A method for evaluating the impact of network attack performance on weapon equipment based on sparse model and ADMM distributed parallel data mining

Through the method based on sparse model and ADMM distributed parallel data mining, a method for evaluating the performance of cyberattacks of weapons and equipment was established, which solved the problems of low accuracy and low efficiency in the existing technology, and achieved the impact on the performance of cyberattacks. Improve accuracy and efficiency.

CN118250085BActive Publication Date: 2024-11-05INFORMATION CENT OF CHINA NORTH IND GRP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410566922.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-09
Publication Date
2024-11-05
Estimated Expiration
2044-05-09

AI Technical Summary

Technical Problem

The existing technology is difficult to effectively evaluate the multi-faceted performance impact of cyber attacks on weapons and equipment, and traditional machine learning algorithms perform poorly in large amounts of data, which cannot improve execution efficiency.

Method used

Using a method based on sparse model and ADMM distributed parallel data mining, an index system for equipment network attack performance impact assessment is established, the index system is sparse through the coefficient of variation method, the time attenuation function is introduced, and the ADMM algorithm is used for global optimization, and it is decomposed into multiple subs. Problem, iterative updates to obtain the minimum value of the performance impact evaluation model.

Benefits of technology

It has achieved the accuracy improvement of the impact on the performance of network attacks, solved the problems of sparseness of indicator systems and lightweighting of models, improved the efficiency and accuracy of the evaluation method, made the evaluation results more reasonable and accurate, and adapted to the distributed computing model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118250085B_ABST
    Figure CN118250085B_ABST
Patent Text Reader

Abstract

The invention discloses a weapon equipment network attack performance impact assessment method based on a sparse model and ADMM distributed parallel data mining, including: establishing an equipment network attack performance impact assessment index system according to potential objects that may be subjected to network attacks in the equipment; sparsely processing the assessment index system by the coefficient of variation method to obtain a core index system; constructing a weapon equipment network attack performance impact assessment model according to the core index system; in the impact assessment model, citing a time decay function to obtain an improved impact assessment model; using the ADMM algorithm to optimize the minimum value of the improved impact assessment model, perform global problem optimization and split it into multiple sub-problems; iteratively update the sub-problem variables until convergence, and obtain the minimum value of the performance impact assessment model function as the optimal solution for the weapon equipment network attack performance impact. The method effectively solves the problems of lightweight assessment model and large-scale computing efficiency improvement, and the assessment value is more accurate.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of weapon equipment network attack performance impact assessment, and more specifically to a weapon equipment network attack performance impact assessment method based on a sparse model and ADMM distributed parallel data mining. Background Art

[0002] At present, with the development of mechanization, informatization and intelligence of weapons and equipment, there are more and more embedded software systems on the equipment, and the degree of informatization of the equipment is getting higher and higher. These software systems rely on wired or wireless networks for data interaction, which also makes weapons and equipment face more and more severe network threats. When weapons and equipment are attacked by network, the performance impact assessment after the attack is an important basis for providing military personnel with equipment application plans. PHM (Prognostics and Health Management) technology can evaluate the performance of weapons and equipment, and its methods can be used as a reference. PHM technology was born at the end of the last century to meet the requirements of information warfare for weapons and equipment to be fast, reliable and accurate, and quickly received high attention from Western military powers represented by the United States. PHM has become a core technology for the development of a new generation of foreign weapons and equipment and the realization of autonomous support. It is a very promising military and civilian dual-use technology for improving the "six properties" of complex systems and reducing life cycle costs in the 21st century. With the development of mechanization, informatization and intelligence of modern weapons and equipment in my country, its structure and mission requirements are becoming increasingly complex, and higher and higher requirements are put forward for its safety and reliability, and its maintenance and support costs are also increasing. The breakthrough of PHM technology and the development and application of system architecture have important practical significance for improving the comprehensive support decision-making ability and combat effectiveness of weapons and equipment and reducing the maintenance and repair costs of weapons and equipment throughout their life cycle.

