A trust-aware secure distributed tracking method for interactive multi-target systems

By employing a trust-aware, secure distributed tracing method, the problems of uncertain dynamic bias and spoofing attacks in interactive multi-target systems are solved, enabling reliable estimation of system state and improving system security and estimation accuracy.

CN122457288APending Publication Date: 2026-07-24HARBIN UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610317601.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-16
Publication Date
2026-07-24

Smart Images

  • Figure CN122457288A_ABST
    Figure CN122457288A_ABST
Patent Text Reader

Abstract

The application provides a trust-aware security distributed tracking method for an interactive multi-target system and belongs to the technical field of state estimation. The application has the following steps: a nonlinear random state space model is established; a memory attack detection mechanism based on trust awareness is designed, an attack detection function is constructed based on measurement residuals, and a limited memory window is used to realize real-time detection of whether the measurement data is attacked; three states of the measurement data are divided, and a measurement credibility factor is constructed; weighted fusion processing is performed to obtain effective measurement information; a security distributed state estimator is constructed to calculate one-step prediction of the system state; an upper bound of a one-step prediction error covariance matrix of each target state is recursively calculated; an estimator gain matrix of each target is calculated; the estimation of each target state is updated by using the gain matrix, and the upper bound of the estimation error covariance matrix is recursively calculated until a preset termination time. The application can improve the security, robustness and reliability of the system in a complex network environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to a trust-aware, secure, distributed tracking method for interactive multi-objective systems, belonging to the field of state estimation technology. Background Technology

[0002] Interactive multi-target systems are widely used in intelligent transportation, unmanned swarm collaboration, intelligent monitoring, and military sensing. Their core task is to continuously and accurately track and estimate the states of multiple interconnected targets. As system scale increases and the interactions between targets become more complex, traditional centralized tracking methods are showing shortcomings in terms of communication overhead, computational complexity, and system robustness, making it difficult to meet the comprehensive requirements of real-time performance, reliability, and scalability in complex dynamic environments. Therefore, distributed tracking methods are gradually becoming an effective technical means to improve the tracking performance of interactive multi-target systems.

[0003] With the development of information and communication technologies, a large number of sensor nodes in interactive multi-target systems typically rely on communication networks for data transmission. However, due to the highly interconnected and open nature of communication networks, measurement information is vulnerable to network attacks during transmission. For example, attackers can inject false data by tampering with or forging measurement data, thereby interfering with the system's state estimation process. This can lead to significant deviations in the estimation results of local nodes or even the entire system, severely impacting the overall performance and security of the system.

[0004] In the process of state tracking and estimation of interactive multi-target systems, relying solely on traditional attack detection methods often makes it difficult to ensure both system security and estimation accuracy. This is especially true when attacks are random and uncertain, which can easily lead to insufficient utilization of effective measurement information or misjudgment of attack detection.

[0005] Furthermore, existing state estimation methods often struggle to simultaneously address uncertain dynamic biases and spoofing attacks within a system. In interactive multi-objective systems, using traditional state estimation methods to estimate the states of multiple objectives can easily lead to decreased estimation accuracy or even estimation failure. Summary of the Invention

[0006] The purpose of this invention is to address the problem of simultaneously handling uncertain dynamic biases and spoofing attacks in existing technologies. It provides a trust-aware, secure distributed tracking method for interactive multi-target systems. By designing an attack detection mechanism with trust awareness capabilities and proposing a memory-based detection strategy based on historical information, it achieves dynamic evaluation of the reliability of measurement data, thereby improving the security, reliability, and robustness of distributed tracking in interactive multi-target systems. Furthermore, this invention designs a security state estimation algorithm within a distributed recursive estimation framework, enabling it to simultaneously cope with uncertain dynamic biases and spoofing attacks, thus possessing significant research value and engineering application significance.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] A trust-aware, secure, distributed tracking method for interactive multi-objective systems, the method comprising the following steps:

[0009] Step 1: Establish a nonlinear stochastic state-space model of the interactive multi-objective system, including: system state equations and measurement equations, and characterize the coupling relationship between the objectives in the system state equations;

[0010] Step 2: Based on the measurement equation established in Step 1, a trust-aware memory-based attack detection mechanism is designed to address potential spoofing attacks in the communication channel. The mechanism constructs an attack detection function based on the measurement residuals and uses a finite memory window to store historical residual information, thereby enabling real-time detection of whether the measurement data has been attacked.

[0011] Step 3: Based on the calculation results of the attack detection function in Step 2, the received measurement data is divided into three states: fully trustworthy measurement, partially trustworthy measurement, and fully untrustworthy measurement, and a measurement trustworthiness factor is constructed. The measurement data is then weighted and fused accordingly to obtain effective measurement information for state estimation.

