Switching system security control method based on two-stage multi-attacker Stacker game
By constructing a handover system security control method based on a two-level multi-attack Stackelberg game, the security control problem of networked handover systems under multi-attack conditions is solved, and proactive security control under deception attacks is achieved, ensuring system stability and security.
Patent Information
- Application Number
- CN202410922607.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-10
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-07-10
AI Technical Summary
Existing game theory models cannot effectively handle the security control problem of networked switching systems by multiple attackers. In particular, the asynchronous behavior of the switching system is more complex under deception attacks, and traditional control schemes are difficult to achieve proactive security control.
A security control method for switching systems based on a two-level multi-attack Stackelberg game is constructed. By building a discrete switching system model and controller, the mode switching signals of the controller and system state information under deception attack are obtained. A two-level multi-attack Stackelberg game model is constructed by combining the Stackelberg game algorithm, and the index functions of SC deception attacker, controller and CA deception attacker are determined. An equilibrium strategy is constructed by using a backward reasoning method to achieve optimal strategy control of the attacker.
It achieves proactive security control of switching systems under deception attacks, ensuring system stability when facing multiple attackers. It guides the behavior of attackers and defenders through a two-level multi-attack Stackelberg game equilibrium strategy, thereby improving system security and stability.
Smart Images

Figure CN118884875A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of switching system safety control, and in particular to a switching system safety control method based on a two-level multi-attacker Stackelberg game. Background Art
[0002] The demand for informatization in modern society is getting higher and higher. The operational stability and security of the system have become the top priority. Due to the long-distance transmission of information, networked control systems have come into being. A networked control system is a control system in which sensors, controllers and actuators communicate through a network. It has the advantages of light weight, low power consumption, high reliability, strong flexibility, easy installation and maintenance, etc., and has attracted widespread attention from the scientific community in the past few decades. A networked switching system consists of a limited number of subsystems and switching laws that determine how these subsystems operate. Due to its special structure, the networked switching system has important applications in describing system models with multi-channel, multi-modal, multi-sensor and unstable dynamics in the actual industrial field, such as aircraft systems, marine vehicle systems, and autonomous robot systems. Since the structure of the network is open, it is easily damaged by network attacks, which causes the information transmitted in the network to be destroyed, thereby affecting the system performance or even instability. Among network attacks, deception attacks are the most difficult to detect and more destructive to the system, which makes the security control problem of networked switching systems under deception attacks a research hotspot. However, when the switching signal in the switching system is also tampered with by deception attacks, more complex asynchronous switching behavior will be generated compared to general control systems. Obviously, the previous passive control scheme alone cannot completely handle the active safety control problem of the switching system under deception attacks. Due to the unique advantages of game theory in simulating the confrontation between malicious attackers and defenders, it has become an effective tool for analyzing the safety issues of networked control systems. However, the existing game theory can only handle the behavior of one attacker and one defender, and cannot obtain an effective solution when multiple attackers appear. Summary of the invention
[0003] The present invention provides a switching system security control method based on a two-level multi-attacker Stackelberg game to overcome the technical problem that when a switching system is subjected to two attacks, the existing Stackelberg game model cannot be used to process the game behavior including two attacks and a controller at the same time.
[0004] In order to achieve the above object, the technical solution of the present invention is:
[0005] A switching system security control method based on two-level multi-attacker Stackelberg game, comprising:
[0006] S1: Construct a discrete switching system model and controller, and obtain the controller mode switching signal, system state information under SC spoofing attack, and control input information under CA spoofing attack;
[0007] S2: constructing a two-level multi-attacker Stackelberg game model according to the Stackelberg game algorithm, wherein the attackers include an SC deception attacker and a CA deception attacker, and obtaining an indicator function of the SC deception attacker, the controller and the CA deception attacker according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller modal switching signal, the system state information and the control input information, wherein the indicator function is used to determine the optimal strategies of the SC deception attacker, the controller and the CA deception attacker;
[0008] S3: Combining the indicator functions of the CA deception attacker, the controller and the SC deception attacker, a two-level multi-attacker Stackelberg equilibrium strategy is constructed by using the reverse method. The two-level multi-attacker Stackelberg equilibrium strategy includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; the two-level multi-attacker Stackelberg equilibrium strategy is used to control the behaviors of the SC deception attacker, the CA deception attacker and the controller;
[0009] S4: controlling the behaviors of the SC deceiving attacker, the controller and the CA deceiving attacker according to the two-level multi-attacker Stackelberg equilibrium strategy to achieve active safety control of the switching system under the deceiving attack.
[0010] Furthermore, S1 constructs a discrete switching system model and controller, and obtains the controller mode switching signal, system state information under the SC deception attack, and control input information under the CA deception attack, including:
[0011] S11. Construct a discrete switching system model, as shown in formula (1):
[0012] x(k+1)=A σ(k) x(k)+B σ(k) u(k) (1)
[0013] in, is the system status, is the system input, A σ(k) is the state matrix, B σ(k) is the input matrix, σ(k)∈S, S∈{0, 1,..., s}, represents the switching signal, s represents the number of subsystems, K(0, k m ) = {k1, k2, ..., k m} represents the set of switching moments, where km is the mth switching moment;
[0014] S12, construct a controller, as shown in formula (2),
[0015]
[0016] Where c(k) represents the switching signal transmitted to the controller, is the input information of the controller, that is, the system status information, H c(k) represents the control gain to be designed;
[0017] S13, obtaining the controller mode switching signal and system status information under the SC spoofing attack, wherein the controller mode switching signal and system status information are as shown in formulas (3) and (4),
[0018]
[0019] in, Indicates the switching signal after being tampered by a spoofing attack, a σ(k)c(k) (k) represents the tampering information of the system status caused by the SC spoofing attack, and denotes σ(k) is attacked by SC spoofing and is not attacked by SC spoofing, respectively, and The mathematical expectation of represents the tampering rate of SC spoofing attack on σ(k);
[0020] S14, obtaining control input information of the controller under CA spoofing attack, as shown in formula (5),
[0021]
[0022] Among them, φ c(k) (k) = 1 and φ c(k) (k)=0 represents the controller The information transmission failure and correct information transmission caused by CA spoofing attack, where φ c(k) The expectation of (k) satisfies E{φ c(k) (k)}=ρ c(k) , E[φ c(k) (k) = 1] = ρ c(k) , E[φ c(k) (k) = 0] = 1 - ρ c(k) , where ρ c(k) The controller outputs information by CA spoofing attack Barrier rate.