[0003] From the perspective of method mechanism, PHM technology is mainly divided into three categories, including physical model-based, empirical model-based and data-driven methods. Physical model-based methods, such as statistical filtering methods such as Kalman filtering and particle filtering, can achieve relatively accurate fault diagnosis and life prediction, but the premise is that an accurate physical or mathematical model needs to be established for the monitored target object, which is often difficult to achieve. Empirical model-based methods are highly dependent on the experience of experts. Once the expert experience is prioritized or deviated, it will have a huge impact on the fault diagnosis and prediction results. Data-driven methods can solve the dependence on models and experience very well, and therefore have received widespread attention and research, but data-driven methods require test and monitoring historical data of the monitored object at all stages.

[0004] At present, some scholars have proposed to organically integrate two or three categories of model-based, expert knowledge-based and data-driven methods, make full use of the advantages of various methods, and achieve more accurate diagnosis / prediction results. This will also be the development trend of PHM technology. Chen Feng et al. used thermal parameters based on the basic database of the equipment integrated monitoring system server to achieve automatic online fault diagnosis of equipment operation status through neural network technology, and can dynamically adjust the knowledge base to adapt to different fault models and diagnosis objects. In order to solve the problem of organic integration of fault diagnosis system and other ship systems, Zheng Yuanzhang used the advantages of neural network and network platform to effectively diagnose and predict equipment, and provide auxiliary decision-making for equipment maintenance and repair management. In order to solve the problem of real-time online fault diagnosis of equipment, Yin Honghao et al. modeled the fault model of ship centrifugal pump and conducted real-time online intelligent fault diagnosis based on SOM (self-organizing feature mapping neural network). Ji Zhe analyzed and mined data based on BP neural network to achieve equipment fault diagnosis, which is feasible.

[0005] Although the rapid development of artificial intelligence and machine learning technology has provided a new solution for the performance impact assessment of network attacks on equipment, the above methods have not established a connection with network attacks, making it difficult to measure the impact of network attack behaviors on equipment. In addition, due to theoretical design limitations, most of the traditional machine learning algorithms used are serial algorithms that can only run in a single-machine environment and perform poorly when faced with large amounts of data. With the development of parallelization technology, how to improve the execution efficiency of algorithms when faced with large amounts of running data for weapons and equipment has become a hot topic in industry research. Summary of the invention

[0006] Cyber ​​attacks can affect many aspects of the performance of weapons and equipment, but there is still little research in the academic community on the evaluation methods of the impact of cyber attacks on equipment performance. In view of this, the purpose of the present invention is to solve the current practical application problem of being unable to understand the various aspects of the impact of cyber attacks on weapons and equipment due to the lack of evaluation technology and methods and low evaluation accuracy, and to provide a method for evaluating the performance impact of cyber attacks on weapons and equipment based on a sparse model and ADMM distributed parallel data mining.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] The present invention provides a method for evaluating the impact of network attack performance of weapons and equipment based on a sparse model and ADMM distributed parallel data mining, comprising the following steps:

[0009] S1. According to the potential objects in the equipment that may be subject to network attacks, establish an equipment network attack performance impact assessment index system and determine m final index collection elements; the potential objects are equipment network resource nodes;

[0010] S2. Sparsely process the evaluation index system by using the coefficient of variation method to obtain a core index system that affects the key performance of the equipment;

[0011] S3. Based on the core indicator system, a weapon equipment network attack performance impact assessment model is constructed;

[0012] S4. In the impact assessment model, a time decay function is referenced to obtain an improved impact assessment model;

[0013] S5. Use the ADMM algorithm to obtain the minimum value of the improved impact assessment model, optimize the global problem and split it into multiple sub-problems; iteratively update the sub-problem variables until convergence, and obtain the minimum value of the performance impact assessment model function as the optimal solution for the performance impact of network attacks on weapons and equipment.