[0012] Step 4: Based on the nonlinear stochastic state-space model established in Step 1 and the effective measurement information obtained in Step 3, construct a secure distributed state estimator, and use the estimator to calculate a one-step prediction of the system state. ;

[0013] Step 5: Based on the secure distributed state estimator constructed in Step 4, recursively calculate the upper bound of the one-step prediction error covariance matrix for each target state. ;

[0014] Step Six: Based on the upper bound of the one-step prediction error covariance matrix obtained in Step Five, calculate the estimator gain matrix for each target. ;

[0015] Step 7: Update the state estimates of each target using the gain matrix calculated in Step 6. And recursively calculate the upper bound of the estimation error covariance matrix. Until the preset termination time is reached.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0017] 1. This invention proposes a trust-aware, secure distributed tracking method for interactive multi-objective systems. Under a distributed estimation structure, the impact of uncertain dynamic bias and spoofing attacks on system state tracking performance is comprehensively considered. By designing a trust-aware, memory-based attack detection mechanism at the network transmission end, the method performs real-time judgment on whether measurement data has been attacked, and filters or weakens the measurement information based on the judgment result. This effectively reduces the impact of abnormal data on the state estimation process and improves the security, robustness, and reliability of the system in complex network environments.

[0018] 2. In the above method, a distributed recursive state estimation structure suitable for interactive multi-objective systems is constructed. By deriving the recursive matrix equation, a computable upper bound of the estimation error covariance matrix is ​​obtained. Based on the minimum variance criterion, the estimator gain matrix is ​​designed to minimize the trace of the upper bound of the estimation error covariance matrix, thereby achieving optimized state estimation performance. Because the proposed method has a clear recursive structure, it can achieve online estimation of the system state. Furthermore, its distributed nature avoids the high computational complexity and communication burden of centralized methods, offering advantages such as low computational complexity, simple implementation, and good scalability.

[0019] 3. The tracking method proposed in this invention can reliably estimate the state of an interactive multi-target system even in the presence of uncertain dynamic biases and spoofed data injection attacks, thereby effectively improving the tracking accuracy and estimation performance of the system under attack environments. Simulation results show that when the attack probability increases from 0.1 to 0.5, the mean cumulative absolute error increases by approximately 9.62%; when the attack probability increases from 0.5 to 0.9, the mean cumulative absolute error increases by approximately 28.7%. Furthermore, at the end of the simulation, the mean cumulative absolute error without an attack detection mechanism increases by approximately 5.89% compared to the case with an attack detection mechanism. These findings indicate that compared to estimation methods without attack detection mechanisms, the method proposed in this invention can significantly reduce the cumulative estimation error, further verifying the feasibility and effectiveness of this tracking method under attack environments. Attached Figure Description

[0020] Figure 1 This is a flowchart of a trust-aware, secure, distributed tracking method for interactive multi-objective systems, as described in this invention.

[0021] Figure 2 This is a comparison diagram of the actual trajectory and the tracking trajectory of target 1 in an interactive multi-target system.

[0022] Figure 3 This is a comparison diagram of the actual trajectory and the tracking trajectory of target 2 in an interactive multi-target system.

[0023] Figure 4 This is a comparison chart of the actual and estimated values ​​of each state component of target 3 in an interactive multi-objective system; where:

[0024] Figure 4 (a) is a comparison chart of the actual and estimated values ​​of the first state component of target 3;

[0025] Figure 4 (b) is a comparison chart of the actual and estimated values ​​of the second state component of target 3;

[0026] Figure 4 (c) is a comparison chart of the actual and estimated values ​​of the third state component of target 3.

[0027] Figure 5 This is a comparison chart of the actual and estimated values ​​of each state component of target 4 in an interactive multi-objective system; where:

[0028] Figure 5 (a) is a comparison chart of the actual and estimated values ​​of the first state component of target 4;

[0029] Figure 5 (b) is a comparison chart of the actual and estimated values ​​of the second state component of target 4;

[0030] Figure 5 (c) is a comparison chart of the actual and estimated values ​​of the third state component of target 4.

[0031] Figure 6 The logarithm of the mean square error of the trajectories of target 1 and target 2 Logarithm of the upper bound trace of the corresponding estimation error covariance matrix Comparison chart; where:

[0032] Figure 6 (a) is the logarithm of the mean square error of the trajectory of target 1. Logarithm of the upper bound trace of the corresponding estimation error covariance matrix Comparison chart;

[0033] Figure 6 (b) is the logarithm of the mean square error of the trajectory of target 2. Logarithm of the upper bound trace of the corresponding estimation error covariance matrix The comparison chart.

[0034] Figure 7 The logarithm of the mean square error of the trajectories of targets 3 and 4 Logarithm of the upper bound trace of the corresponding estimation error covariance matrix Comparison chart; where:

[0035] Figure 7 (a) is the logarithm of the mean square error of the trajectory of target 3. Logarithm of the upper bound trace of the corresponding estimation error covariance matrix Comparison chart;

[0036] Figure 7 (b) is the logarithm of the mean square error of the trajectory of target 4. Logarithm of the upper bound trace of the corresponding estimation error covariance matrix The comparison chart.

[0037] Figure 8 The mean cumulative absolute error of four targets under different probabilities of injecting fake data The comparison chart.

[0038] Figure 9 The mean cumulative absolute error of the four targets under the conditions of having and not having a trust-aware attack detection mechanism. The comparison chart. Detailed Implementation

[0039] The present invention will be further described in detail below with reference to the accompanying drawings: This embodiment is implemented based on the technical solution of the present invention and provides detailed implementation methods, but the protection scope of the present invention is not limited to the following embodiments.

