Industrial grade control cabinet safety control method

Abstract

The invention relates to the technical field of intelligent control, and discloses an industrial-grade control cabinet safety control method, which comprises the following steps: step 1, acquiring quantum arbitrary number generator data to generate a key and evaluating key entropy by editing distance measurement; 2, according to the key generated in the step 1, a Markov decision is adopted to adjust a key updating strategy, and the key is updated when an attack is detected; 3, collecting network traffic according to the updating strategy in the step 2 to construct a topology analysis model so as to persistently and coherently judge the network attack and send out an alarm; and 4, constructing an adjacent matrix according to the flow data and the equipment connection information in the step 3, and calculating a graph Laplacian matrix. According to the method, a quantum arbitrary number generation and Markov decision optimization key updating strategy is adopted, the key is dynamically updated when the attack is detected, the key arbitrariness and the anti-attack capability are ensured, and compared with an existing static key management scheme, the key leakage risk can be effectively reduced, and the quantum computing attack resisting capability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and specifically to a safety control method for industrial control cabinets. Background Art

[0002] Industrial control cabinets are an important part of modern industrial automation and intelligent manufacturing, and are widely used in the power, communication, transportation, and intelligent manufacturing industries. Their core functions include equipment control, data processing, and network communication. However, with the continuous upgrading of network attack means, the increasing complexity of system topologies, and the improvement of human-machine interaction requirements, existing control cabinets have many defects in terms of security, maintainability, and interaction experience, restricting the pace of intelligent development.

[0003] Traditional control cabinets mainly rely on static key mechanisms for communication encryption and identity authentication. Once the static key is leaked or cracked, the security of the system will be seriously threatened. In addition, in the face of the development of quantum computing technology, the anti-attack ability of traditional encryption algorithms has decreased significantly, and the system is vulnerable to quantum computing attacks. Existing solutions usually rely on fixed-period key updates, but lack an intelligent dynamic adjustment mechanism and cannot respond in a timely manner when an attack is detected, resulting in potential security vulnerabilities.

[0004] The topological structure of industrial control networks is not fixed, and the addition or removal of devices, changes in connection status, and dynamic adjustment of network loads will all affect the overall architecture of the system. Currently, most control cabinet systems use manual topology configuration, which is cumbersome and error-prone. In addition, most existing topology learning methods are based on static graph models and cannot adapt to the real-time changing network environment, resulting in limited system scalability and high maintenance costs.

[0005] The display interface of industrial control cabinets is usually used for status monitoring and operation instruction input. Different environmental lighting conditions will affect the display effect, resulting in unidentifiable information and even causing misoperations. Existing interface layouts are usually preset and lack the ability to adaptively adjust to lighting conditions, and cannot maintain the best readability in different environments. Some systems can use automatic brightness adjustment technology, but they do not fully combine system topology changes, resulting in poor interface adaptability and limited user experience.

[0006] Therefore, those skilled in the art provide a safety control method for industrial control cabinets to solve the above-mentioned problems. Summary of the Invention

[0007] In view of the deficiencies of the prior art, the present invention provides a safety control method for industrial control cabinets to solve the problems raised in the above background art.

[0008] To achieve the above objectives, the present invention is realized through the following technical solutions: A safety control method for industrial control cabinets, comprising:

[0009] Step 1, collect the data of the quantum random number generator to generate a key and evaluate the key entropy with the edit distance metric;

[0010] Step 2, adjust the key update strategy by Markov decision based on the key generated in Step 1 and update the key when an attack is detected;

[0011] Step 3, collect network traffic according to the update strategy in Step 2 to construct a topology analysis model, determine network attacks by persistent homology, and issue an alarm;

[0012] Step 4, construct an adjacency matrix based on the traffic data and device connection information in Step 3 and calculate the graph Laplacian matrix;

[0013] Step 5, train a topology model using a graph convolutional network based on the graph Laplacian matrix in Step 4 to determine module connections and update the system topology;

[0014] Step 6, collect new data to construct an adjacency matrix when the device connection changes according to the system topology in Step 5 and adjust the system structure according to the updated topology;

[0015] Step 7, collect the illumination data of the display interface according to the system structure in Step 6 to construct an illumination model and calculate the interface layout by differentiable rendering to complete the display update.

[0016] Preferably, in Step S1, when the data output by the quantum random number generator satisfies the stable sampling condition, it further includes:

[0017] Step 1.1, define the data set output by the quantum random number generator:

[0018] R = {r 1 , r 2 ,..., r n},

[0019] where r 1 represents the random number obtained from the first sampling, r 2 represents the random number obtained from the second sampling, r n represents the random number obtained from the nth sampling, and n represents the number of samplings;

[0020] Calculate the key candidates:

[0021] where a i represents the weight factor corresponding to the i-th sampling, r i represents the random number obtained from the i-th sampling, and K 1 represents the key candidate;

[0022] Step 1.2, define the reference sequence: T = {t 1 , t2 ,...,t m}, where t 1 represents the first element of the reference sequence, and t 2 represents the second reference value, and t m represents the last element of the reference sequence, m represents the length of the reference sequence, and T represents the reference sequence;

[0023] Calculate the edit distance:

[0024] where D represents the key candidate K 1 and the sum of the mismatch indicators at each position with the reference sequence T, and β j represents the mismatch indicator at the j-th position, and L represents the comparison length;

[0025] Step 1.3, define the normalized entropy: H = D / L,

[0026] where H represents the entropy value of the key candidate, and D represents the key candidate K 1 and the sum of the mismatch indicators at each position with the reference sequence T, and L represents the comparison length;

[0027] Set a high-entropy threshold y. If H ≥ y, determine that the key candidate meets the high-entropy requirement.

