Console identity recognition method based on block chain

Abstract

The invention discloses a console identity recognition method based on a block chain, relates to the technical field of computer security and human-computer interaction, and realizes non-sensibility and precision of console identity recognition through fusion of behavior feature modeling and a block chain technology. In a multi-person alternate scene, an operator does not need to actively execute a biological verification action, and the system converts behavior characteristics into encrypted operation fingerprints by continuously capturing physical operation actions and graphic interaction logic and carries out uplink storage on the encrypted operation fingerprints; the block chain smart contract compares the time sequence difference between the current operation and the historical fingerprint in real time, and only when the behavior is abnormal, the biological auxiliary verification is triggered, so that the operation interruption caused by the factors such as gloves and light in the industrial environment is remarkably reduced; a dynamic permission label generation mechanism breaks through static authorization limitation; by analyzing the semantic intention of the control instruction and combining a preset equipment permission topological graph, the minimum equipment operation range is automatically delimited, and the tag validity period is set according to the instruction risk level.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer security and human-computer interaction, and in particular to a console identity recognition method based on a block chain. BACKGROUND

[0002] In multi-party collaborative conference scheduling, emergency command and other scenarios, the console often needs to be operated by multiple people in rotation. For example, in a major conference, the chairman, interpreter and technical support personnel need to quickly take over the operation of the same console. In an industrial dispatching center, there is a need for shift operators to hand over. Such scenarios require seamless connection during identity switching to avoid interruptions in meetings or delays in commands. The current mainstream solution uses biometric recognition, such as facial or fingerprint recognition, to bind the operator's identity, which meets the continuous verification principle of the zero trust architecture.

[0003] To meet the need for rapid switching, the mainstream solution mainly uses two types of technology: one is based on environmental variables for rapid identity loading, such as an Azure console that binds to a terminal through a preset key and automatically loads the corresponding user configuration when switching. The second is dynamic biometric detection, which refers to the live detection mechanism of the real person authentication of Ali Cloud, which periodically triggers facial comparison during operation. Some systems combine the OIDC protocol to implement cross-session permission synchronization and update access tokens in real time through a centralized identity provider.

[0004] However, in actual applications, the explicit biometric verification of existing solutions requires interrupting the operation process. For example, fingerprint recognition requires the user to remove their gloves and touch the sensor, which causes a delay in industrial scenarios. In addition, there is a risk of confusion in the handover of permissions. When multiple people quickly rotate, the centralized identity provider may cause the permissions of the previous and subsequent operators to overlap due to synchronization delays. At the same time, physical conditions limit the accuracy of identity recognition, such as changes in lighting affecting the accuracy of facial recognition in conference consoles, or the inability to collect biometric data when operators wear protective equipment. These defects are particularly prominent in scheduling and cross-country meetings that require high continuity. SUMMARY

[0005] In view of the above-mentioned existing problems, the present application is proposed.

[0006] The present application provides a console identity recognition method based on a block chain to solve the problems of interrupting the operation during biometric verification, delaying the synchronization of centralized permissions, and limiting the accuracy of recognition due to physical conditions in multi-party rotation scenarios.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] The present application provides a console identity recognition method based on a block chain, which includes,

[0009] Step S1, the operation behavior characteristics of the operator are collected through the console sensor, and the operation behavior characteristics include a physical operation action sequence and a graphical interface interaction logic sequence;

[0010] Step S2, the operation behavior characteristics are encrypted to generate an operation fingerprint, and a hash value of the operation fingerprint is written into a blockchain network;

[0011] Step S3, the console operation flow is captured in real time, and the current operation behavior characteristics are extracted to generate a temporary verification fingerprint;

[0012] Step S4, the temporary verification fingerprint is compared with historical operation fingerprints stored on the chain through a smart contract deployed on the blockchain;

[0013] Step S5, when the smart contract determines that the matching degree is lower than a dynamic threshold, a biological feature auxiliary verification process is started;

[0014] Step S6, a dynamic permission tag is generated after verification succeeds, and the tag includes an operator digital identity identifier, an operable device range and a tag valid time length;

[0015] Step S7, a control instruction is executed within the permission boundary defined by the dynamic permission tag.

[0016] As a preferred scheme of the console identity recognition method based on the blockchain, in step S1, the physical operation action sequence is collected through a non-contact sensor, and includes gesture space trajectory characteristics and a knob rotation angular velocity sequence.

[0017] The graphical interface interaction logic sequence includes a multi-window cooperative operation event stream, and the event stream is composed of a window focus switching time sequence and a cross-application data dragging path.

[0018] As a preferred scheme of the console identity recognition method based on the blockchain, in step S2, the blockchain network adopts a consortium chain architecture, and consensus nodes include a console terminal, a cloud permission server and an audit terminal.