[0014] In one embodiment, step S2 includes:

[0015] S201. Use multiple attack methods to carry out threat intrusion behaviors on the equipment, and obtain the mean, square deviation and coefficient of variation of each final-level indicator after n network penetration tests;

[0016] S202, summing up the coefficients of variation of each item, and calculating the weight W of each final indicator according to the proportion of the coefficients of variation of each item in the summation. 3i ;

[0017] S203. Compare the weight of each final-level indicator with the average weight of the final-level indicators, take out the final-level indicator items whose weights are less than the average weight of the final-level indicators, obtain sparse final-level indicator items, and determine the core indicator system that affects the key performance of the equipment.

[0018] In one embodiment, step S3 includes:

[0019] S301, using the final index item after sparseness and the secondary index weight W predetermined according to the equipment application standard 2j , calculate the secondary index evaluation value Among them, j is the number of secondary index items, u is the jth item of the secondary index P 2j The corresponding initial number of final indicators; v is the jth secondary indicator P 2j The number of end items of the corresponding final-level indicator; the value range of u and v varies with the selection and sparsification of the final-level indicator items; It represents the final indicators after n network penetration tests; i is the number of final indicators;

[0020] S302, according to the secondary index evaluation value P 2j , and the predetermined number of first-level indicators k, weight W 1k , build a weapon equipment network attack performance impact assessment model K represents the number of end items of the first-level indicator.

[0021] In one embodiment, step S4 includes:

[0022] S401: Introduce the attack time parameter t into the impact assessment model. i , establish the performance impact time decay function A(t) = a·e -η·t ; a is the performance loss gradient in the ideal case where there is no performance impact time decay after the equipment network resource node is attacked; η is the decay factor; e is a constant;

[0023] S402. Determine the improved weapon equipment network attack performance impact assessment model as follows:

[0024]

[0025] t i It is the time difference between the current network attack time and the time when the attack purpose is achieved for the i-th final indicator.

[0026] In one embodiment, step S5 includes:

[0027] S501. Solve the minimum value of the improved weapon equipment network attack performance impact assessment model:

[0028]

[0029] S502, using the alternating direction multiplier method, decomposing the global optimization of the minimum value into multiple local sub-problems, and determining the optimization problem as:

[0030]

[0031] st x+bt=bx,t∈R d , b∈R q

[0032] R d is the set of all d-dimensional vectors, R q is the set of all q-dimensional vectors; b is the average time performance loss gradient when the equipment network resource node is attacked and the performance is affected by time decay; t is the time difference between the current network attack time and the time to achieve the attack purpose;

[0033] S503. Introduce the corresponding augmented Lagrangian function, divide the data into blocks, and initialize the global variables. Solve the original variables, dual variables, and Lagrangian variables according to the iterative process until the iteration is terminated and the convergence conditions are met. Solve the minimum value of the improved weapon equipment network attack performance impact assessment model function as the optimal solution for the weapon equipment network attack performance impact.

[0034] In one embodiment, in step S503, the augmented Lagrangian function is:

[0035]

[0036] p∈R k

[0037] R k is the set of all k-dimensional vectors, and p is the Lagrange multiplier.

[0038] Through the above technical scheme, it can be known that compared with the prior art, the present invention discloses a method for evaluating the impact of network attack performance of weapons and equipment based on sparse model and ADMM distributed parallel data mining, establishes an equipment network attack performance impact evaluation index system, and sparses the index system by the coefficient of variation method to obtain a core index system that affects the key performance of the equipment, and then establishes an evaluation model based on the sparse index system, which effectively solves the problem of index system sparseness and model lightweight. In order to make the evaluation value more accurate, the time decay function is introduced to eliminate the performance loss gain caused by the continuous attack after the network attack purpose is achieved, and the performance impact degree measurement value matching the penetration test is obtained, and the model is improved accordingly to make the evaluation result more reasonable and accurate. With this as the basic assumption background and conditions, an improved ADMM distributed parallel data mining method is then used to realize the distributed training of the algorithm and the improvement of the model accuracy, which solves the problem that the single-machine algorithm cannot be parallelized, so that the original single-machine algorithm is better adapted to the distributed computing mode, so that the evaluation method has better performance than other methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0040] Figure 1 It is a flow chart of the method for evaluating the impact of network attack performance of weapons and equipment provided by the present invention;