[0028] Preferably, in step S2, after the key candidate determined in step S1 meets the high-entropy requirement, it further includes:

[0029] Step 2.1, based on the key candidate K calculated in step S1 1 and the reference sequence T, define the key update probability distribution: p K ={p 1 , p 2 ,..., p n},

[0030] where p 1 represents the probability that the key candidate K 1 takes the first state, and p 2 represents the probability that the key candidate K 1 takes the second state, and p n represents the probability that the key candidate K 1 takes the n-th state, and construct a Markov decision process: M = (S, A, P, R, γ),

[0031] where S is the set of key states, A is the set of update strategies, P is the state transition probability matrix, R is the state reward function, and γ is the discount factor;

[0032] Step 2.2, based on the Markov decision process constructed in step 2.1, calculate the state transition probability:

[0033]

[0034] Among them, P(s′∣s,a) represents the probability that the current state s is transferred to the next state s′ by the policy a, s represents the current state, s′ represents the next state, and a represents the adopted key update policy, δ i represents the adjustment factor of the i-th state, p n represents the key candidate K 1 takes the value of the probability of the n-th state, represents the summation calculation of the key update probability distribution p K ;

[0035] Optimize the policy A to maximize the cumulative reward function:

[0036] J(A) = Σ (t＝0) ∧∞γ t R(s t ,a t ),

[0037] Among them, J(A) represents the cumulative reward function, γ represents the discount factor, and R(s t ,a t ) represents the reward value obtained by taking the policy a t and being in the state s t at the time step t, Σ (t＝0) ∧∞ represents the cumulative calculation for all time steps t;

[0038] Step 2.3, after the update policy A optimized in Step 2.2 0 is determined, according to the key update trigger condition Θ, determine whether to perform key update, and define the trigger condition:

[0039] Θ = f(D,H,η),

[0040] Among them, D represents the sum of the mismatch indicators of the key candidate K 1 and the reference sequence T at each position, H represents the entropy value of the key candidate, and η represents the attack detection feedback signal;

[0041] If Θ exceeds the set threshold, perform key update and replace the old key K 0 with the new key K 1 .

[0042] Preferably, in step S3, after the key update policy A optimized in step S2 0 is determined, it further includes:

[0043] Step 3.1, according to the new key K determined in step S2 1 and the update policy A0 , collect network traffic data: X = {x 1 , x 2 ,..., x m}, where x m represents the data packet collected in the m-th time slice, m represents the total number of time slices collected, and define the network traffic feature vector:

[0044] F = g(X, K 1 ),

[0045] where F is the network traffic feature vector, g(·) is a feature extraction function weighted by a key to generate traffic topology data for subsequent attack detection;

[0046] Step 3.2, construct a topology data analysis model based on the traffic feature vector F calculated in Step 3.1:

[0047] G = (V, E, W), where V is the set of network nodes, E is the set of connection relationships between nodes, W is the edge weight matrix, and G is the network traffic topology data analysis model;

[0048] Calculate the topological persistent homology feature: P = h(G), where P is the topological persistent homology feature and h(·) is a topological feature extraction function to identify potential abnormal patterns;

[0049] Step 3.3, based on the topological feature P extracted in Step 3.2, define an attack detection decision function:

[0050] θ = ψ(P, K 1 , A 0 ),

[0051] where θ is the attack detection decision function, ψ(·) is a detection function that comprehensively considers topological features, keys, and update strategies, and set the attack decision threshold θ 0 , if θ ≥ θ 0 , trigger an alarm signal and feedback the attack information η to Step S2 to adjust the key update strategy.

[0052] Preferably, in Step S4, after determining the network attack feature and triggering the alarm signal in Step S3, it further includes:

[0053] Step 4.1, based on the topology data analysis model constructed in Step S3:

[0054] G = (V, E, W), where V is the set of network nodes, E is the set of connection relationships between nodes, G is the network traffic topology data analysis model, W is the edge weight matrix, and W = [w ij , where w ij represents the node v i and vj The connection strength between;

[0055] Define the adjacency matrix A: A = [a ij , where, a ij represents the connection relationship between nodes v i and v j . If there is a direct connection: a ij = w ij , otherwise a ij = 0, where v i and v j represent the i-th and j-th network nodes in the topological model;

[0056] Calculate the graph Laplacian matrix: L = D - A, where D represents the degree matrix, L represents the graph Laplacian matrix, which is used to characterize the global information of the network topology and serves as the input for subsequent graph convolutions;

[0057] Step 4.2, based on the graph Laplacian matrix L calculated in Step 4.1, use a graph convolutional network to train the system topological model: M t = f(L, X),

[0058] where X is the node feature matrix, f(·) is the graph convolution operation, and calculate the node embedding representation:

[0059] Z = σ(LXW), where W is the training parameter matrix and σ(·) is the activation function to learn the module connection method in the topological structure;

[0060] Step 4.3, based on the system topological model M t trained in Step 4.2, define the topology update strategy:

[0061] Ω = argmax g(Z, P), where g(·) is the objective function optimized based on node embedding and topological features. Adjust the system topology according to Ω and feedback the topology change information to Step S3 to optimize the subsequent network attack detection and key update strategies.