[0040] like Figure 1 As shown in the figure, the trust-aware secure distributed tracking method for interactive multi-target systems involved in this embodiment includes the following steps:

[0041] Step 1: Establish a nonlinear stochastic state-space model of the interactive multi-objective system, including: system state equations and measurement equations, and characterize the coupling relationship between the objectives in the system state equations;

[0042] Step 11: Establish the system state equations for the interactive multi-objective system, whose expression is:

[0043] (1)

[0044] In the formula, Indicates the first One goal is The state vector at time t, with superscript " " indicates matrix transpose. For target labeling, Indicates the number of targets. Represents the first in the Cartesian coordinate system One goal is Location information at any given time Represents the first in the Cartesian coordinate system One goal is Directional information at any given time; Indicates the first One goal is The state vector at time t, Represents the first in the Cartesian coordinate system One goal is Location information at any given time Represents the first in the Cartesian coordinate system One goal is Directional information at any given time; and They represent the first One goal is Displacement velocity and angular velocity at any given moment; and These represent the sine and cosine functions, respectively; " is the summation symbol; Indicates the first The first goal and the first The coupling weights between objectives are used to characterize the interactive effects between different objectives. This is the inner coupling matrix in an interactive multi-target system, used to describe the interaction between target position information and orientation information; Indicates the first One goal is The state vector at time t, Represents the first in the Cartesian coordinate system One goal is Location information at any given time Represents the first in the Cartesian coordinate system One goal is Directional information at any given time; express The dynamic deviation intensity matrix in the time-series system; express The evolution of uncertain dynamic deviations in a time-series system satisfies a dynamic equation. , express Uncertain dynamic deviations in a time-based system Indicates the deviation term The state transition matrix, Let it be a disturbance term, and satisfy... and ,in For a given positive scalar, For an identity matrix of appropriate dimension, for The transpose of the matrix, for" "expectations"; The mean is zero and the covariance matrix is Gaussian white noise; for The noise intensity matrix in the time-series system; For zero-mean process noise in an interactive multi-objective system, its covariance matrix is: ;

[0045] Step 1 and Step 2: Establish the measurement equations for the interactive multi-objective system, whose expression is:

[0046] (2)

[0047] In the formula, For the first One goal is The measurement output at time, The first one was obtained by the airborne camera. One goal is The image plane coordinates at that moment; For the first The coordinates of each camera in the target's local coordinate system; The coordinates of the visual features in the world coordinate system; The distance from the camera's optical center to the ceiling plane; and This is the pixel magnification factor; The principal point position of the camera in the image plane; For zero-mean measurement noise in an interactive multi-objective system, its covariance matrix is: .

[0048] Step 2: Based on the measurement equation established in Step 1, a trust-aware memory-based attack detection mechanism is designed to address potential spoofing attacks in the communication channel. The mechanism constructs an attack detection function based on the measurement residuals and uses a finite memory window to store historical residual information, thereby enabling real-time detection of whether the measurement data has been attacked.

[0049] Step 2.1: Based on the interactive multi-target system measurement equations established in Step 1, a model is constructed for spoofing data injection attacks in the communication channel. The attack model is expressed as follows:

[0050] (3)

[0051] In the formula, For the first One goal is The measurement output after being transmitted through the communication channel at any time; For a bounded attack signal, satisfying ,in Given a positive scalar, for The transpose of the matrix; Let be a set of random variables following a Bernoulli distribution, used to characterize whether an attack is activated, and its probability distribution satisfies ... and ,in, For the event "The probability of occurrence, i.e. For the first One goal is The probability that the measurement output at any given time is subject to a spoofed data injection attack. For the first One goal is The probability that the measurement output at a given time moment is subject to a spoofed data injection attack. Used to characterize the uncertainty in the probability of an attack occurring, and satisfying , For a given nonnegative scalar, express" The absolute value of ";

[0052] Step 22: Construct an attack detection function based on the measurement residuals, the expression of which is:

[0053] (4)

[0054] In the formula, For the residual vector, for Time for nonlinear function A step of prediction, , for Time for the first A one-step prediction of a target state. For an identity matrix of appropriate dimension, for The transpose of the matrix; Let be a known weighted matrix; where

[0055]

[0056] Steps 2 and 3: Introduce a finite memory window to store measurement residuals that have recently been determined not to have been attacked, in order to improve the reliability of attack detection.

[0057] The length of the finite memory window The preset positive integer is used when the amount of residual data stored in the cache reaches the window length. At that time, the oldest stored residual data is deleted, and the latest residual is added to the cache, thereby achieving dynamic updates of the cache content. Furthermore, let... Indicates time The latest moment that was previously determined to be attack-free has the following update rules:

[0058] (5)

[0059] In the formula, As an indicator variable, it is used to represent the first... One goal is Is the time-based measurement output under attack? express The next moment after the current moment that is determined to be without attack; It represents the infimum of a set.