[0023] Further, S2 constructs a two-level multi-attacker Stackelberg game model according to the Stackelberg game algorithm, wherein the attackers include an SC deception attacker and a CA deception attacker, and obtains an indicator function of the SC deception attacker, the controller, and the CA deception attacker according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller modal switching signal, the system state information, and the control input information. The indicator function is used to determine the optimal strategy of the SC deception attacker, the controller, and the CA deception attacker, including:
[0024] S21. Construct a two-level multi-attacker Stackelberg game model based on the Stackelberg game algorithm;
[0025] S22, obtaining information sets of the SC spoofing attacker, the controller and the CA spoofing attacker, as shown in formulas (6), (7) and (8),
[0026]
[0027] in, represents the information set of SC deceiving the attacker, I u Represents the information set of the controller, Indicates the information set of the CA spoofing attacker, represents the tampering strategy of the SC deceiving attacker, the control strategy of the controller, and the attack strategy of the CA deceiving attacker. represents the indicator function of SC deceiving the attacker, the indicator function of the controller and the indicator function of CA deceiving the attacker;
[0028] S23. According to the information set of the SC deception attacker, the control gain and the action of the CA deception attacker are obtained, and the switching system equation under the SC deception attack is constructed, as shown in formula (9):
[0029] x1(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ) (9)
[0030] Wherein, σ(k) represents the switching signal in the information set of the SC deceiving the attacker, c(k) represents the switching signal transmitted to the controller in the information set, and the information set here is the information set of the SC deceiving the attacker, H c(k) represents the control gain, φ c(k) The expectation of (k) satisfies E{φ c(k) (k)}=ρ c(k)ρ c(k) φ c(k) (k) represents the action of CA deceiving the attacker;
[0031] According to the information set of the SC deception attacker and the switching system equation under the SC deception attack, the index function of the SC deception attacker is constructed as shown in formulas (10) and (11):
[0032]
[0033] Among them, P σ(k) and S σ(k) is the weight matrix of SC deceiving the attacker and P σ(k) >0,S σ(k) >0,γ d is a constant related to the signal-to-noise ratio of the communication network channel and γ d >0;
[0034] S24, according to the information set of the controller, obtain the action of the CA deceiving the attacker and construct the switching system equation of the controller, as shown in formula (12),
[0035]
[0036] Represents the system status information in the information set. Represents the controller in the information set, represents the control input information under CA spoofing attack; c(k) represents the switching signal transmitted to the controller in the information set; the information set here is the information set of the controller;
[0037] The index function of the controller is constructed according to the information set of the controller and the switching system equation of the controller as shown in formula (13):
[0038]
[0039] Among them, Q c(k) , U c(k) is the weight matrix of the controller and Q c(k) >0,U c(k) >0;
[0040] S25. Construct a switching system equation under CA spoofing attack according to the information set of the CA spoofing attacker, as shown in formula (14):
[0041]
[0042] Among them, the switching system equation of the controller is consistent with the switching system equation under the CA deception attack, c(k) represents the switching signal transmitted to the controller in the information set; the information set here is the information set of the CA deception attacker;
[0043] The indicator function of the CA deception attacker is constructed according to the information set of the CA deception attacker and the switching system equation under the CA deception attack, as shown in formula (15):
[0044]
[0045] Among them, G c(k) and R c(k) is the weight matrix of the CA spoofing attacker and G c(k) >0 and R c(k) >0.
[0046] Furthermore, in the two-level multi-attacker Stackelberg game model, the controller and the CA deception attacker are modeled as the leader and follower in the first-level Stackelberg game, the SC deception attacker is modeled as the leader in the second-level game, and the controller and the CA deception attacker are jointly modeled as two followers in the second-level game.
[0047] Furthermore, S3 combines the indicator functions of the CA deception attacker, the controller and the SC deception attacker, and adopts the reverse method to construct a two-level multi-attacker Stackelberg equilibrium strategy, wherein the two-level multi-attacker Stackelberg equilibrium strategy includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; the two-level multi-attacker Stackelberg equilibrium strategy is used to control the behaviors of the SC deception attacker, the CA deception attacker and the controller, including:
[0048] S31, the CA deceiving attacker minimizes the indicator function of the CA deceiving attacker according to the switching system equation of the CA deceiving attacker, as shown in formula (16),
[0049]
[0050] Assume that the current controller mode is c(k), solve formula (16) and obtain the optimal attack strategy for minimizing the CA deception attacker indicator function, as shown in formula (17):
[0051]
[0052] S32, the controller acts as a leader, obtains the optimal attack strategy of the CA deceiving the attacker, and minimizes the index function of the controller according to the optimal attack strategy of the CA deceiving the attacker and the switching system equation of the controller, as shown in formula (18),
[0053]
[0054] According to the optimal attack strategy of the CA deceiving attacker and the indicator function of the controller, formula (18) is solved to obtain the optimal strategy, as shown in formula (19):
[0055]
[0056] Satisfy constraints
[0057] S33, the SC deception attacker acts as the common leader of the controller and the CA deception attacker, obtains the optimal attack strategy of the CA deception attacker and the optimal tampering strategy of the controller, and minimizes the indicator function of the SC deception attacker, as shown in formulas (20) and (21),
[0058]
[0059] stx1(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ), (20)
[0060]
[0061] stx1(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ) (twenty one)
[0062] The SC deception attacker solves the minimum value of the indicator function of formulas (20) and (21) according to the optimal attack strategy of the CA deception attacker, the optimal tampering strategy of the controller and the indicator function of the SC deception attack, and obtains the optimal tampering strategy, as shown in formulas (22) and (23).