[0062] Preferably, in Step S5, after the graph Laplacian matrix L calculated in Step S4 and training the system topological model M t , it further includes:

[0063] Step 5.1, based on the topological model M t trained in Step S4, collect device connection change data:

[0064] Y = {y 1 , y 2 ,..., y k}, where, y kDenote the device connection information recorded in the k-th time slice, and define a new adjacency matrix: A′ = [a′ ij , where a′ ij represents the new connection relationship between nodes v i and v j . If there is a direct connection: a′ ij = w′ ij , otherwise: a′ ij = 0. Calculate the updated graph Laplacian matrix: L′ = D′ - A′, where D′ is the updated degree matrix;

[0065] Step 5.2, based on the updated graph Laplacian matrix L′ calculated in Step 5.1, use a graph convolutional network to calculate the new topological model: M t ′ = f(L′, X′), where X′ is the updated node feature matrix, f(·) is the graph convolution operation, and calculate the new node embedding representation:

[0066] Z′ = σ(L′X′W′), where W′ is the new training parameter matrix, σ(·) is the activation function, to optimize the system topological structure;

[0067] Step 5.3, based on the new topological model M t ′ calculated in Step 5.2, define the topological optimization objective:

[0068] Ω′ = argmax g(Z′, ξ),

[0069] where g(·) is the optimization function calculated based on the node embedding and topological change ξ. Re-adjust the system topological structure according to Ω′, and feedback the new topological adjustment information to Step S4 to further optimize the subsequent network analysis and attack detection.

[0070] Preferably, in Step S6, after the new topological structure calculated in Step S5 and adjusting the system topology, it further includes:

[0071] Step 6.1, based on the updated graph Laplacian matrix L′ and the optimized topological structure calculated in Step S5, collect the illumination data of the display interface:

[0072] I = {i 1 , i 2 ,..., i m}, where i m represents the illumination data at the m-th time slice, and define the illumination model:

[0073] M I = φ(I, L′), where φ(·) represents the mapping function based on the illumination data and the topological structure, and is used to calculate the influence of illumination on the interface display;

[0074] Step 6.2, based on the lighting model M calculated in Step 6.1 I , construct a lighting influence matrix:

[0075] G = [g ij , where g ij represents the lighting influence degree between interface elements e i and e j . Define a differentiable rendering calculation formula: C = σ(GX I W I ),

[0076] where X I is the lighting feature matrix of the interface element, W I is the training parameter matrix, σ(·) is the activation function, and calculate the interface layout adjustment plan;

[0077] Step 6.3, based on the interface layout adjustment plan C calculated in Step 6.2, define the final interface update strategy: Ω I = argmax f(C, Z′),

[0078] where f(·) is the optimization objective function, perform interface adjustment based on the calculated interface lighting information and topological embedding Z′, and feedback the interface adjustment information to Step S5 to optimize the system topological structure and interaction experience.

[0079] Preferably, in Step S7, after the interface lighting model M calculated in Step S6 I and the interface layout is adjusted, it further includes:

[0080] Step 7.1, based on the lighting influence matrix G calculated in Step S6 and the optimized interface layout plan C, collect the final interface display data:

[0081] D I = {d 1 , d 2 ,..., d n}, where d n represents the interface display state at the k-th time slice, and define the interface display feature matrix: X D = ψ(D I , C), where ψ(·) is a mapping function based on the display data and the layout plan to calculate the interface stability and user interaction experience;

[0082] Step 7.2, based on the interface display feature matrix X D calculated in Step 7.1, construct an interface stability evaluation function:

[0083]

[0084] Among them, α ij is the interaction weight between the interface elements e i and e j to evaluate the consistency and stability of the interface display; is used to calculate the relative change degree of the interface elements to evaluate the consistency and stability of the interface display;

[0085] Step 7.3, based on the interface stability S calculated in Step 7.2, define the final interface optimization objective: Ω D = argmax g(S, C, G),

[0086] where g(·) is an optimization objective function that comprehensively considers interface stability, layout scheme, and lighting effects. Based on Ω D calculate the final interface adjustment scheme, and feedback interface optimization information to Step S6 to further optimize the lighting model and interface layout.

[0087] A terminal device includes a processor, a memory, and a communication interface that execute the industrial control cabinet security control method.

[0088] A storage medium stores a computer program, and when the program is executed, it enables the terminal device to implement the industrial control cabinet security control method.

[0089] The present invention provides an industrial control cabinet security control method. It has the following beneficial effects:

[0090] 1. The present invention adopts a quantum arbitrary number generation and Markov decision optimization key update strategy to dynamically update the key when detecting an attack, ensuring the randomness of the key and the anti-attack ability. Compared with the existing static key management scheme, it can effectively reduce the risk of key leakage and improve the system's anti-quantum computing attack ability.