[0060] Step 3: Based on the calculation results of the attack detection function in Step 2, the received measurement data is divided into three states: fully trustworthy measurement, partially trustworthy measurement, and fully untrustworthy measurement, and a measurement trustworthiness factor is constructed. The measurement data is then weighted and fused accordingly to obtain effective measurement information for state estimation.

[0061] Based on the calculation results of the attack detection function, the processing of measurement data includes the following three scenarios:

[0062] First scenario: Fully reliable measurement

[0063] When attack detection function When the following first condition is met:

[0064] (6)

[0065] think The received measurement outputs at any given time have not been attacked and are in a completely trustworthy state. At this point, let... And the current measurement value is directly used for state estimation. Where, For a length of Within the limited memory window, the first The mean of the target residual vectors, For the length of a finite memory window, For the first One goal is The residual vector at time step; Let be a column vector composed of the standard deviations of each component, where the _i_i is the _i ... Each component Defined as , and These are the residual vectors. and its mean vector The One portion, , To measure the dimension of the output, For time index variables within a finite memory window, , for The column vector formed; A pre-defined positive scalar threshold; for The transpose of the matrix; for The transpose of the matrix; for The square of; This indicates the square root operation.

[0066] Second scenario: Partially reliable measurement

[0067] When attack detection function When the following second condition is met:

[0068] (7)

[0069] think The measurement output received at any given time is in a questionable state and is classified as a partially reliable measurement. At this time, let... ,in The confidence factor is used to characterize the validity of the current measurement information. These are the pre-set initial parameters; A constant in the interval 0 to 1; For A logarithmic function with base 0; Let be the vector norm. In the formula, For a pre-set positive scalar threshold, and satisfying .

[0070] In partially reliable measurement scenarios, a weighted fusion mechanism is used to update the measurement data, and the update format is as follows: ,in For the first An estimator in The final measurement value used at that time.

[0071] Third scenario: Completely unreliable measurement

[0072] When attack detection function When the following third condition is met:

[0073] (8)

[0074] think The measurement output received at any given time is under attack, making it a completely untrusted measurement. At this point... It discards the current measurement data and does not participate in the state estimation update.

[0075] To compensate for the above situation, the present invention adopts a zero-order hold strategy, which enables the estimator to perform state estimation at the current time using the most recently determined reliable measurement data.

[0076] By integrating the above three scenarios, the first... An estimator in The final measured value used at time is expressed as follows:

[0077] (9)

[0078] Step 4: Based on the nonlinear stochastic state-space model established in Step 1 and the effective measurement information obtained in Step 3, construct a secure distributed state estimator, and use the estimator to calculate a one-step prediction of the system state. ;

[0079] Step 41: Define the following symbols:

[0080]

[0081] In the formula, For nonlinear functions in interactive multi-objective systems; For the state vector Dynamic deviation and measurement output The augmented state vector; For the reason The nonlinear function obtained by augmenting the zero vector; For the zero vector and The augmented nonlinear function; For the dynamic deviation intensity matrix State transition matrix of deviation term Indicator variable matrix And the system matrix composed of the appropriate-dimensional zero matrix; For process noise Gaussian noise Measurement noise The noise signal obtained by augmentation; This is the augmented attack signal; For the indicator variable matrix The system matrix is ​​composed of the appropriate-dimensional zero matrix; The system matrix is ​​composed of the identity matrix and the zero matrix of appropriate dimension; For an appropriate dimension zero matrix and perturbation terms And a block diagonal matrix formed by using the zero matrix of the appropriate dimension as diagonal block elements; For the inner coupling strength matrix And a block diagonal matrix formed by using the zero matrix of the appropriate dimension as diagonal block elements; For the noise intensity matrix The appropriate-dimensional identity matrix and indicator variable matrix A block diagonal matrix composed of diagonal block elements;

[0082] Step 4.2, for the first objective in an interactive multi-objective system To achieve the following objectives, a secure distributed state estimator is constructed:

[0083] (10)

[0084] (11)

[0085] In the formula, for Time for the first A one-step prediction of a target state; for Time of the first The objective is based on the nonlinear function of an interactive multi-objective system. The estimation form; for Time of the first The objective is based on the nonlinear function of an interactive multi-objective system. The estimation form; for Time for the first State estimation of each objective; for Time for the first State estimation of each objective; for Time for the first State estimation of each objective; for Time of the first Gain matrix for estimating target state; For the first An estimator in The final measurement value used at that time; For the indicator variable matrix The system matrix is ​​composed of the appropriate-dimensional zero matrix; for Time of the first The objective is based on the nonlinear function of an interactive multi-objective system. The predictive form.