[0063]
[0064] Satisfy constraints
[0065] The present invention designs a two-level multi-attacker Stackelberg game model, obtains an indicator function, determines the optimal strategy of SC deceiving attackers and controlling CA deceiving attackers, constructs a two-level multi-attacker Stackelberg equilibrium strategy according to the indicator function, uses the equilibrium solution of the two-level game to guide the behaviors of two attackers and one defender, and realizes the effect of active safety control of the switching system based on game theory under deception attacks. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0067] Figure 1 It is a method flow chart of the switching system security control method based on two-level multi-attacker Stackelberg game of the present invention;
[0068] Figure 2 This is the control system block diagram under deception attack;
[0069] Figure 3 A modal diagram of a subsystem and a controller of an embodiment of the present invention;
[0070] Figure 4 A Stackelberg equilibrium action line chart of a CA deceiving attacker according to an embodiment of the present invention;
[0071] Figure 5 This is a Stackelberg equilibrium action line diagram of an embodiment of the present invention where an SC deceives an attacker to tamper with a switching signal;
[0072] Figure 6 This is a Stackelberg equilibrium action line diagram of an embodiment of the present invention, in which an SC deceives an attacker to tamper with the system state;
[0073] Figure 7 A system status trajectory diagram of an embodiment of the present invention;
[0074] Figure 8 This is a Stackelberg equilibrium action line chart of a controller according to an embodiment of the present invention. DETAILED DESCRIPTION
[0075] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0076] This embodiment provides a switching system security control method based on a two-level multi-attacker Stackelberg game. Figure 1 As shown, including:
[0077] S1: Construct a discrete switching system model and controller, and obtain the controller mode switching signal, system state information under SC spoofing attack, and control input information under CA spoofing attack;
[0078] S2: constructing a two-level multi-attacker Stackelberg game model according to the Stackelberg game algorithm, wherein the attackers include an SC deception attacker and a CA deception attacker, and obtaining an indicator function of the SC deception attacker, the controller and the CA deception attacker according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller modal switching signal, the system state information and the control input information, wherein the indicator function is used to determine the optimal strategies of the SC deception attacker, the controller and the CA deception attacker;
[0079] S3: Combining the indicator functions of the CA deception attacker, the controller and the SC deception attacker, a two-level multi-attacker Stackelberg equilibrium strategy is constructed by using the reverse method. The two-level multi-attacker Stackelberg equilibrium strategy includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; the two-level multi-attacker Stackelberg equilibrium strategy is used to control the behaviors of the SC deception attacker, the CA deception attacker and the controller;
[0080] S4: controlling the behaviors of the SC deceiving attacker, the controller and the CA deceiving attacker according to the two-level multi-attacker Stackelberg equilibrium strategy to achieve active safety control of the switching system under the deceiving attack.
[0081] Specifically, firstly, a discrete switching system model and a controller are constructed, and the controller modal switching signal, system state information and control input information under the SC deception attack are obtained, which is helpful for the design of subsequent active safety control schemes; secondly, a two-level multi-attacker Stackelberg game model is constructed according to the Stackelberg game algorithm, and the attackers include SC deception attackers and CA deception attackers, and the indicator functions of the SC deception attacker, the controller and the CA deception attacker are obtained according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller modal switching signal, the system state information and the control input information. The indicator function is used to determine the optimal strategies of the SC deception attacker, the controller and the CA deception attacker. Through the indicator function, it is possible to accurately describe how each party makes decisions based on its own strategies and the opponent's strategies. The leader makes a decision first, and the followers optimize their own strategies according to the leader's decision; combining the indicator functions of the CA deception attacker, the controller and the SC deception attacker, the reverse method is used to construct a two-level multi-attacker Stackelberg equilibrium strategy, which includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; the two-level multi-attacker Stackelberg equilibrium strategy is used to control the behaviors of the SC deception attacker, the CA deception attacker and the controller, and the two-level multi-attacker Stackelberg game equilibrium analysis helps to find the optimal strategies for the attacker and the defender, so that the system can remain relatively stable when facing attacks; finally, the behaviors of the SC deception attacker, the controller and the CA deception attacker are controlled according to the two-level multi-attacker Stackelberg equilibrium strategy to achieve active safety control of the switching system under deception attacks.
[0082] Figure 2 is a system block diagram of the switching system of this embodiment, and multiple attackers refer to multiple attack behaviors of SC deceiving attackers and CA deceiving attackers;
[0083] In a specific embodiment, a discrete switching system model and a controller are constructed, and a scheme for obtaining a controller mode switching signal, system state information, and control input information under a CA spoofing attack is:
[0084] S11. Construct a discrete switching system model, as shown in formula (24):
[0085] x(k+1)=A σ(k) x(k)+B σ(k) u(k) (24)
[0086] in, is the system status, is the system input, A σ(k) is the state matrix, Bσ(k) is the input matrix, σ(k)∈S, S∈{0, 1,..., s}, represents the switching signal, s represents the number of subsystems, K(0, k m ) = {k1, k2, ..., k m} represents the set of switching moments, where k m is the mth switching moment;
[0087] S12. Construct a controller as shown in formula (25):
[0088]
[0089] Where c(k) represents the switching signal transmitted to the controller, is the input information of the controller, that is, the value of the system state when it reaches the controller, is the system state information, and H c(k) represents the control gain to be designed;
[0090] S13, obtaining the controller mode switching signal and system status information under the SC spoofing attack, wherein the controller mode switching signal and system status information are as shown in formulas (26) and (27),
[0091]
[0092] in, Indicates the switching signal after being tampered by a spoofing attack, a σ(k)c(k) (k) represents the tampering information of the system status caused by the SC spoofing attack, and denotes σ(k) is attacked by SC spoofing and is not attacked by SC spoofing, respectively, and The mathematical expectation of
[0093] is the tampering rate of SC spoofing attack on σ(k);
[0094] S14. In the CA network, when the CA spoofing attacker tampers with the controller and causes the information to become invalid, the control input u(k) will be cleared, and the system will run in an open loop to obtain the control input information of the controller under the CA spoofing attack, as shown in formula (28).
