follower 1 (t)~e N (t), and the tracking error corresponding to any follower is expressed as e i (t) = x i (t)-x 0 (t); Laplace matrix The weight coefficient a of the communication network ij ∈{0,1}, and a ij =1 indicates that there is a communication flow between the ith follower and the jth follower, otherwise there is no communication flow, a i0 =1 indicates that there is a communication flow between the ith follower and the leader, otherwise there is no communication flow, and the diagonal matrix Δ = diag{a 10 ,...,a N0}.

[0061] The linear function observer system is used to obtain the actual input control signal u of each follower i Estimated value of (t)

[0062] It should be noted that the variable It is considered from the perspective of the original unmanned system, that is, the expected value of the observer state, so the observation state error is expressed as Then the observer matrix is ​​designed to satisfy the error e when time approaches infinity. w (t) approaches 0.

[0063] S5: Determine the attack detection criteria for each follower of the cluster unmanned system as follows:

[0064]

[0065] Among them, i represents the cumulative error metric of the i-th unmanned system, measuring the deviation between the actual control input and the control input estimate at time [0, T o ], Ψ i As the attack detection criterion, the subsequent algorithm will be based on this criterion for detection, μ i and T o are known parameters;

[0066] Furthermore, the attack detection algorithm of the present invention is:

[0067] S51: Define attack detection algorithm variables and initial values ​​of variables:

[0068]

[0069] Among them, i represents the attack detection metric corresponding to the i-th follower, γ represents the flag variable, γ∈{0,1}, γ=1 means resetting the above variables, α i Indicates the cumulative variable used to store Ψ i Reaching the threshold i,max The number of times, s(t) and r(t) represent clock functions, s(t) is used to record and judge the attack detection metric Ψ i Whether the cumulative calculation time exceeds the first time period T o The clock function, r(t) is used to record and judge the cumulative variable α i Whether the cumulative calculation time exceeds the second time period T c The clock function of ; the specific form of the clock function is as follows:

[0070]

[0071] represents the first-order derivative of s(t), represents the first-order derivative of r(t);

[0072] S52: Continuously calculate the attack detection metric Ψ according to the set time step i as follows:

[0073]

[0074] in, is the actual input control signal u i The estimated value of (t), μ i is the integral weight corresponding to the ith follower, τ is the integral variable;

[0075] S53: Determine the currently calculated attack detection metric Ψ i Is it greater than the measurement threshold Ψ i,max And whether s(t) exceeds the first time period T o If the judgment results are all negative, then directly continue to execute steps S52 to S53. i Not greater than the measurement threshold Ψ i,max And s(t) exceeds the first time period T o , will iand s(t) are reset to 0 and then steps S52 to S53 are executed again. i Greater than the measurement threshold Ψ i,max , then α i Add 1 and go to step S54;

[0076] It should be noted that the dynamic calculation function Ψ i In the process, under safe conditions, with u i The values ​​of are approximately equal, resulting in Ψ i The value of is relatively small, and when an attack occurs, Significant deviation from u i , so that Ψ i The value of is relatively large. i The value of does not exceed the threshold value Ψ i,max When the clock function s(t) is o Reset i and s, and recalculate Ψ in the new cycle i Once i The value of reaches the threshold, then it is necessary to judge α i Whether the conditions are met, see step S54 for details;

[0077] S54: Judgment α i Whether it exceeds the cumulative threshold α i,max And whether r(t) exceeds the second time period T c If the judgment results are all negative, then directly continue to execute steps S52 to S54. i Does not exceed the cumulative threshold α i,max When r(t) exceeds the second time period T c , then i , s(t), r(t) and α i After resetting to 0, re-execute steps S52 to S54. i Exceeding the cumulative threshold α i,max And r(t) does not exceed the second time period T c , it means that the current follower is under attack, so the control input signal is changed from u i (t) Switch to

[0078] S6: First, to simplify the writing, this equation and some formulas in the following text omit the time variable t, for example, using x i Represents x i (t), in order to achieve consistency control between the leader and the follower, the traditional leader-follower consistency control protocol is adopted, and its control signal is defined as follows:

[0079]

[0080] u i =Ku' i

[0081] Among them, u' i (t) is the intermediate control signal;

[0082] It should be noted that, in the present invention, the control input designed by the virtual layer is defined as a reliable control input signal. Under the monitoring of the attack detection strategy, when a potential attack is detected, the control input signal can be switched in time.