[0091] 2. The present invention adopts a topology learning method based on graph convolutional networks, which can automatically learn and optimize the device connection structure, enabling the system to adapt to the dynamically changing device topology. Traditional methods rely on manual adjustment of the topology, which is inefficient and error-prone, while this solution can achieve intelligent adjustment, improve operation and maintenance efficiency, and support flexible system expansion.

[0092] 3. The present invention combines light perception and differentiable rendering technology to achieve adaptive interface adjustment, ensuring that the interface is clearly visible under different environmental lighting conditions. Compared with the traditional fixed interface layout method, the present invention can dynamically optimize the display effect according to the lighting change, improve the user's visual comfort, and reduce the risk of misoperation. Brief Description of the Drawings

[0093] Figure 1 is a flowchart of the present invention. Detailed Embodiment

[0094] To enable those skilled in the art to understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0095] The present invention will be described in detail below with reference to the accompanying drawings:

[0096] Embodiment:

[0097] Please refer to the attached Figure 1 , the embodiment of the present invention provides an industrial control cabinet safety control method, including:

[0098] Step 1, collecting quantum random number generator data to generate a key and evaluating the key entropy with the edit distance metric;

[0099] Step 1.1, defining the output data set of the quantum random number generator:

[0100] R = {r 1 , r 2 ,..., r n},

[0101] where r 1 represents the random number obtained from the first sampling, r 2 represents the random number obtained from the second sampling, r n represents the random number obtained from the nth sampling, and n represents the number of samplings;

[0102] Calculating key candidates:

[0103] where a i represents the weight factor corresponding to the ith sampling, r i represents the random number obtained from the ith sampling, and K 1 represents the key candidate;

[0104] Step 1.2, defining a reference sequence: T = {t 1 , t 2 ,..., t m}, where t 1 represents the first element of the reference sequence, t 2 represents the second reference value, t m represents the last element of the reference sequence, m represents the length of the reference sequence, and T represents the reference sequence;

[0105] Calculating the edit distance:

[0106] Among them, D represents the key candidate K 1 The sum of the mismatch indicators at each position with the reference sequence T, β j represents the mismatch indicator at the j-th position, and L represents the comparison length;

[0107] Step 1.3, define the normalized entropy: H = D / L,

[0108] where H represents the entropy value of the key candidate, and D represents the key candidate K 1 The sum of the mismatch indicators at each position with the reference sequence T, and L represents the comparison length;

[0109] Set a high-entropy threshold y. If H ≥ y, it is determined that the key candidate meets the high-entropy requirement;

[0110] Step 2, adjust the key update strategy using Markov decision based on the key generated in Step 1 and update the key when an attack is detected;

[0111] Step 2.1, based on the key candidate K calculated in Step S1 1 and the reference sequence T, define the key update probability distribution: p K ={p 1 , p 2 ,..., p n},

[0112] where p 1 represents the probability that the key candidate K 1 takes the value of the first state, p 2 represents the probability that the key candidate K 1 takes the value of the second state, p n represents the probability that the key candidate K 1 takes the value of the n-th state, and construct a Markov decision process: M = (S, A, P, R, γ),

[0113] where S is the set of key states, A is the set of update strategies, P is the state transition probability matrix, R is the state reward function, and γ is the discount factor;

[0114] Step 2.2, based on the Markov decision process constructed in Step 2.1, calculate the state transition probability:

[0115]

[0116] where P(s′∣s,a) represents the probability that the current state s transfers to the next state s′ by the strategy a, s represents the current state, s′ represents the next state, a represents the adopted key update strategy, and δ i represents the adjustment factor of the i-th state, and p n represents the key candidate K1 The probability of taking the nth state represents the summation calculation of the key update probability distribution p K ;

[0117] Optimize policy A to maximize the cumulative reward function:

[0118] J(A) = Σ (t＝0) ∧∞γ t R(s t , a t ),

[0119] where J(A) represents the cumulative reward function, γ represents the discount factor, and R(s t , a t ) represents the reward value obtained when taking policy a t and being in state s t at time step t, and Σ (t＝0) ∧∞ represents the cumulative calculation for all time steps t;

[0120] Step 2.3, after the updated policy A 0 optimized in Step 2.2 is determined, according to the key update trigger condition Θ, determine whether to perform key update, and define the trigger condition:

[0121] Θ = f(D, H, η),

[0122] where D represents the sum of the mismatch indicators of the key candidate K 1 and the reference sequence T at each position, H represents the entropy value of the key candidate, and η represents the attack detection feedback signal;

[0123] If Θ exceeds the set threshold, perform key update and replace the old key K 0 with the new key K 1 ;

[0124] Step 3, based on the updated policy in Step 2, collect network traffic to construct a topological analysis model to determine network attacks by persistent homology and issue an alarm;

[0125] Step 3.1, based on the new key K 1 and the updated policy A 0 determined in Step S2, collect network traffic data: X = {x 1 , x 2 ,..., x m}, where x m represents the data packet collected in the mth time slice, and m represents the total number of time slices collected, and define the network traffic feature vector:

[0126] F = g(X, K 1 ),

[0127] Among them, F is the network traffic feature vector, and g(·) is the feature extraction function weighted by the key to generate traffic topology data for subsequent attack detection;