[0095]
[0096] Among them, φ c(k) (k) = 1 and φ c(k) (k)=0 represents the controller Information transmission failure and correct information transmission caused by CA spoofing attacks;φ c(k) The expectation of (k) satisfies E{φc(k) (k)}=ρ c(k) ; E[φ c(k) (k) = 1] = ρ c(k) , E[φ c(k) (k) = 0] = 1 - ρ c(k) , where ρ c(k) It is the blocking rate of controller output information caused by CA spoofing attack.
[0097] The present invention constructs a switching model and obtains switching signals and control input information under SC spoofing attacks and CA spoofing attacks, which is helpful for the design of subsequent active safety control solutions.
[0098] In a specific embodiment, a two-level multi-attacker Stackelberg game model is constructed according to the Stackelberg game algorithm, the attackers include an SC deception attacker and a CA deception attacker, and the indicator functions of the SC deception attacker, the controller and the CA deception attacker are obtained according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller mode switching signal, the system state information and the control input information. The indicator function is used to determine the optimal strategy of the SC deception attacker, the controller and the CA deception attacker:
[0099] S21. Construct a two-level multi-attacker Stackelberg game model based on the Stackelberg game algorithm:
[0100] In the two-level multi-attacker Stackelberg game model, the controller and the CA deception attacker are modeled as the leader and follower in the first-level Stackelberg game, the SC deception attacker is modeled as the leader in the second-level game, and the controller and the CA deception attacker are jointly modeled as two followers in the second-level game;
[0101] S22, obtaining information sets of the SC spoofing attacker, the controller and the CA spoofing attacker, as shown in formulas (29), (30) and (31),
[0102]
[0103] in, represents the information set of SC deceiving the attacker, I u Represents the information set of the controller, Indicates the information set of the CA spoofing attacker, represents the tampering strategy of the SC deceiving attacker, the control strategy of the controller, and the attack strategy of the CA deceiving attacker. represents the indicator function of the SC deceiving attacker, the indicator function of the controller and the indicator function of the CA deceiving attacker; due to the leading position in the game, the SC deceiving attacker can know the indicator function J of the controller and the CA deceiving attacker u , and Strategy S u , The controller is in the middle of the two-level multi-attacker Stackelberg game and can learn the indicator function of the CA to deceive the attacker. and strategies
[0104] S23. According to the information set of the SC deception attacker, the control gain and the action of the CA deception attacker are obtained and the switching system equation under the SC deception attack is constructed, as shown in formula (32).
[0105] x1(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ) (32)
[0106] Wherein, σ(k) represents the switching signal in the information set of the SC deceiving the attacker, c(k) represents the switching signal transmitted to the controller in the information set, and the information set here is the information set of the SC deceiving the attacker, H c(k) represents the control gain, φ c(k) The expectation of (k) satisfies E{φ c(k) (k)}=ρ c(k) ρ c(k) φ c(k) (k) represents the action of the CA deceiving attacker. The SC deceiving attacker can obtain the indicator function and strategy of the controller and the CA deceiving attacker. The control gain of the controller and the action of the CA deceiving attacker can be derived through the indicator function and strategy.
[0107] According to the information set of the SC deception attacker and the switching system equation under the SC deception attack, the index function of the SC deception attacker is constructed, as shown in formulas (33) and (34),
[0108]
[0109] Among them, P σ(k) and S σ(k) is the weight matrix of SC deceiving the attacker and P σ(k) >0,S σ(k) >0,γ d is a constant related to the signal-to-noise ratio of the communication network channel and γ d >0;
[0110] S24. According to the information set of the controller, obtain the action of the CA deceiving the attacker and construct the switching system equation of the controller, as shown in formula (35),
[0111]
[0112] Represents the system status information in the information set. Represents the controller in the information set, represents the control input information under the CA deception attack; c(k) represents the switching signal transmitted to the controller in the information set; the information set here is the information set of the controller; the controller can obtain the indicator function and strategy of the CA deception attacker, and the action of the CA deception attacker can be derived through the indicator function and strategy of the CA deception attacker;
[0113] The index function of the controller is constructed according to the information set of the controller and the switching system equation of the controller as shown in formula (36):
[0114]
[0115] Among them, Q c(k) , U c(k) is the weight matrix of the controller and Q c(k) >0,U c(k) >0;
[0116] S25. Construct a switching system equation under CA spoofing attack according to the information set of the CA spoofing attacker, as shown in formula (37):
[0117]
[0118] Among them, the switching system equation of the controller is consistent with the switching system equation under the CA deception attack, c(k) represents the switching signal transmitted to the controller in the information set; the information set here is the information set of the CA deception attacker;
[0119] The indicator function of the CA deception attacker is constructed according to the information set of the CA deception attacker and the switching system equation under the CA deception attack, as shown in formula (38):
[0120]
[0121] Among them, G c(k) >0 and R c(k) >0 is the weight matrix of the CA spoofing attacker.
[0122] In this embodiment, the indicator functions of SC deceiving attackers, controllers and CA deceiving attackers are designed. Through the indicator functions, it is possible to accurately describe how each party makes decisions based on its own strategy and the opponent's strategy. The leader makes a decision first, and the followers optimize their own strategies based on the leader's decision.