[0041] Figure 2It is a diagram of the equipment network attack performance impact evaluation index system provided by the present invention;

[0042] Figure 3 It is a diagram of the evaluation index system of equipment network attack performance impact after sparseness provided by the present invention;

[0043] Figure 4 It is a distributed parallel data mining algorithm diagram based on ADMM provided by the present invention;

[0044] Figure 5 This is a weighted representation of the sparse equipment network attack performance impact evaluation index system provided by the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0046] The embodiment of the present invention discloses a method for evaluating the impact of network attack performance of weapons and equipment based on a sparse model and ADMM distributed parallel data mining, referring to Figure 1 As shown, the following steps are included:

[0047] 1) According to the potential objects that can be attacked by network in the equipment, that is, the equipment network resource nodes, a three-level indicator system for evaluating the impact of equipment network attack performance is established. For example, taking a certain equipment of a digital mechanized infantry division as an example, 53 terminal indicator collection elements x1, x2, ..., x 53 ,like Figure 2 shown.

[0048] For example, the first-level index is: Firepower performance impact index P 11 , other performance influencing indicators P 12 , protection performance impact index P 13 , maneuverability performance impact index P 14 , Information performance impact index P 15 .

[0049] Firepower performance impact index P 11 , the secondary index is: P 21 (x1~x8),P 22 (x9~x 12 ).

[0050] 2) Use the coefficient of variation method to sparse the index system, for example, reduce the final index items to 34 items, and obtain the core index system that affects the key performance of the equipment, thereby reducing the amount of subsequent numerical calculations, such as Figure 3 shown.

[0051] In order to verify the impact of network attacks on equipment performance, such as injection attacks, access control unauthorized attacks, malicious sample attacks, channel attacks, password attacks, etc., typical threat means are used to implement threat intrusion behaviors on equipment. After n network penetration tests, the final indicators are i is the number of final index items (i=1,2,3…,53), and its mean

[0052] 3) By calculating the square difference of each final indicator after n network penetration tests

[0053] 4) The coefficient of variation of each final indicator is: (i=1,2,3…,53). i is the coefficient of variation of the ith final indicator, and then the coefficients of variation of the 53 final indicators are summed to obtain the comprehensive coefficient of variation m is the number of items collected by the final indicator (m = 53);

[0054] 5) Calculate the weight of each final indicator based on the comprehensive coefficient of variation. (m=53);

[0055] The weight of each final-level indicator is compared with the mean weight of the final-level indicator, and the final-level indicator items with weights less than the mean weight of the final-level indicator are removed to obtain the sparse final-level indicator items, and the sparse equipment network attack performance impact evaluation index system is determined.

[0056] 6) Use the final index item after sparseness and the secondary index weight W pre-determined according to the equipment application standard 2j , calculate the secondary index evaluation value j is the number of secondary index items, u is the jth item of secondary index P 2j The corresponding initial number of final indicators, refer to Figure 3 As shown, u∈(1,7,11,13,15,17,22,24,30); v is the j-th secondary index P 2j The corresponding number of end items of the final level indicator is v∈(6,10,12,14,16,21,23,29,34); the value range of u and v varies with the selection and sparsification of the final level indicator items.

[0057] 7) The sparse weapon equipment network attack performance impact assessment model is obtained as follows: k is the number of first-level indicators, and the number of ending items of k is K=(1,2,3,4,5), which indicates the number of ending items of first-level indicators; W 1k is the predetermined weight of the first-level indicator items.