[0128] Step 3.2, construct a topology data analysis model based on the traffic feature vector F calculated in step 3.1:

[0129] G = (V, E, W), where V is the set of network nodes, E is the set of connection relationships between nodes, W is the edge weight matrix, and G is the network traffic topology data analysis model;

[0130] Calculate the topological persistent homology feature: P = h(G), where P is the topological persistent homology feature and h(·) is the topological feature extraction function to identify potential abnormal patterns;

[0131] Step 3.3, based on the topological feature P extracted in step 3.2, define an attack detection decision function:

[0132] θ = ψ(P, K 1 , A 0 ),

[0133] where θ is the attack detection decision function, ψ(·) is the detection function that comprehensively considers topological features, keys, and update strategies, and set the attack decision threshold θ 0 , if θ ≥ θ 0 , trigger an alarm signal and feedback the attack information η to step S2 to adjust the key update strategy;

[0134] Step 4, construct an adjacency matrix based on the traffic data and device connection information in step 3 and calculate the graph Laplacian matrix;

[0135] Step 4.1, based on the topology data analysis model constructed in step S3:

[0136] G = (V, E, W), where V is the set of network nodes, E is the set of connection relationships between nodes, G is the network traffic topology data analysis model, W is the edge weight matrix, and W = [w ij , where w ij represents the connection strength between nodes v i and v j ;

[0137] Define the adjacency matrix A: A = [a ij , where a ij represents the connection relationship between nodes v i and v j . If there is a direct connection: a ij = w ij , otherwise a ij = 0, v i, v j represent the i-th and j-th network nodes in the topological model;

[0138] Calculate the graph Laplacian matrix: L = D - A, where D represents the degree matrix, L represents the graph Laplacian matrix, which is used to characterize the global information of the network topology and serves as the input for subsequent graph convolutions;

[0139] Step 4.2, based on the graph Laplacian matrix L calculated in Step 4.1, use a graph convolutional network to train the system topological model: M t = f(L, X),

[0140] where X is the node feature matrix, f(·) is the graph convolution operation, and calculate the node embedding representation:

[0141] Z = σ(LXW), where W is the training parameter matrix, σ(·) is the activation function, to learn the module connection method in the topological structure;

[0142] Step 4.3, based on the system topological model M trained in Step 4.2 t , define the topology update strategy:

[0143] Ω = argmax g(Z, P), where g(·) is the objective function optimized based on node embedding and topological features. Adjust the system topology according to Ω and feedback the topology change information to Step S3 to optimize the subsequent network attack detection and key update strategies;

[0144] Step 5, use a graph convolutional network to train the topological model based on the graph Laplacian matrix in Step 4 to determine the module connection and update the system topology;

[0145] Step 5.1, based on the topological model M trained in Step S4 t , collect the device connection change data:

[0146] Y = {y 1 , y 2 ,..., y k}, where y k represents the device connection information recorded at the k-th time slice. Define a new adjacency matrix: A' = [a' ij , where a' ij represents the new connection relationship between nodes v i and v j . If there is a direct connection: a' ij = w' ij , otherwise: a' ij = 0. Calculate the updated graph Laplacian matrix: L' = D' - A', where D' is the updated degree matrix;

[0147] Step 5.2, according to the updated Laplacian matrix L′ calculated in Step 5.1, use a graph convolutional network to calculate the new topological model: M t ′ = f(L′, X′), where X′ is the updated node feature matrix, f(·) is the graph convolution operation, and calculate the new node embedding representation:

[0148] Z′ = σ(L′X′W′), where W′ is the new training parameter matrix, σ(·) is the activation function, to optimize the system topology structure;

[0149] Step 5.3, according to the new topological model M t ′ calculated in Step 5.2, define the topological optimization objective:

[0150] Ω′ = argmax g(Z′, ξ),

[0151] where g(·) is the optimization function calculated based on the node embedding and topological change ξ, re-adjust the system topology structure according to Ω′, and feedback the new topological adjustment information to Step S4 to further optimize the subsequent network analysis and attack detection;

[0152] Step 6, according to the system topology in Step 5, collect new data to construct an adjacency matrix when the device connection changes and adjust the system structure according to the updated topology;

[0153] Step 6.1, according to the updated Laplacian matrix L′ and the optimized topological structure calculated in Step S5, collect the illumination data of the display interface:

[0154] I = {i 1 , i 2 ,..., i m}, where i m represents the illumination data at the m-th time slice, and define the illumination model:

[0155] M I = φ(I, L′), where φ(·) represents the mapping function based on the illumination data and topological structure, and is used to calculate the influence of illumination on the interface display;

[0156] Step 6.2, according to the illumination model M I calculated in Step 6.1, construct the illumination influence matrix:

[0157] G = [g ij , where g ij represents the illumination influence degree between the interface elements e i and e j , and define the differentiable rendering calculation formula: C = σ(GX I W I ),

[0158] Among them, X I is the illumination feature matrix of the interface element, W I is the training parameter matrix, σ(·) is the activation function, and the interface layout adjustment scheme is calculated;

[0159] Step 6.3, according to the interface layout adjustment scheme C calculated in step 6.2, define the final interface update strategy: Ω I = argmax f(C, Z′),