[0086] Step 5: Based on the secure distributed state estimator constructed in Step 4, recursively calculate the upper bound of the one-step prediction error covariance matrix for each target state. ;

[0087] Upper bound of the one-step prediction error covariance matrix for each target state Calculate using the following formula:

[0088] (12)

[0089] in,

[0090]

[0091]

[0092] In the formula, , , , , , , , , , , , , , and The scaling parameters are known. , , , , , , , , , , , , , and Scaling parameters , , , , , , , , , , , , , and The reciprocal of; for Time of the first The nonlinear function corresponding to the state of each objective In estimation Jacobian matrix at the location; and nonlinear function The known error matrix obtained based on the Taylor expansion formula; for Time of the first The upper bound of the one-step prediction error covariance matrix for each target state; for Time of the first The upper bound of the estimation error covariance matrix of each target state; for The inverse matrix; It is a positive scalar and satisfies ; for The reciprocal of; for The inverse matrix; for The square of; for Time of the first The nonlinear function corresponding to the state of each objective In estimation Jacobian matrix at the location; and nonlinear function The known error matrix obtained based on the Taylor expansion formula; It is a positive scalar and satisfies ; for The reciprocal of; for The inverse matrix; The upper bound of the estimated error covariance matrix is ​​given by... and state estimation The intermediate matrix formed; intermediate variables For the first The sum of the coupling weights between each objective and other objectives; For process noise covariance matrix Gaussian noise covariance matrix Measurement noise covariance matrix A block diagonal matrix composed of diagonal block elements; for Time of the first The upper bound of the estimation error covariance matrix of each target state; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; It is an identity matrix with appropriate dimensions; Indicates to " "Seeking traces."

[0093] Step Six: Based on the upper bound of the one-step prediction error covariance matrix obtained in Step Five, calculate the estimator gain matrix for each target. ;

[0094] Estimator gain matrix for each target Calculate using the following formula:

[0095] (13)

[0096] in,

[0097]

[0098] In the formula, , and Given the scaling parameters, , and They are respectively , and The reciprocal of; for The transpose of the matrix; for Time of the first The nonlinear function corresponding to the state of each objective In prediction Jacobian matrix at the location; and nonlinear function The known error matrix obtained based on the Taylor expansion formula; It is a positive scalar and satisfies ; for The reciprocal of; for The inverse matrix; for The inverse matrix; For process noise covariance matrix Gaussian noise covariance matrix Measurement noise covariance matrix A block diagonal matrix composed of diagonal block elements; For the first The measurement output of each target is in The probability of being subjected to a fake data injection attack at any time; for The inverse matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of .

[0099] Step 7: Update the state estimates of each target using the gain matrix calculated in Step 6. And recursively calculate the upper bound of the estimation error covariance matrix. Until the preset termination time is reached.

[0100] Upper bound of the estimation error covariance matrix for each target state Calculate using the following formula:

[0101] (14)

[0102] In the formula, for Time of the first The upper bound of the estimation error covariance matrix of each target state; for The transpose of the matrix; for The transpose of .

[0103] Example 1:

[0104] This embodiment uses the distributed tracking problem of an interactive multi-object system as an example to simulate and verify the method proposed in this invention.

[0105] This invention selects a system with four objectives for simulation. The system parameters are as follows: the initial states of the four objectives are respectively... , , and "T" represents the transpose of the vector; the initial values ​​of the dynamic deviations of the four targets are respectively , , and The initial state estimates of the estimator are the same as the initial values ​​of the target states; the initial values ​​of the upper bound of the estimation error covariance matrix of the four targets are... , , It is a seven-dimensional identity matrix; the dynamic deviation intensity matrices are respectively , , and ; Indicates that the element " The diagonal matrix formed by these targets; the coupling weights between the targets are... , The inner coupling matrix is , The matrix is ​​a three-dimensional identity matrix; the noise intensity matrices are respectively...

[0106]

[0107] Displacement velocity , angular velocity , The camera's coordinates in the target's local coordinate system are: The distance from the camera's optical center to the ceiling plane is ; pixel magnification factor is and The principal point position of the camera in the image plane is: The state transition matrix for the deviation term is: , It is a two-dimensional identity matrix, a positive scalar. , The covariance matrix of the process noise is , The covariance matrix of the measured noise is: , The covariance matrix of Gaussian noise is , The probability of a deduced data injection attack is: positive scalar nonnegative scalar , The finite memory window length of attack detection mechanisms Weighted matrix , The positive scalar thresholds are respectively and , The initial value for measuring the confidence factor is... , The scaling parameter is and Nonlinear functions The error matrix obtained at the estimation point based on the Taylor expansion formula is: , Nonlinear functions The error matrix obtained at the estimation point based on the Taylor expansion formula is: , Nonlinear functions The error matrix obtained at the prediction point based on the Taylor expansion formula is: and , .

[0108] This invention uses mean squared error to evaluate the performance of the tracking algorithm. Indicates the first One goal is The mean square error at time t is calculated using the following formula:

[0109]

[0110] In the formula, The number of Monte Carlo experiments; and They represent the first In the experiment, the first One goal is The actual state and the estimated state at any given time; for The transpose of the matrix. Furthermore, to further analyze the algorithm's performance, cumulative absolute error is used for comparison. Indicates the first One goal is The cumulative absolute error at time t is calculated using the following formula:

[0111]

[0112] In the formula, For the first One goal is Time estimation; Represents the vector norm. express The mean of the cumulative absolute errors of the four target state estimates at time 1.