[0058] 8) Considering that in network attacks, different attack methods take different attack times and achieve different attack effects. The longer the attack time, the higher the attack cost. Therefore, when achieving the same attack purpose, the shorter the attack time, the better, and the greater the performance loss, the better. The attack time parameter t should be introduced in the evaluation model. i , to comprehensively consider the impact of the attack. In actual network penetration testing, different attack methods are used to carry out overall continuous attacks on multiple network resource nodes. After a certain network resource node achieves the attack purpose, the attack on the node continues due to the overall continuity of the attack behavior until the pre-designed penetration test task is completed. Therefore, it is also necessary to establish a performance impact time decay function A(t), where t is the time difference between the current network attack time and the time when the attack purpose is achieved, so that the evaluation value is more accurate and reasonable. As the performance impact time decay function A(t) decays from the initial 1 to 0 over time, the closer the time is, the smaller the decay is, and the larger A(t) is, so a half-life factor ζ can be defined, that is, the time required for A(t) to decay from the beginning to half, then its function expression is,

[0059]

[0060] A(ζ) is the time taken for the half-life to decay from 1 to 0.

[0061] 9) Attenuation Factor Therefore, the performance impact time decay function A(t) is:

[0062] A(t)=a·e -η·t

[0063] 10) The final improved weapon equipment network attack performance impact assessment model is:

[0064]

[0065]

[0066] STx+bt=b

[0067] a is the performance loss gradient in the ideal case where there is no performance impact time decay after the equipment network resource node is attacked,

[0068] b is the average time performance loss gradient when the equipment network resource node is attacked and the performance is affected by time decay.

[0069] 11) For n penetration tests, the evaluation value can be obtained by inputting the above model based on the collected data. However, different values ​​of n and different penetration test methods, that is, x in The performance impact assessment value of the equipment will be different. Therefore, we can also solve the minimum value of the performance impact assessment model function to obtain the minimum impact of network attacks on equipment performance, which can be expressed as

[0070]

[0071] 12) Using the alternating direction multiplier method (ADMM), refer to Figure 4 As shown, the global problem is decomposed into multiple smaller and easier to solve local sub-problems, and the solution to the global problem is obtained by coordinating the solutions to the sub-problems. The problem to be optimized is:

[0072]

[0073] st x+bt=bx,t∈R d , b∈R q

[0074] R d is the set of all d-dimensional vectors, R d ={(x1, x2, ..x d )|x i ∈R i=1,2…d}

[0075] R q is the set of all q-dimensional vectors, R q ={(b1, b2, ..b q )|b i ∈R i=1,2…q}

[0076] Among them, based on the ADMM distributed parallel data mining algorithm, the mapping function and the reduction function are used to partition the problem function, allowing multiple nodes to execute concurrently and merge the final results. By controlling two variables to update the third variable, the partition calculation is cycled until the convergence condition of the function is met, the algorithm terminates, and the optimal solution is obtained.

[0077] 13) Its augmented Lagrangian function is:

[0078]

[0079] p∈R k

[0080] R k is the set of all k-dimensional vectors, R k ={(p1, p2, ..p k )|pi ∈R i=1,2…k}; p is the Lagrange multiplier (vector); capital T is the transpose symbol.

[0081] 14) Iterate using the update step until convergence:

[0082] (1) Divide the data into blocks and initialize global variables;

[0083] (2) According to the iterative process, solve the original variable: x k+1 , each block is calculated in parallel; here k represents the number of iterations of the ALM method, that is, the kth iteration;

[0084] (3) According to the iterative process, solve the dual variable: t k+1 , the summation part performs aggregation operation;

[0085] (4) According to the iterative process, solve the Lagrangian variable: p k+1 =p k +η(x k+1 +bt k+1 -b);

[0086] (5) Iterate 2-3 steps until convergence.

[0087] In this embodiment, the ADMM distributed parallel data mining method is adopted to determine the problem to be optimized, introduce its corresponding augmented Lagrangian function, divide the data into blocks, and initialize the global variables. According to the iterative process, the original variable, the dual variable, and the Lagrangian variable are solved until the iteration is terminated and the convergence condition is met. The minimum value of the performance impact assessment model function is solved, that is, the minimum impact of the network attack on the equipment performance is obtained, thereby decomposing a single large, complex, and computationally intensive problem into multiple smaller and easier to solve local sub-problems, and obtaining a global solution by coordinating the solutions of the sub-problems, thereby improving the computational efficiency and obtaining improved accuracy.