[0083] Considering that the control input channel has two situations: safe and attacked, the control input signal is expressed as:

[0084]

[0085] in, represents the control input signal of the i-th unmanned system under attack;

[0086] Furthermore, the virtual variable dynamic equation in the virtual network layer of the present invention is:

[0087] S61: The dummy variable dynamic equation is as follows:

[0088]

[0089] Among them, v i ,ω i and θ i All represent the three dummy variables corresponding to the ith follower, v j ,ω j and θ j All represent the three dummy variables corresponding to the ith follower, T vi is the dummy variable v i The corresponding coefficient, T θi is the dummy variable θ i The corresponding coefficient, T ωi is the dummy variable ω i The corresponding coefficient, η i represents the weight coefficient, γ i,j ∈{0, 1},γ i,j With γ j,i is the weight coefficient of the topological structure between any two followers in the virtual network layer, γ i0 is the proportional coefficient of the topological structure between followers and leaders in the virtual network layer, k is the sum weight of the topological structure between any two followers, and β is the proportional coefficient of the topological structure between followers and leaders.

[0090] It should be noted that the sensors in the virtual network layer are assumed to be safe, and only the actuators may be attacked. This assumption is reasonable because in the presence of a large number of variables, attackers cannot effectively attack all variables.

[0091] S7: The attack detection criterion in step S5 estimates the control input signal through the attack detection algorithm in steps S51-S54, and triggers the conversion of the control signal when an attack is detected, that is, Then the control signal Substitute into step S2 to realize each follower's response to the leader's state x 0 tracking.

[0092] In summary, the present invention provides an attack detection and security defense control method for a cluster unmanned system under network attacks. Although the traditional consensus control protocol is kept simpler and clearer, and communication resources can be saved, it lacks resilience to attacks, resulting in the system being unable to guarantee overall security and stability when attacked. Therefore, the present invention designs an attack detection algorithm and a security defense strategy respectively. Under the monitoring of the attack detection strategy, when the system is subjected to a malicious attack, the destroyed control input signal will be replaced with a reliable control input signal. The defense strategy is implemented based on a virtual network layer, and the physical state variables are not actually exchanged. Only by exchanging information between virtual nodes and neighbor nodes in the virtual network layer, reliable control input is obtained, and safe tracking and control of the cluster unmanned system under network attacks is realized.

[0093] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art may certainly make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for attack detection and security defense control of a cluster unmanned system under network attack, characterized in that: According to the set attack detection criteria, each follower is judged whether it is attacked. For followers whose judgment result is no, the actual input control signal u is used. i (t) Track the leader's status and use control signals for followers whose judgment result is yes Track the status of the leader, where i = 1, 2, ..., N, N represents the number of followers in the swarm unmanned system, δ i (t) represents the false data injection attack signal on the i-th follower. At the same time, when the attacked followers track the status of the leader, they use virtual variables to exchange their respective status information in the virtual network layer. is the control signal of the ith follower obtained by exchanging respective status information according to the virtual network layer, and x i (t) represents the physical state of the ith follower, k 1,i , k 2,i With k 3,i Both represent coefficients, v i and ω i Both represent the dummy variables used by the i-th follower in the virtual network layer.

2. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 1, characterized in that: The method of judging whether any follower is attacked according to the set attack detection criteria is as follows: Step 1: Define the initial values ​​of the variables of the attack detection criteria as follows: Among them, i represents the attack detection metric corresponding to the i-th follower, α i To represent T c Within the cycle i Reaching the measurement threshold Ψ i,max The cumulative variable of the number of times, s(t) is used to record and judge the attack detection metric Ψ i Whether the cumulative calculation time exceeds the first time period T o The clock function, r(t) is used to record and judge the cumulative variable α i Whether the cumulative calculation time exceeds the second time period T c The clock function of Step 2: Continuously calculate the attack detection metric Ψ according to the set time step i as follows: in, is the actual input control signal u i The estimated value of (t), μ i is the integral weight corresponding to the ith follower, τ is the integral variable; Step 3: Determine the currently calculated attack detection metric Ψ i Is it greater than the measurement threshold Ψ i,max And whether s(t) exceeds the first time period T o If the judgment results are all negative, then directly continue to step 2 to step 3. i Not greater than the measurement threshold Ψ i,max And s(t) exceeds the first time period T o , will i and s(t) are reset to 0 and then steps 2 to 3 are executed again. If Ψ i Greater than the measurement threshold Ψ i,max , then α i Add 1 and go to step S4; Step 4: Determine α i Whether it exceeds the cumulative threshold α i,max And whether r(t) exceeds the second time period T c If the judgment results are all negative, then directly continue to step 2 to step 4. If α i Does not exceed the cumulative threshold α i,max When r(t) exceeds the second time period T c , then i , s(t), r(t) and α i After resetting to 0, execute steps 2 to 4 again. If α i Exceeding the cumulative threshold α i,max And r(t) does not exceed the second time period T c , it means the current follower is under attack.

3. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 2, characterized in that: The dynamic model of the leader of the swarm unmanned system is as follows: Among them, x0(t) represents the physical state of the leader, y0(t) represents the output of the leader's dynamic model, represents the first-order derivative of x0(t), A represents the state coefficient matrix, and C represents the output coefficient matrix.

4. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 3, characterized in that: The dynamic model of each follower in the swarm unmanned system is as follows: Among them, x i (t) represents the physical state of the ith follower, u i (t) represents the control input signal actually input by the ith follower, y i (t) represents the output of the dynamics model of the ith follower, Represents x i (t), A represents the state coefficient matrix of the follower dynamics model, B represents the input coefficient matrix of the follower dynamics model, C represents the output coefficient matrix of the follower dynamics model, and matrices A, B and C are known.

5. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 4, characterized in that: The actual input control signal u of any follower i Estimated value of (t) The method to obtain is as follows: The linear function observer system for estimating the control input signal is constructed for each follower of the cluster unmanned system as follows: in, Represents the estimated value of the actual input control signal of each follower The compact form of (·) T represents the transpose of , The actual input control signals u1(t) to u N The compact form of (t), represents the observer state w1(t)~w of each follower N The compact form of (t), represents the first-order derivative of w(t), T, F, J and P represent the linear function observer system matrix selected according to the stability of the linear function observer system, the matrix S is the Hurwitz matrix, and the linear function observer system matrix satisfies the following equation: in, A′ represents the compact form of the state coefficient matrix A, and the symbol represents the Kronecker product, I N represents the N-order identity matrix, B′ represents the compact form of the input coefficient matrix B, C′ represents the compact form of the output coefficient matrix C, K′ represents the compact form of the first feedback gain matrix K, H′ represents the compact form of the second feedback gain matrix H, is the Laplace matrix, Δ is the diagonal matrix, represents the estimate of the observer state in compact form w(t), L is the error coefficient matrix, represents the tracking error of each follower e1(t)~e N (t), and the tracking error corresponding to any follower is expressed as e i (t) = x i (t)-x0(t); The linear function observer system is used to obtain the actual input control signal u of each follower i Estimated value of (t) 6. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 5, characterized in that: Laplacian Matrix The weight coefficient a of the communication network ij ∈{0,1}, and a ij =1 indicates that there is a communication flow between the ith follower and the jth follower, otherwise there is no communication flow, a i0 =1 indicates that there is a communication flow between the ith follower and the leader, otherwise there is no communication flow, and the diagonal matrix Δ = diag{a 10 ,...,a N0 }.

7. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 6, characterized in that: The actual input control signal u of any follower i The method for obtaining (t) is: u i (t)=Ku' i (t) Among them, u' i (t) is the intermediate control signal.

8. The attack detection and security defense control method of a cluster unmanned system under network attack as claimed in claim 6, characterized in that: The variable dynamics equation in the virtual network layer corresponding to any follower is as follows: Among them, v i ,ω i and θ i All represent the three dummy variables corresponding to the ith follower, v j ,ω j and θ j All represent the three dummy variables corresponding to the ith follower, T vi is the dummy variable v i The corresponding coefficient, T θi is the dummy variable θ i The corresponding coefficient, T ωi is the dummy variable ω i The corresponding coefficient, η i represents the weight coefficient, γ i,j ∈{0, 1},γ i,j With γ j,i is the weight coefficient of the topological structure between any two followers in the virtual network layer, γ i0 is the proportional coefficient of the topological structure between followers and leaders in the virtual network layer, k is the sum weight of the topological structure between any two followers, and β is the proportional coefficient of the topological structure between followers and leaders.

Images