[0113] The following are the multi-target tracking results using the trust-aware secure distributed tracking method for interactive multi-target systems proposed in this invention:

[0114] like Figure 2 and Figure 3 As shown, the attack probabilities are respectively given. The motion tracking trajectories of target 1 and target 2 are shown in the figure. As can be seen from the figure, even under the influence of false data injection attacks, the proposed distributed tracking method with trust awareness can still estimate the state of each target in the interactive multi-target system in real time and effectively, demonstrating good tracking performance.

[0115] like Figure 4 and Figure 5 As shown, the attack probabilities are respectively given. The estimation results of the three state components of targets 3 and 4 are shown in the figure. As can be seen from the figure, the proposed estimation method has good estimation accuracy at each state component level.

[0116] like Figure 6 and Figure 7 As shown, the logarithm of the mean square error of the four target trajectories is given respectively. And the logarithm of the upper bound trace of the corresponding estimation error covariance matrix. The comparison results show that the mean squared error is consistently lower than its upper bound, thus verifying the effectiveness of the proposed method.

[0117] like Figure 8 As shown, the mean cumulative absolute error of four targets under different probabilities of fake data injection attacks is presented. As shown in the graph, with increasing attack probability... The increase, The obvious upward trend indicates that the probability of attack is high. The higher the probability, the lower the accuracy of the system's state estimation. Furthermore, when the attack probability... The mean cumulative absolute error of the four target state estimates at the simulation termination time is calculated when the values ​​are 0.1, 0.5, and 0.9, respectively. The values ​​were 257.515, 282.287, and 363.303, respectively. When the attack probability increased from 0.1 to 0.5, the mean cumulative absolute error increased by approximately 9.62%; when the attack probability increased from 0.5 to 0.9, the mean cumulative absolute error increased by approximately 28.7%.

[0118] like Figure 9As shown, the mean cumulative absolute error of four targets under the conditions of having and not having a trust-aware attack detection mechanism is given. As shown in the figure, the tracking method incorporating a trust-aware attack detection mechanism significantly outperforms the tracking method without such a mechanism in terms of estimation accuracy. This indicates that the designed attack detection mechanism can effectively mitigate the impact of attacks on state estimation performance, thereby improving the system's estimation accuracy. Furthermore, at the simulation termination time, the mean cumulative absolute error of the four target state estimations is 286.455 with attack detection and 303.316 without, representing an increase of approximately 5.89%.

[0119] In summary, the trust-aware security distributed tracking method for interactive multi-target systems proposed in this invention can still effectively estimate the target state of the system even in the presence of uncertain dynamic biases and false data injection attacks.

[0120] The above description is merely a preferred embodiment of the present invention. These specific embodiments are different implementations based on the overall concept of the present invention, and the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A trust-aware, secure, distributed tracking method for interactive multi-objective systems, characterized in that, The method includes the following steps: Step 1: Establish a nonlinear stochastic state-space model of the interactive multi-objective system, including: system state equations and measurement equations, and characterize the coupling relationship between the objectives in the system state equations; Step 2: Based on the measurement equation established in Step 1, a trust-aware memory-based attack detection mechanism is designed to address potential spoofing attacks in the communication channel. The mechanism constructs an attack detection function based on the measurement residuals and uses a finite memory window to store historical residual information, thereby enabling real-time detection of whether the measurement data has been attacked. Step 3: Based on the calculation results of the attack detection function in Step 2, the received measurement data is divided into three states: fully trustworthy measurement, partially trustworthy measurement, and fully untrustworthy measurement, and a measurement trustworthiness factor is constructed. The measurement data is then weighted and fused accordingly to obtain effective measurement information for state estimation. Step 4: Based on the nonlinear stochastic state-space model established in Step 1 and the effective measurement information obtained in Step 3, construct a secure distributed state estimator, and use the estimator to calculate a one-step prediction of the system state. ; Step 5: Based on the secure distributed state estimator constructed in Step 4, recursively calculate the upper bound of the one-step prediction error covariance matrix for each target state. ; Step Six: Based on the upper bound of the one-step prediction error covariance matrix obtained in Step Five, calculate the estimator gain matrix for each target. ; Step 7: Update the state estimates of each target using the gain matrix calculated in Step 6. And recursively calculate the upper bound of the estimation error covariance matrix. Until the preset termination time is reached.