[0123] In a specific embodiment, in combination with the indicator functions of the CA deception attacker, the controller and the SC deception attacker, a two-level multi-attacker Stackelberg equilibrium strategy is constructed by using a reverse method. The two-level multi-attacker Stackelberg equilibrium strategy includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; the two-level multi-attacker Stackelberg equilibrium strategy is used to control the behavior of the SC deception attacker, the CA deception attacker and the controller:
[0124] S31. Constraints for constructing a two-level multi-attacker Stackelberg equilibrium strategy:
[0125] (1) For a given First and second level Stackelberg game common follower CA cheating attacker There is a Can minimize the indicator function Followers There is always a mapping Satisfy the conditions
[0126]
[0127] in, is the optimal response of the follower in the first-level Stackelberg game;
[0128] (2) For a given Follower P of the second-level Stackelberg game u The controller (leader of the first level) has a Can minimize the indicator function That is, the followers of the second-level Stackelberg game always have a mapping Satisfy the conditions
[0129]
[0130] in, It is called the optimal response of the follower in the second-level Stackelberg game;
[0131] (3) In the second-level game, the leader SC deceives the attacker There are also strategies Its indicator function can be minimized So that for all Satisfy the conditions
[0132]
[0133] is the optimal leadership strategy in the second-level Stackelberg game, It is also the optimal leadership strategy for leaders in the first-level Stackelberg game and the optimal following strategy for followers in the second-level Stackelberg game. is the optimal following strategy in the first and second level Stackelberg game;
[0134] S32, the CA deceiving attacker minimizes the indicator function of the CA deceiving attacker according to the switching system equation of the CA deceiving attacker, as shown in formula (39),
[0135]
[0136] Assume that the current controller mode is c(k), solve formula (39) and obtain the optimal attack strategy for minimizing the CA deception attacker indicator function, as shown in formula (40):
[0137]
[0138] Satisfy constraints (1);
[0139] Proof: Substitute into the objective function And simplify formula (41),
[0140]
[0141] Among them, Z c(k) The CA deception attacker indicator function in formula (39) The minimization problem of Z is transformed into c(k) In the independent variable ρ c(k) Next, let Z c(k) P c(k) Taking the derivative as 0, we get formula (42):
[0142]
[0143] Consider the packet loss rate ρ in the actual situation c(k) It should be between (0, 1), and the optimal strategy for CA to deceive attackers in the two-level multi-attacker Stackelberg game is formula (40);
[0144] S33, the controller acts as a leader, obtains the optimal attack strategy of the CA deceiving the attacker, and minimizes the index function of the controller according to the optimal attack strategy of the CA deceiving the attacker and the switching system equation of the controller, as shown in formula (43),
[0145]
[0146] According to the optimal attack strategy of the CA deceiving attacker and the indicator function of the controller, formula (43) is solved to obtain the optimal strategy, as shown in formula (44):
[0147]
[0148] Among them, the constraints are met and (2) in the constraints;
[0149] Proof: Substitute into J u Then, we can simplify the formula (45)
[0150]
[0151] From the constraints, we know that the independent variable The second-order derivative of the quadratic function is greater than zero, and the controller index function in formula (43) The minimization problem can be transformed into In the independent variable The optimal solution problem of the next-order derivative equal to 0 is right The derivative of is equal to 0, and we get formula (46),
[0152]
[0153] Solving the above equation, we can obtain the optimal strategy of the controller in the two-level multi-attacker Stackelberg game as formula (44);
[0154] S34, the SC deception attacker acts as the common leader of the controller and the CA deception attacker, obtains the optimal attack strategy of the CA deception attacker and the optimal tampering strategy of the controller, and minimizes the indicator function of the SC deception attacker, as shown in formulas (47) and (48),
[0155]
[0156] The SC deception attacker solves the minimum value of the indicator function of formulas (47) and (48) according to the optimal attack strategy of the CA deception attacker, the optimal tampering strategy of the controller and the indicator function of the SC deception attack, and obtains the optimal tampering strategy, as shown in formulas (49) and (50).
[0157]
[0158] Among them, the constraints are met and constraints (3);
[0159] Proof: Find the optimal tampering information that SC uses to deceive the attacker about the system state x(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ) into We can obtain formula (51) from
[0160]
[0161] From the constraints, we know that a σ(k)c(k) The coefficient before the quadratic term is positive, and the indicator function in formula (47) The minimization problem can be transformed into Tampering information a σ(k)c(k) The optimal solution problem with derivative equal to 0 is to a σ(k)c(k) Taking the derivative equal to 0, we get formula (52),
[0162]
[0163] According to formula (52), formula (49) is valid;
[0164] Find the optimal tampering strategy of the SC spoofing attacker on the switching signal Assume that the σ(k)th subsystem is currently activated. When the switching signal is tampered with, the controller switches to in x1(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ) into Then, we can simplify formula (53) by combining them.
[0165]
[0166] By γ d >0, the indicator function in formula (48) The minimization problem is then transformed into Packet loss function for the switching signal σ(k) Find the optimal value problem when the derivative is equal to 0. Let right The derivative is equal to 0, and we get formula (54),
[0167]
[0168] According to formula (54), formula (50) is the optimal tampering strategy for SC deception attack on σ(k);
[0169] In summary, combined with the constraints, the optimal attack strategy for SC to deceive the attacker is The optimal control strategy of the controller as a defender is The best attack strategy for CA spoofing attackers is and Together they form the equilibrium strategy under the two-level multi-attacker Stackelberg game;
[0170] The present invention constructs an equilibrium strategy under the two-level multi-attacker Stackelberg game, which can guide the behavior of the attacker and the controller. Under the guidance of the equilibrium strategy, the controller can still maximize the security and stability of the system when the attacker adopts the best attack strategy. The equilibrium analysis of the two-level multi-attacker Stackelberg game helps to find the optimal strategy for the attacker and the defender, so that the system can remain relatively stable when facing attacks.