[0160] where f(·) is the optimization objective function, the interface is adjusted based on the calculated interface illumination information and topological embedding Z′, and the interface adjustment information is fed back to step S5 to optimize the system topology structure and interaction experience;

[0161] Step 7, based on the system structure in step 6, collect the interface illumination data to construct an illumination model and complete the display update by differentiable rendering to calculate the interface layout;

[0162] Step 7.1, according to the illumination influence matrix G calculated in step S6 and the optimized interface layout scheme C, collect the final interface display data:

[0163] D I = {d 1 , d 2 ,..., d n}, where d n represents the interface display state at the k-th time slice, and define the interface display feature matrix: X D = ψ(D I , C), where ψ(·) is a mapping function based on the display data and the layout scheme to calculate the interface stability and user interaction experience;

[0164] Step 7.2, according to the interface display feature matrix X D calculated in step 7.1, construct an interface stability evaluation function:

[0165]

[0166] where α ij is the interaction weight between interface elements e i and e j , is to calculate the relative change degree of the interface elements to evaluate the consistency and stability of the interface display;

[0167] Step 7.3, according to the interface stability S calculated in step 7.2, define the final interface optimization objective: Ω D = argmax g(S, C, G),

[0168] Among them, g(·) is the optimization objective function that comprehensively considers interface stability, layout scheme and illumination influence. D The final interface adjustment solution is calculated, and the interface optimization information is fed back to step S6 to further optimize the lighting model and interface layout.

[0169] The benefit of step S1 is that the quantum arbitrary number generator is used to ensure the arbitrariness of the key and improve the encryption security. In addition, the edit distance entropy evaluation method is combined to ensure that the generated key meets the high entropy requirements and effectively resists quantum computing attacks.

[0170] The benefits of step S2 are that the key update strategy is dynamically adjusted through the Markov decision process, the intelligence level of key management is improved, and the key can be quickly adjusted when an attack is detected to avoid security risks caused by long-term unchanged keys;

[0171] The benefits of step S3 are to collect real-time traffic data, extract topological features in combination with persistent coherence methods, accurately detect abnormal traffic patterns, and feed back to the key management system when an alarm signal is triggered, so as to achieve linkage optimization of keys and security detection;

[0172] The benefits of step S4 are to build a network topology model based on traffic data and device connection relationships, improve the understanding of the system structure, and calculate the graph Laplacian matrix to obtain global topology features, providing a basis for subsequent topology optimization;

[0173] The benefits of step S5 are that the graph convolutional network is used to train the topology model, which can realize automatic topology learning and optimization, improve network adaptability, and when the device connection changes, the system can automatically adjust the topology structure, reduce manual intervention, and improve operation and maintenance efficiency;

[0174] The benefits of step S6 are that the differentiable rendering technology is used in combination with the ambient lighting data to achieve dynamic adjustment of the interface and improve readability. Compared with the traditional static interface solution, this method can automatically adapt to lighting changes and reduce the risk of misoperation;

[0175] Benefits of step S7: Through the interface stability evaluation function, it is ensured that the interface layout can maintain a good user experience after adjustment, and the interface display scheme is optimized in combination with the illumination model to improve the comfort and operation accuracy of human-computer interaction;

[0176] In summary, this invention achieves enhanced security of industrial control cabinets, improved system maintainability and optimized interactive experience through quantum key management, network topology optimization and intelligent interface adjustment technology. Unlike traditional solutions, this method can dynamically adapt to environmental changes and optimize the system structure by combining intelligent learning technology, providing a reliable and efficient security control solution for intelligent manufacturing, industrial Internet of Things and critical infrastructure.

[0177] A terminal device, comprising a processor, a memory, and a communication interface that execute an industrial control cabinet security control method.

[0178] A storage medium stores a computer program, which, when executed, enables the terminal device to implement an industrial control cabinet security control method.

[0179] The terminal device and the storage medium provided by the present invention enhance system security, improve maintainability, and optimize the interaction experience by executing an industrial control cabinet security control method. Compared with traditional solutions, this solution can adaptively adjust keys, intelligently optimize the topology structure, and dynamically adapt the interface, and is widely applicable to intelligent manufacturing, industrial Internet of Things, and critical infrastructure, providing a safe, efficient, and intelligent solution for industrial control systems.

[0180] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A safety control method for an industrial control cabinet, characterized in that: include: Step 1, collect quantum arbitrary number generator data to generate a key and evaluate the key entropy using the edit distance metric; Step 2: Using Markov decision making to adjust the key update strategy based on the key generated in step 1 and update the key when an attack is detected; Step 3: According to the updated strategy in step 2, collect network traffic to build a topology analysis model to persistently synchronize and determine network attacks and issue an alarm; Step 4: construct an adjacency matrix based on the traffic data and device connection information in step 3 and calculate the graph Laplacian matrix; Step 5: Based on the graph Laplacian matrix in step 4, a graph convolutional network is used to train the topology model to determine the module connection and update the system topology; Step 6, according to the system topology in step 5, when the device connection changes, collect new data to build an adjacency matrix and adjust the system structure according to the updated topology; Step 7, collect the display interface illumination data according to the system structure of step 6 to build an illumination model and use differentiable rendering to calculate the interface layout to complete the display update.