2. The trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 1, characterized in that, Step one includes the following steps: Step 11: Establish the system state equations for the interactive multi-objective system, whose expression is: In the formula, Indicates the first One goal is The state vector at time t, Indicates matrix transpose. For target labeling, Indicates the number of targets. Represents the first in the Cartesian coordinate system One goal is Location information at any given time Represents the first in the Cartesian coordinate system One goal is Directional information at any given time; Indicates the first One goal is The state vector at time t, Represents the first in the Cartesian coordinate system One goal is Location information at any given time Represents the first in the Cartesian coordinate system One goal is Directional information at any given time; and They represent the first One goal is Displacement velocity and angular velocity at any given moment; Indicates the first The first goal and the first The coupling weights between objectives are used to characterize the interactive effects between different objectives. This is the inner coupling matrix in an interactive multi-target system, used to describe the interaction between target position information and orientation information; Indicates the first One goal is The state at any given moment, Represents the first in the Cartesian coordinate system One goal is Location information at any given time Represents the first in the Cartesian coordinate system One goal is Directional information at any given time; express The dynamic deviation intensity matrix in the time-series system; express The evolution of uncertain dynamic deviations in a time-series system satisfies a dynamic equation. , express Uncertain dynamic deviations in a time-based system Indicates the deviation term The state transition matrix, Let it be a disturbance term, and satisfy... and ,in For a given positive scalar, For an identity matrix of appropriate dimension, for The transpose of the matrix, Expressing expectations; The mean is zero and the covariance matrix is Gaussian white noise; for The noise intensity matrix in the time-series system; For zero-mean process noise in an interactive multi-objective system, its covariance matrix is: ; Step 1 and Step 2: Establish the measurement equations for the interactive multi-objective system, whose expression is: In the formula, For the first One goal is The measurement output at time, The first one was obtained by the airborne camera. One goal is The image plane coordinates at that moment; For the first The coordinates of each camera in the target's local coordinate system; The coordinates of the visual features in the world coordinate system; The distance from the camera's optical center to the ceiling plane; and This is the pixel magnification factor; The principal point position of the camera in the image plane; For zero-mean measurement noise in an interactive multi-objective system, its covariance matrix is: .

3. The trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 2, characterized in that, In step two, a model is constructed for spoofed data injection attacks occurring in the communication channel. The attack model is represented as follows: In the formula, For the first One goal is The measurement output after being transmitted through the communication channel at any time; For a bounded attack signal, satisfying ,in Given a positive scalar, for The transpose of the matrix; Let be a set of random variables following a Bernoulli distribution, used to characterize whether an attack is activated, and its probability distribution satisfies ... and ,in, The probability of an event occurring, i.e. For the first One goal is The probability that the measurement output at any given time is subject to a spoofed data injection attack. For the first One goal is The probability that the measurement output at a given time moment is subject to a spoofed data injection attack. Used to characterize the uncertainty in the probability of an attack occurring, and satisfying , For a given nonnegative scalar.

4. The trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 3, characterized in that, The attack detection function constructed based on the measurement residuals in step two is expressed as follows: In the formula, For the residual vector, for Time for nonlinear function A step of prediction, , for Time for the first A one-step prediction of a target state. For an identity matrix of appropriate dimension, for The transpose of the matrix; The weighting matrix is ​​known. in 。 5. A trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 4, characterized in that, The second step involves using a finite memory window to store historical residual information, wherein the historical residual information is measurement residual information that has recently been determined not to have been attacked; The length of the finite memory window The preset positive integer is used when the amount of residual data stored in the cache reaches the window length. When the time comes, the oldest stored residual data is deleted and the latest residual is added to the cache, thereby achieving dynamic updates of the cache content; make Indicates time The latest moment that was previously determined to be attack-free has the following update rules: In the formula, As an indicator variable, it is used to represent the first... One goal is Is the time-based measurement output under attack? express The next moment after the current moment that is determined to be without attack; It represents the infimum of a set.