[0171] In a specific embodiment, the state of the SC deception attacker, the controller and the CA deception attacker is controlled according to the two-stage multi-attacker Stackelberg equilibrium strategy, and the scheme for realizing active safety control of the switching system under deception attack is:
[0172] The behaviors of the three participants are guided by the equilibrium solution of the Stackelberg game. Assume that the current i-th subsystem is activated, the controller is in mode j, the switching signal under the deception attack is tampered with to p, and The switching system is expressed as formula (55),
[0173] x(k+1)=A i x(k)+(1-φ j (k))B i H j (x(k)+a ij (k)) (55)
[0174] Theorem 1: For a given κ i ∈(-∞,0), μ i ∈(1,∞), if there exists a positive definite matrix Satisfying the conditions of formulas (56), (57), (58), (59) and (60),
[0175]
[0176]
[0177] Then the switching system of formula (24) is mean square exponentially stable under the conditions of satisfying the three optimal strategies;
[0178] Proof: Depending on whether the SC spoofing attack tampers with σ(t) and causes system modal asynchrony, it can be divided into the following two forms of synchronization and asynchronous for discussion.
[0179] (1) When the switching signal σ(k) is not attacked (case A), The system modes are synchronized, and the controller mode j = i. The system state equation is shown in formula (61):
[0180]
[0181] (2) When the switching signal σ(k) transmitted in the SC network is attacked by spoofing (case B), The system mode is asynchronous, the controller mode is j=p, and the system state equation under the system mode is shown in formula (62):
[0182]
[0183] In summary, we select the Lyapunov universal function as shown in formula (63):
[0184]
[0185] Case A: When j = i, the deceptive attacker in the SC network does not tamper with the switching signal transmitted in the network. The deceptive attacker's deceptive information about the system state is Formula (63) under modal synchronization can be mainly divided into the following two cases:
[0186] A1: The spoofing attack in the CA network successfully blocks the controller output information The system operates in open loop;
[0187] A2: Failure to block spoofing attackers in CA network The system is only affected by the behavior of the SC deceiving attacker;
[0188] When the system is in situation A1, take φ i Substituting (k)=1 into formula (60), and combining with formula (56), we can derive the differential Lyapunov function ΔV i (k) Satisfy the conditions
[0189]
[0190] That is, formula (64),
[0191] V i (k+1)≤(1-κ i )V i (k) (64)
[0192] When the system is in situation A2, φ i Substitute (k) = 0 into formula (60) and combine with formula (57) to derive the differential Lyapunov function ΔV i (k)
[0193]
[0194] Simplifying formula (65),
[0195]
[0196] According to formulas (64) and (65), when the system modes are synchronized, the inequality of formula (66) holds true
[0197]
[0198] Case B: When σ(k) is tampered, The system mode is asynchronous and j=p, which can be mainly divided into the following two cases:
[0199] B1: CA spoofing attack successfully blocks controller output information The system operates in open loop.
[0200] B2: CA spoofs attacker and does not attack The health of the closed-loop system depends only on the behavior of the SC spoofing attack.
[0201] When the system is in situation B1, φ p Substituting (k)=1 into formula (62) and combining it with formula (56), we get formula (67):
[0202] V i (k+1)≤(1-κ i )V i (k) (67)
[0203] When the system is in situation B2, φ p (k) = 0 and Substituting into formula (62) and combining with formula (58), the differential Lyapunov function ΔV can be derived i (k)
[0204]
[0205] That is, formula (68),
[0206]
[0207] In summary, when the system is asynchronous, we get formula (69)
[0208]
[0209] Where j = p. Since the attacker’s tampering rate of σ(k) is Combining formula (66) and (69) we get formula (70),
[0210]
[0211] Considering the jump of formula (63) at the switching point, substitute formula (59) into formula (70) to obtain formula (71),
[0212]
[0213]
[0214] make represents the time of switching to subsystem i for the qth time, represents the last moment in its activation time, and the formula (63) in the entire system running time can be derived as
[0215]
[0216] according to The above formula can be simplified to
[0217]
[0218] Similarly, according to φ j (k)∈{0,1}, continue to simplify the above formula to get
[0219]
[0220] According to the optimal attack strategy of the CA deceiving attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deceiving attacker, it can be seen that under the equilibrium strategy of the two-level multi-attacker Stackelberg game, the optimal tampering rate of the SC deceiving attacker on the switching signal σ(t) is CA deceives attackers to output information to the controller The optimal barrier rate is make represents the residence time of the qth switch to subsystem i. During the entire system operation time, the expectation of formula (62) satisfies
[0221]
[0222] Among them, T i (k0, k) represents the length of time that the ith subsystem is activated within the interval (k0, k). For the ith subsystem,
[0223]
[0224] definition Will and Substituting into the above formulas, we get formula (72):
[0225]
[0226] Combining equations (60) and (72), it can be concluded that the switching system equation (24) is mean square exponentially stable when the switching signal σ(k) satisfies the mode-dependent average dwell time switching condition.
[0227] In this embodiment, a switching RLC circuit is used to verify the effectiveness of the active safety control scheme in this chapter. The switching RLC circuit can be modeled as a continuous-time switching system, where x = [q c i L ] T ,q c represents the charge of the capacitor, i L is the magnetic flux in the inductor. The corresponding system matrix can be expressed as
[0228]
[0229] Select L = 1H, R = 1Ω, c1 = 1F, c2 = 0.5F, c3 = 0.67F, then the specific system matrix value is expressed as
[0230]
[0231] The switching RLC system matrix of the continuous-time system is discretized accordingly, and the discrete time step is selected as T = 1s. The discretized switching system is obtained, and its system matrix is
[0232] In the two-level multi-attacker Stackelberg game, the weight matrices of the SC deceiving attacker, the controller, and the CA deceiving attacker are set as
[0233] S1=S2=S3=I, U1=U2=U3=I, G1=G2=G3=I, Q1=Q2=Q3=1, R1=R2=R3=1,
[0234] According to the obtained two-level multi-attacker Stackelberg game equilibrium solution, the optimal control gain can be obtained for
[0235] H1=[0.3185 0.2275],H2=[0.3770 0.2828],H3=[0.0183 -0.1187]. CA deceives the attacker to output information to the controller The optimal barrier rate It can be expressed as
[0236]
[0237] The parameters in Theorem 3.1 are chosen as κ1 = -0.1, κ2 = -0.2, κ3 = -0.2, μ1=1.7,μ2=1.8,μ3=1.4,then the positive definite matrix and the modal correlation average dwell time are expressed as follows;
[0238]
[0239] T 1A ≥5.0269, T 2A ≥2.5986, T 3A ≥3.0489.