2. The industrial control cabinet safety control method according to claim 1, characterized in that: In the step S1, when the output data of the quantum arbitrary number generator meets the stable sampling condition, the step further includes: Step 1.1, define the output data set of the quantum arbitrary number generator: R＝{r1,r2,...,r n}， Among them, r1 represents an arbitrary number obtained by the first sampling, r2 represents an arbitrary number obtained by the second sampling, and r n It represents any number obtained by the nth sampling, where n represents the number of sampling times; Compute key candidates: Among them, a i Represents the weight factor corresponding to the i-th sampling, r i represents any number obtained by sampling at the i-th time, and K1 represents a key candidate; Step 1.2, define the reference sequence: T = {t1, t2, ..., t m }, where t1 represents the first element of the reference sequence, t2 represents the second reference value, and t m represents the last element of the reference sequence, m represents the length of the reference sequence, and T represents the reference sequence; Calculate the edit distance: Where D represents the mismatch index between the key candidate K1 and the reference sequence T at each position and β j represents the mismatch index of the jth position, and L represents the comparison length; Step 1.3, define normalized entropy: H = D / L, Where H represents the entropy value of the key candidate, D represents the sum of mismatch indicators between the key candidate K1 and the reference sequence T at each position, and L represents the comparison length; Set the high entropy threshold y. If H≥y, it is determined that the key candidate meets the high entropy requirement.

3. The industrial control cabinet safety control method according to claim 1, characterized in that: In the step S2, after the key candidate determined in step S1 meets the high entropy requirement, the step further includes: Step 2.1, based on the key candidate K1 calculated in step S1 and the reference sequence T, define the key update probability distribution: p K ={p1,p2,...,p n }, Among them, p1 represents the probability that the key candidate K1 takes the first state, p2 represents the probability that the key candidate K1 takes the second state, and p n represents the probability that the key candidate K1 takes the value of the nth state, and constructs the Markov decision process: M = (S, A, P, R, γ), Among them, S is the key state set, A is the update strategy set, P is the state transition probability matrix, R is the state reward function, and γ is the discount factor; Step 2.2, based on the Markov decision process constructed in step 2.1, calculate the state transition probability: Among them, P(s′|s,a) represents the probability that the current state s is transferred to the next state s′ by strategy a, s represents the current state, s′ represents the next state, a represents the key update strategy adopted, and δ i represents the adjustment factor of the i-th state, p n represents the probability that the key candidate K1 takes the value of the nth state, Denotes the probability distribution p of key update K Perform sum calculations; Optimize policy A to maximize the cumulative reward function: J(A)=Σ (t＝0) ∧∞γ t R(s t ,a t ), Among them, J(A) represents the cumulative reward function, γ represents the discount factor, and R(s t ,a t ) means taking strategy a at time step t t and is in state s t The reward value obtained when (t＝0) ∧∞ means cumulative calculation of all time steps t; Step 2.3: After the update strategy A0 optimized in step 2.2 is determined, determine whether to perform the key update according to the key update trigger condition Θ, and define the trigger condition: Θ=f(D,H,η), Where D represents the sum of mismatch indicators between the key candidate K1 and the reference sequence T at each position, H represents the entropy value of the key candidate, and η represents the attack detection feedback signal; If Θ exceeds the set threshold, a key update is performed and the old key K0 is replaced by the new key K1.

4. The industrial control cabinet safety control method according to claim 1, characterized in that: In the step S3, after the key update strategy A0 optimized in the step S2 is determined, the following steps are further included: Step 3.1, based on the new key K1 and update strategy A0 determined in step S2, collect network traffic data: X = {x1, x2, ..., x m }, where x m Represents the data packet collected in the mth time slice, m represents the total number of collected time slices, and defines the network traffic feature vector: F=g(X,K1), Where F is the network traffic feature vector, g(·) is the key-weighted feature extraction function to generate traffic topology data for subsequent attack detection; Step 3.2: Construct a topology data analysis model based on the traffic feature vector F calculated in step 3.1: G = (V, E, W), where V is the set of network nodes, E is the set of connection relationships between nodes, W is the edge weight matrix, and G is the network traffic topology data analysis model; Calculate the topological persistent homology feature: P = h(G), where P is the topological persistent homology feature and h(·) is the topological feature extraction function to identify potential abnormal patterns; Step 3.3, based on the topological feature P extracted in step 3.2, define the attack detection decision function: θ=ψ(P,K1,A0), Wherein, θ is the attack detection function, ψ(·) is the detection function that comprehensively considers topological characteristics, keys and update strategies, and the attack judgment threshold θ0 is set. If θ≥θ0, an alarm signal is triggered, and the attack information η is fed back to step S2 to adjust the key update strategy.