6. A trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 5, characterized in that, In step three, based on the calculation results of the attack detection function, the received measurement data is divided into three states: fully trustworthy measurement, partially trustworthy measurement, and fully untrustworthy measurement, and a measurement trustworthiness factor is constructed. The measurement data is then weighted and fused accordingly to obtain effective measurement information for state estimation. The specific method is as follows: Scenario 1: Fully Reliable Measurement When attack detection function When the following first condition is met: think The received measurement outputs at any given time have not been attacked and are in a completely trustworthy state. At this point, let... And the current measurement value is directly used for state estimation, where, For a length of Within the limited memory window, the first The mean of the target residual vectors, For the length of a finite memory window, For the first One goal is The residual vector at time step; Let be a column vector composed of the standard deviations of each component, where the _i_i is the _i ... Each component Defined as , and These are the residual vectors. and its mean vector The One portion, , To measure the dimension of the output, For time index variables within a finite memory window, , for The column vector formed; A pre-defined positive scalar threshold; for The transpose of the matrix; for The transpose of the matrix; Second scenario: Partially reliable measurement When attack detection function When the following second condition is met: think The received measurement output is in a questionable state and is classified as a partially reliable measurement. At this point, let... ,in The confidence factor is used to characterize the validity of the current measurement information. These are the pre-set initial parameters; A constant within the interval 0 to 1; For A logarithmic function with base 0; Let be the vector norm, where For a pre-defined positive scalar, and satisfying ; In partially reliable measurement scenarios, a weighted fusion mechanism is used to update the measurement data, and the update format is as follows: ,in For the first An estimator in The final measurement value used at that time; Third scenario: Completely unreliable measurement When attack detection function When the following third condition is met: think If the received measurement output is under attack and is considered a completely untrusted measurement, then... And discard the current measurement data, and do not participate in the state estimation update; To compensate for the above situation, a zero-order hold strategy is adopted, which enables the estimator to perform state estimation at the current time using the most recently determined reliable measurement data; By integrating the above three scenarios, the first... An estimator in The final measured value used at time is expressed as follows: 。 7. A trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 6, characterized in that, Step four involves constructing a secure distributed state estimator based on the nonlinear stochastic state-space model established in step one and the effective measurement information obtained in step three. The estimator is then used to calculate a one-step prediction of the system state. The specific steps are as follows: Step 41: Define the following symbols: In the formula, For nonlinear functions in interactive multi-objective systems; For the state vector Dynamic deviation and measurement output The augmented state vector; For the reason The nonlinear function obtained by augmenting the zero vector; For the zero vector and The augmented nonlinear function; For the dynamic deviation intensity matrix State transition matrix of deviation term Indicator variable matrix And the system matrix composed of the appropriate-dimensional zero matrix; For process noise Gaussian noise With measurement noise The noise signal obtained by augmentation; This is the augmented attack signal; For the indicator variable matrix The system matrix is ​​composed of the appropriate-dimensional zero matrix; The system matrix is ​​composed of the identity matrix and the zero matrix of appropriate dimension; For an appropriate dimension zero matrix and perturbation terms And a block diagonal matrix formed by using the zero matrix of the appropriate dimension as diagonal block elements; For the inner coupling strength matrix And a block diagonal matrix formed by using the zero matrix of the appropriate dimension as diagonal block elements; For the noise intensity matrix The appropriate-dimensional identity matrix and indicator variable matrix A block diagonal matrix composed of diagonal block elements; Step 4.2, for the first objective in an interactive multi-objective system To achieve the following objectives, a secure distributed state estimator is constructed: In the formula, for Time for the first A one-step prediction of a target state; for Time of the first The objective is based on the nonlinear function of an interactive multi-objective system. The estimation form; for Time of the first The objective is based on the nonlinear function of an interactive multi-objective system. The estimation form; for Time for the first State estimation of each objective; for Time for the first State estimation of each objective; for Time for the first State estimation of each objective; for Time of the first Gain matrix for estimating target state; For the first An estimator in The final measurement value used at that time; For the indicator variable matrix The system matrix is ​​composed of the appropriate-dimensional zero matrix; for Time of the first The objective is based on the nonlinear function of an interactive multi-objective system. The predictive form.

8. A trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 7, characterized in that, In step five, based on the secure distributed state estimator built in step four, the upper bound of the one-step prediction error covariance matrix for each target state is recursively calculated. The calculation formula is as follows: in, In the formula, , , , , , , , , , , , , , and The scaling parameters are known. , , , , , , , , , , , , , and Scaling parameters , , , , , , , , , , , , , and The reciprocal of; for Time of the first The nonlinear function corresponding to the state of each objective In estimation Jacobian matrix at the location; and nonlinear function The known error matrix is ​​obtained based on the Taylor expansion formula; for Time of the first The upper bound of the one-step prediction error covariance matrix for each target state; for Time of the first The upper bound of the estimation error covariance matrix of each target state; for The inverse matrix; It is a positive scalar and satisfies ; for The reciprocal of; for The inverse matrix; for The square of; for Time of the first The nonlinear function corresponding to the state of each objective In estimation Jacobian matrix at the location; and nonlinear function The known error matrix is ​​obtained based on the Taylor expansion formula; It is a positive scalar and satisfies ; for The reciprocal of; for The inverse matrix; The upper bound of the estimated error covariance matrix. and state estimation The intermediate matrix formed; intermediate variables For the first The sum of the coupling weights between each objective and other objectives; For process noise covariance matrix Gaussian noise covariance matrix With measurement noise covariance matrix A block diagonal matrix composed of diagonal block elements; for Time of the first The upper bound of the estimation error covariance matrix of each target state; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; It is an identity matrix with appropriate dimensions; It indicates the search for a trace.

9. A trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 8, characterized in that, Step six involves calculating the estimator gain matrix for each target based on the upper bound of the one-step prediction error covariance matrix obtained in step five. The calculation formula is as follows: in, In the formula, , and Given the scaling parameters, , and They are respectively , and The reciprocal of; for The transpose of the matrix; for Time of the first The nonlinear function corresponding to the state of each objective In prediction Jacobian matrix at the location; and nonlinear function The known error matrix is ​​obtained based on the Taylor expansion formula; It is a positive scalar and satisfies ; for The reciprocal of; for The inverse matrix; for The inverse matrix; For process noise covariance matrix Gaussian noise covariance matrix With measurement noise covariance matrix A block diagonal matrix composed of diagonal block elements; For the first The measurement output of each target is in The probability of being subjected to a fake data injection attack at any time; for The inverse matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of the matrix; for The transpose of .

10. A trust-aware secure distributed tracking method for interactive multi-objective systems according to claim 9, characterized in that, Step seven involves updating the target state estimates using the gain matrix calculated in step six. And recursively calculate the upper bound of the estimation error covariance matrix. The process continues until a preset termination time is reached, as follows: Upper bound of the estimation error covariance matrix for each target state Calculate according to the following formula: In the formula, for Time of the first The upper bound of the estimation error covariance matrix of each target state; for The transpose of the matrix; for The transpose of .