[0240] The results of the simulation experiment are as follows Figure 3-Figure 8 As shown,
[0241] SC spoofing attack modifies the subsystem and controller modes under the switching signal σ(t) as follows Figure 3 As shown, when σ(t) is tampered, the system mode is asynchronous;
[0242] CA deceives attackers to output information to the controller The blocking rate like Figure 4 As shown in the figure, the blue solid line represents the optimal attack strategy of CA spoofing attack on the controller output information, and the red dotted line represents the controller switching signal. According to the optimal strategy of the SC deceiving attacker, the optimal tampering rate of the SC deceiving attacker on the switching signal σ(k) is obtained. And the attack information on the system state x(k), as shown in Table 1, Figure 5 and Figure 6 As shown,
[0243] Table 1 Stackelberg equilibrium actions of SC spoofing attacker to tamper with switching signals
[0244]
[0245] From the analysis of Table 1, we can see that when the SC spoofing attacker steals the switching signal σ(k), he will choose to tamper σ(k) to other modes different from the subsystem mode, and the probability of successful tampering to different modes is also different;
[0246] The green solid line frame As shown in and σ(k), when the subsystem is in mode 2, the SC deception attacker may tamper with the switching signal to 1 or 3 at different times, and the tampering rate corresponding to tampering with different modes is also different. The SC deception attacker uses the stolen x(k) information to imitate the action of x(k) to launch an attack, as shown in Figure 6 , Figure 7 As shown; Figure 6 The blue dashed line in the middle represents the optimal strategy for SC to deceive the attacker to tamper with the system state x(k) The upper half of the matrix, the solid red line represents the optimal strategy for SC to deceive the attacker to tamper with the system state x(k) The lower half of the matrix;
[0247] When x(k) tends to be stable, the SC spoofing attacker will also give up tampering with x(k), that is, Under the guidance of the actions of the two-level switching Stackelberg game equilibrium, the controller finally reaches a stable state, such as Figure 8 shown.
[0248] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A switching system security control method based on two-level multi-attacker Stackelberg game, characterized in that: include: S1: Construct a discrete switching system model and controller, and obtain the controller mode switching signal, system state information under SC spoofing attack, and control input information under CA spoofing attack; S2: constructing a two-level multi-attacker Stackelberg game model according to the Stackelberg game algorithm, wherein the attackers include an SC deception attacker and a CA deception attacker, and obtaining an indicator function of the SC deception attacker, the controller and the CA deception attacker according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller modal switching signal, the system state information and the control input information, wherein the indicator function is used to determine the optimal strategies of the SC deception attacker, the controller and the CA deception attacker; S3: Combining the indicator functions of the CA deception attacker, the controller and the SC deception attacker, a two-level multi-attacker Stackelberg equilibrium strategy is constructed by using the reverse method. The two-level multi-attacker Stackelberg equilibrium strategy includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; the two-level multi-attacker Stackelberg equilibrium strategy is used to control the behaviors of the SC deception attacker, the CA deception attacker and the controller; S4: controlling the behaviors of the SC deceiving attacker, the controller and the CA deceiving attacker according to the two-level multi-attacker Stackelberg equilibrium strategy to achieve active safety control of the switching system under the deceiving attack.
2. The switching system security control method based on two-level multi-attacker Stackelberg game according to claim 1 is characterized in that: S1 builds a discrete switching system model and controller, and obtains the controller mode switching signal, system state information and control input information under the SC deception attack, including: S11. Construct a discrete switching system model, as shown in formula (1): x(k+1)=A σ(k) x(k)+B σ(k) u(k) (1) in, is the system status, is the system input, A σ(k) is the state matrix, B σ(k) is the input matrix, σ(k)∈S, S∈{0,1,...,s}, represents the switching signal, s represents the number of subsystems, K(0,k m )={k1,k2,…,k m } represents the set of switching moments, where k m is the mth switching moment; S12, construct a controller, as shown in formula (2), Where c(k) represents the switching signal transmitted to the controller, is the input information of the controller, that is, the system status information, H c(k) represents the control gain to be designed; S13, obtaining the controller mode switching signal and system status information under the SC spoofing attack, wherein the controller mode switching signal and system status information are as shown in formulas (3) and (4), in, Indicates the switching signal after being tampered by a spoofing attack, a σ(k)c(k) (k) represents the tampering information of the system status caused by the SC spoofing attack, and denotes σ(k) is attacked by SC spoofing and is not attacked by SC spoofing, respectively, and The mathematical expectation of represents the tampering rate of SC spoofing attack on σ(k); S14, obtaining control input information of the controller under CA spoofing attack, as shown in formula (5), Among them, φ c(k) (k) = 1 and φ c(k) (k)=0 represents the controller The information transmission failure and correct information transmission caused by CA spoofing attack, where φ c(k) The expectation of (k) satisfies E{φ c(k) (k)}=ρ c(k) , E[φ c(k) (k) = 1] = ρ c(k) , E[φ c(k) (k) = 0] = 1 - ρ c(k) , where ρ c(k) The controller outputs information by CA spoofing attack Barrier rate.