[0088] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0089] The present invention provides a method for evaluating the impact of network attack performance of weapons and equipment based on a sparse model and ADMM distributed parallel data mining, establishes an equipment network attack performance impact evaluation index system, and sparses the index system through the coefficient of variation method to obtain a core index system that affects the key performance of the equipment, and then establishes an evaluation model based on the sparse index system, effectively solving the problems of index system sparseness and model lightweight. In order to make the evaluation value more accurate, a time decay function is introduced to eliminate the performance loss gain caused by the continuous attack after the network attack purpose is achieved, and a performance impact degree measurement value matching the penetration test is obtained, and the model is improved accordingly to make the evaluation result more reasonable and accurate. With this as the basic assumption background and conditions, an improved ADMM distributed parallel data mining method is then used to realize the distributed training of the algorithm and the improvement of the model accuracy, solving the problem that the single-machine algorithm cannot be parallelized, so that the original single-machine algorithm is better adapted to the distributed computing mode, so that the evaluation method has better performance than other methods, and the specific implementation steps are as follows:

[0090] (1) Taking a certain equipment of a digital mechanized infantry division as an example, according to the above step 1), based on the potential objects in the equipment that can be attacked by network attacks, that is, the equipment network resource nodes, an equipment network attack performance impact evaluation index system is established, and 53 terminal index collection elements x1, x2, ..., x3 are determined. 53 ,See Figure 2 shown.

[0091] (2) According to the above steps 2) to 5), typical threat means such as injection attack, access control unauthorized attack, malicious sample attack, channel attack, password attack, etc. are used to implement threat intrusion behavior on the equipment. After two network penetration tests, the index system is sparsely divided by the coefficient of variation method, so that the final index items become 34 items. A model is established to obtain the core index system that affects the key performance of the equipment. Figure 3 shown.

[0092] (3) According to the above steps 6) to 7), the secondary index evaluation value is calculated using the 34 final index items after sparseness and the secondary index weights predetermined according to the equipment operation standards. j is the number of secondary index items, j = (1, 2, 3, 4, 5, 6, 7, 8, 9), u is the jth item of the secondary index P 2j The corresponding initial number of the final index, u∈(1,7,11,13,15,17,22,24,30); v is the jth secondary index P 2j The corresponding number of terminal items of the final level indicator is v∈(6, 10, 12, 14, 16, 21, 23, 29, 34); thus the sparse weapon equipment network attack performance impact assessment model is obtained as follows: Where k is the number of first-level indicators, K = (1, 2, 3, 4, 5) The weight combination relationship of indicators at all levels is shown in Figure 5 shown.

[0093] (4) According to the above steps 8) to 10), introduce the time decay function A(t) = x·e -η·t , the final improved weapon equipment network attack performance impact assessment model is obtained as follows

[0094]

[0095] (5) According to the above step 11), the minimum value of the performance impact assessment model function is solved to obtain the minimum impact of the network attack on the equipment performance, which can be expressed as:

[0096]

[0097] (6) According to the above steps 12) to 14), the alternating direction multiplier method (ADMM) is used, see Figure 4 , transforming the problem into

[0098]

[0099] st x+bt=bx,t∈R d , b∈R q

[0100] (7) By deriving its augmented Lagrangian function, the update step is used to iterate until convergence, and the result is used to solve the minimum value of the performance impact assessment model function, and finally obtain the minimum impact of network attacks on equipment performance. The effectiveness of this method is verified through equipment network attack performance impact assessment and comparative analysis. The results show that the equipment network attack performance impact assessment method based on the core indicator system and the distributed parallel mining algorithm of ADMM has better performance than other methods.