5. The industrial control cabinet safety control method according to claim 1, characterized in that: In the step S4, after determining the network attack characteristics and triggering the alarm signal in step S3, the following steps are further included: Step 4.1, based on the topological data analysis model constructed in step S3: G = (V, E, W), where V is the set of network nodes, E is the set of connections between nodes, G is the network traffic topology data analysis model, W is the edge weight matrix, W = [w ij ], where w ij Represents node v i With v j The strength of the connection between Define the adjacency matrix A: A = [a ij ], where a ij Represents node v i With v j If there is a direct connection: ij =w ij , otherwise a ij =0,v i 、v j represents the i-th and j-th network nodes in the topology model; Calculate the graph Laplacian matrix: L = DA, where D represents the degree matrix and L represents the graph Laplacian matrix, which is used to represent the global information of the network topology and serves as the input for subsequent graph convolution. Step 4.2: Based on the graph Laplacian matrix L calculated in step 4.1, use the graph convolutional network to train the system topology model: M t =f(L,X), Where X is the node feature matrix, f(·) is the graph convolution operation, and the node embedding representation is calculated as: Z = σ(L × W), where W is the training parameter matrix and σ(·) is the activation function to learn how the modules in the topology are connected. Step 4.3: Based on the system topology model M trained in step 4.2 t , define the topology update strategy: Ω=argmaxg(Z,P), where g(·) is the objective function based on node embedding and topological feature optimization. The system topology is adjusted according to Ω, and the topology change information is fed back to step S3 to optimize the subsequent network attack detection and key update strategy.

6. The industrial control cabinet safety control method according to claim 1, characterized in that: In step S5, the graph Laplacian matrix L calculated in step S4 and the training system topology model M t The following further includes: Step 5.1: Based on the topological model M obtained by training in step S4 t , collect device connection change data: Y={y1,y2,...,y k }, where y k Represents the device connection information recorded in the kth time slice, and defines a new adjacency matrix: A′=[a′ ij ], where a′ ij Represents node v i With v j If there is a direct connection: a′ ij =w′ ij , otherwise: a′ ij = 0, calculate the updated graph Laplacian matrix: L′=D′-A′, where D′ is the updated degree matrix; Step 5.2: Based on the updated graph Laplacian matrix L′ calculated in step 5.1, a graph convolutional network is used to calculate the new topological model: M t ′=f(L′,X′), where X′ is the updated node feature matrix and f(·) is the graph convolution operation to calculate the new node embedding representation: Z′=σ(L′X′W′), where W′ is the new training parameter matrix and σ(·) is the activation function to optimize the system topology; Step 5.3: The new topological model M calculated in step 5.2 t ′, define the topology optimization goal: Ω′=argmax g(Z′,ξ), Wherein, g(·) is an optimization function calculated based on node embedding and topology change ξ. The system topology is readjusted according to Ω′, and the new topology adjustment information is fed back to step S4 to further optimize the subsequent network analysis and attack detection.

7. The industrial control cabinet safety control method according to claim 1, characterized in that: In step S6, after the new topology structure is calculated in step S5 and the system topology is adjusted, the following steps are further included: Step 6.1, based on the updated graph Laplacian matrix L′ and the optimized topological structure calculated in step S5, collect the display interface illumination data: I={i1,i2,...,i m }, where i m Represents the illumination data of the mth time slice and defines the illumination model: M I =φ(I,L′), where φ(·) represents a mapping function based on illumination data and topological structure, which is used to calculate the effect of illumination on interface display; Step 6.2: The illumination model M calculated in step 6.1 I , construct the lighting influence matrix: G=[g ij ], where g ij Indicates interface element e i and e j The degree of illumination influence between the two is defined as follows: C = σ(GX I W I ), Among them, X I is the illumination feature matrix of the interface elements, W I is the training parameter matrix, σ(·) is the activation function, and the interface layout adjustment scheme is calculated; Step 6.3, based on the interface layout adjustment scheme C calculated in step 6.2, define the final interface update strategy: Ω I =argmaxf(C,Z′), Wherein, f(·) is the optimization objective function, and the interface adjustment is performed based on the calculated interface illumination information and topological embedding Z′, and the interface adjustment information is fed back to step S5 to optimize the system topology structure and interactive experience.

8. The industrial control cabinet safety control method according to claim 1, characterized in that: In step S7, the interface illumination model M calculated in step S6 is I And after adjusting the interface layout, it further includes: Step 7.1, based on the illumination influence matrix G calculated in step S6 and the optimized interface layout scheme C, collect the final interface display data: D I ={d1,d2,...,d n }, where d n Indicates the interface display status of the kth time slice, and defines the interface display feature matrix: X D =ψ(D I ,C), where ψ(·) is a mapping function based on display data and layout scheme to calculate interface stability and user interaction experience; Step 7.2: The interface display feature matrix X calculated in step 7.1 D , construct the interface stability evaluation function: Among them, α ij For interface elements i and e j The interaction weight between To calculate the relative degree of change of interface elements and evaluate the consistency and stability of interface display; Step 7.3, based on the interface stability S calculated in step 7.2, define the final interface optimization target: Ω D =argmaxg(S,C,G), Among them, g(·) is the optimization objective function that comprehensively considers interface stability, layout scheme and illumination influence. D The final interface adjustment solution is calculated, and the interface optimization information is fed back to step S6 to further optimize the lighting model and interface layout.

9. A terminal device, characterized in that: The invention comprises a processor, a memory and a communication interface for executing the industrial control cabinet safety control method according to any one of claims 1 to 8.

10. A storage medium, characterized in that: A computer program is stored, and when the program is executed, the terminal device implements the industrial control cabinet safety control method as described in any one of claims 1 to 8.

Images