3. The switching system security control method based on two-level multi-attacker Stackelberg game according to claim 1 is characterized in that: S2 constructs a two-level multi-attacker Stackelberg game model according to the Stackelberg game algorithm, wherein the attackers include an SC deceiving attacker and a CA deceiving attacker, and obtains an index function of the SC deceiving attacker, the controller, and the CA deceiving attacker according to the two-level multi-attacker Stackelberg game model, the switching system model, the controller modal switching signal, the system state information, and the control input information. The index function is used to determine the optimal strategy of the SC deceiving attacker, the controller, and the CA deceiving attacker, including: S21. Construct a two-level multi-attacker Stackelberg game model based on the Stackelberg game algorithm; S22, obtaining information sets of the SC spoofing attacker, the controller and the CA spoofing attacker, as shown in formulas (6), (7) and (8), in, represents the information set of SC deceiving the attacker, I u Represents the information set of the controller, Represents the information set of the CA spoofing attacker, represents the tampering strategy of the SC deceiving attacker, the control strategy of the controller, and the attack strategy of the CA deceiving attacker. represents the indicator function of SC deceiving the attacker, the indicator function of the controller and the indicator function of CA deceiving the attacker; S23. According to the information set of the SC deception attacker, the control gain and the action of the CA deception attacker are obtained, and the switching system equation under the SC deception attack is constructed, as shown in formula (9): x1(k+1)=A σ(k) x(k)+(1-φ c(k) (k))B σ(k) H c(k) (x(k)+a σ(k)c(k) ) (9) Wherein, σ(k) represents the switching signal in the information set of the SC deceiving the attacker, c(k) represents the switching signal transmitted to the controller in the information set, and the information set here is the information set of the SC deceiving the attacker, H c(k) represents the control gain, φ c(k) The expectation of (k) satisfies E{φ c(k) (k)}=ρ c(k) ρ c(k) φ c(k) (k) represents the action of CA deceiving the attacker; According to the information set of the SC deception attacker and the switching system equation under the SC deception attack, the index function of the SC deception attacker is constructed as shown in formulas (10) and (11): Among them, P σ(k) and S σ(k) is the weight matrix of SC deceiving the attacker and P σ(k) >0,S σ(k) >0,γ d is a constant related to the signal-to-noise ratio of the communication network channel and γ d >0; S24, according to the information set of the controller, obtain the action of the CA deceiving the attacker and construct the switching system equation of the controller, as shown in formula (12), Represents the system status information in the information set. Represents the controller in the information set, represents the control input information under CA spoofing attack; c(k) represents the switching signal transmitted to the controller in the information set; the information set here is the information set of the controller; The index function of the controller is constructed according to the information set of the controller and the switching system equation of the controller as shown in formula (13): Among them, Q c(k) , U c(k) is the weight matrix of the controller and Q c(k) >0,U c(k) >0; S25. According to the information set of the CA spoofing attacker, a switching system equation under the CA spoofing attack is constructed, as shown in formula (14): Among them, the switching system equation of the controller is consistent with the switching system equation under the CA deception attack, c(k) represents the switching signal transmitted to the controller in the information set; the information set here is the information set of the CA deception attacker; The indicator function of the CA deception attacker is constructed according to the information set of the CA deception attacker and the switching system equation under the CA deception attack, as shown in formula (15): Among them, G c(k) and R c(k) is the weight matrix of the CA spoofing attacker and G c(k) >0 and R c(k) >0.
4. The switching system security control method based on two-level multi-attacker Stackelberg game according to claim 3 is characterized in that: In the two-level multi-attacker Stackelberg game model, the controller and the CA deception attacker are modeled as the leader and follower in the first-level Stackelberg game, the SC deception attacker is modeled as the leader in the second-level game, and the controller and the CA deception attacker are jointly modeled as two followers in the second-level game.
5. The switching system security control method based on two-level multi-attacker Stackelberg game according to claim 4 is characterized in that: S3 combines the indicator functions of the CA deception attacker, the controller and the SC deception attacker, and adopts the reverse method to construct a two-level multi-attacker Stackelberg equilibrium strategy, wherein the two-level multi-attacker Stackelberg equilibrium strategy includes the optimal attack strategy of the CA deception attacker, the optimal control strategy of the controller and the optimal tampering strategy of the SC deception attacker; The two-level multi-attacker Stackelberg equilibrium strategy is used to control the behaviors of the SC deception attacker, the CA deception attacker and the controller, including: S31, the CA deceiving attacker minimizes the indicator function of the CA deceiving attacker according to the switching system equation of the CA deceiving attacker, as shown in formula (16), Assume that the current controller mode is c(k), solve formula (16) and obtain the optimal attack strategy for minimizing the CA deception attacker indicator function, as shown in formula (17): S32, the controller acts as a leader, obtains the optimal attack strategy of the CA deceiving the attacker, and minimizes the index function of the controller according to the optimal attack strategy of the CA deceiving the attacker and the switching system equation of the controller, as shown in formula (18), According to the optimal attack strategy of the CA deceiving attacker and the indicator function of the controller, formula (18) is solved to obtain the optimal strategy, as shown in formula (19): Satisfy constraints S33, the SC deception attacker acts as the common leader of the controller and the CA deception attacker, obtains the optimal attack strategy of the CA deception attacker and the optimal tampering strategy of the controller, and minimizes the indicator function of the SC deception attacker, as shown in formulas (20) and (21), The SC deception attacker solves the minimum value of the indicator function of formulas (20) and (21) according to the optimal attack strategy of the CA deception attacker, the optimal tampering strategy of the controller and the indicator function of the SC deception attack, and obtains the optimal tampering strategy, as shown in formulas (22) and (23). Satisfy constraints
Citation Information
Patent Citations
Network defense method based on multi-stage attack and defense signals
CN114024738A
Low-cost detection and isolation method for spoofing attack in cyber-physical system
CN115048625A
Unmanned surface vessel path tracking time triggering control method under spoofing attack task
CN115079698A
Network spoofing defense decision-making method and system based on Flipit intelligent game
CN116962050A
Method for predicting spoofing attack intention in multi-agent system based on inverse reinforcement learning
CN117155616A