[0101] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0102] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating the performance impact of network attacks on weapons and equipment based on a sparse model and ADMM distributed parallel data mining, characterized in that: The following steps are involved: S1. According to the potential targets in the equipment that may be attacked by cyber attacks, establish an equipment cyber attack performance impact assessment index system and determine the m final index collection elements; The potential object is an equipment network resource node; wherein, according to the potential object in the equipment that may be subjected to network attacks, three levels of indicators are divided into: a primary indicator, a plurality of secondary indicators corresponding to each primary indicator, and a plurality of final indicators corresponding to each secondary indicator; the plurality of final indicators constitute m items of final indicator collection elements; S2. Sparsely process the evaluation index system by using the coefficient of variation method to obtain a core index system that affects the key performance of the equipment; S3. Based on the core indicator system, a weapon equipment network attack performance impact assessment model is constructed; S4. In the impact assessment model, a time decay function is referenced to obtain an improved impact assessment model; S5. Using the ADMM algorithm to optimize the minimum value of the improved impact assessment model, perform global problem optimization and split it into multiple sub-problems; iteratively update the sub-problem variables until convergence, and obtain the minimum value of the performance impact assessment model function as the optimal solution for the performance impact of the weapon equipment network attack; The step S2 comprises: S201. Use multiple attack methods to carry out threat intrusion behaviors on the equipment, and obtain the mean, square deviation and coefficient of variation of each final-level indicator after n network penetration tests; S202, summing up the coefficients of variation of each item, and calculating the weight W of each final indicator according to the proportion of the coefficients of variation of each item in the summation. 3i ; S203, comparing the weight of each final-level indicator with the average weight of the final-level indicator, taking out the final-level indicator items whose weights are less than the average weight of the final-level indicator, obtaining sparse final-level indicator items, and determining the core indicator system that affects the key performance of the equipment; The step S3 comprises: S301, using the final index item after sparseness and the secondary index weight W predetermined according to the equipment application standard 2j , calculate the secondary index evaluation value Among them, j is the number of secondary index items, u is the jth item of the secondary index P 2j The corresponding initial number of final indicators; v is the jth secondary indicator P 2j The number of end items of the corresponding final-level indicator; the value range of u and v varies with the selection and sparsification of the final-level indicator items; It represents the final indicators after n network penetration tests; i is the number of final indicators; S302, according to the secondary index evaluation value P 2j , and the predetermined number of first-level indicators k and weight W1k, to build a weapon equipment network attack performance impact assessment model K represents the number of end items of the first-level indicator.

2. The method according to claim 1, characterized in that The step S4 comprises: S401: Introduce the attack time parameter t into the impact assessment model. i , establish the performance impact time decay function A(t) = a·e -η·t ; a is the performance loss gradient in the ideal case where there is no performance impact time decay after the equipment network resource node is attacked; η is the decay factor; e is a constant; S402. Determine the improved weapon equipment network attack performance impact assessment model as follows: t i It is the time difference between the current network attack time and the time when the attack purpose is achieved for the i-th final indicator.

3. The method according to claim 2, characterized in that The step S5 comprises: S501. Solve the minimum value of the improved weapon equipment network attack performance impact assessment model: S502, using the alternating direction multiplier method, decomposing the global optimization of the minimum value into multiple local sub-problems, and determining the optimization problem as: s.tx+bt=bx,t∈R d ,b∈R q R d is the set of all d-dimensional vectors; R q is the set of all q-dimensional vectors; b is the average time performance loss gradient when the equipment network resource node is attacked and the performance is affected by time decay; t is the time difference between the current network attack time and the time to achieve the attack purpose; S503. Introduce the corresponding augmented Lagrangian function, divide the data into blocks, and initialize the global variables. Solve the original variables, dual variables, and Lagrangian variables according to the iterative process until the iteration is terminated and the convergence conditions are met. Solve the minimum value of the improved weapon equipment network attack performance impact assessment model function as the optimal solution for the weapon equipment network attack performance impact.

4. The method according to claim 3, characterized in that In step S503, the Lagrangian function is augmented: p∈R k R k is the set of all k-dimensional vectors; p is the Lagrange multiplier.