[0019] The operation fingerprint storage adopts a sharding strategy, and physical operation characteristics are stored in the console terminal node in a sharded manner, and logical operation characteristics are stored in the cloud permission server node in a sharded manner.

[0020] As a preferred scheme of the console identity recognition method based on the blockchain, in step S4, when the smart contract performs comparison:

[0021] A decryption key of the historical operation fingerprint is called;

[0022] A time sequence difference measure of the temporary verification fingerprint and the historical operation fingerprint is calculated;

[0023] Step S5 is activated when the time sequence difference metric exceeds the adaptive threshold.

[0024] In a preferred embodiment of the blockchain-based console identity recognition method described in this invention, in step S4, the calculation method of the time-series difference measure is as follows:

[0025] {P h (t),G h (t)} and {P c (t),G c Linear interpolation of (t)} to a sampling frequency of 100Hz yields discrete sequences of lengths N and M, respectively.

[0026] Construct the local distance matrix:

[0027]

[0028] Among them, D i,j w represents the local distance between row index i and column index j. p P is a 2×2 diagonal matrix of physical feature weights. h (i) is the historical physics operation feature vector, P c (j) is the current physical operation feature vector, ||·||2 is the Euclidean norm, and w g G is a 2×2 diagonal matrix of graphic feature weights. h (i) represents the historical graphical interaction feature vector, G c (j) is the current graphical interaction feature vector, i∈[1,N], j∈[1,M];

[0029] Dynamic programming cumulative distance:

[0030] When i = j = 1, DTW(i,j) = D 1,1 ,

[0031] In other cases, DTW(i,j) = D i,j +min{DTW(i-1,j),DTW(i,j-1),DTW(i-1,j-1)},

[0032] Where DTW(i,j) are the elements of the cumulative distance matrix, D 1,1 Let `min{·}` be the initial local distance, and `min{·}` denotes the operation of selecting the minimum value.

[0033] Normalized variance measure:

[0034]

[0035] Where, Δ TSFor the time difference measure, DTW(N,M) is the cumulative distance right bottom element, N is the history sequence length, M is the current sequence length, exp(·) is the natural exponential function, alpha is the time length mismatch amplification coefficient, |N-M| is the length difference of two sequences, max(N,M) is the larger one;

[0036] The adaptive weight is calculated as:

[0037] w p =diag(k s ,k r ),

[0038] w g =diag(k f ,k d ),

[0039] where w p is the physical feature weight matrix, diag(·) is the diagonal matrix operator, k s is the gesture trajectory weight coefficient, k r is the knob angular velocity weight coefficient, w g is the graphical feature weight matrix, k f is the window focus switching weight coefficient, k d is the data dragging path weight coefficient,

[0040] In the formula:

[0041]

[0042] where k s , k r , k f , k d are the feature weight coefficients, sigma s , sigma r , sigma f , sigma d are the historical variances of gesture trajectory, knob angular velocity, window focus switching, and data dragging path respectively, epsilon is a fine-tuning constant, and beta is a predefined constant.

[0043] As a preferred scheme of the console identity recognition method based on the blockchain, in step S5, the biological feature assisted verification process comprises:

[0044] Collecting biological feature data through a local security chip of the console;

[0045] Generating a zero-knowledge proof of biological features in a trusted execution environment;

[0046] Submitting the zero-knowledge proof to the blockchain for on-chain verification.

[0047] In a preferred embodiment of the blockchain-based console identity recognition method described in this invention, step S6, the generation of the dynamic permission tag includes:

[0048] Analyze the semantic intent of the current control command;

[0049] Map the device's operating range according to semantic operational intent;

[0050] The effective duration of the label is set based on the risk level of the instruction.

[0051] As a preferred embodiment of the blockchain-based console identity recognition method described in this invention, the semantic operation intent parsing adopts a pre-trained natural language processing model, and the input sources include voice command text and graphical interface operation logs.

[0052] The mapping of device operation scope is executed based on the device permission topology map stored in the blockchain.

[0053] As a preferred embodiment of the blockchain-based console identity recognition method described in this invention, the semantic operation intent parsing process in step S6 includes:

[0054] Perform dual-source word segmentation and embedding of instructions:

[0055]

[0056] in, Let E be the i-th initial embedding vector, and E be the size |V|×d. e The word vector lookup table matrix, τ i The terms are discrete word units resulting from the fusion of speech transcription and interface logs, where |V| is the vocabulary size, and d is the word set. e For the embedded dimension;

[0057] Perform context encoding:

[0058]

[0059] Where, r i For contextual representation, BiGRU(·) is a bidirectional gated cyclic update operator;

[0060] Calculate the global attention weights:

[0061]

[0062] Where, β i Let L be the attention weight of the i-th element, u be the trainable global query vector, and L be the weight of the i-th element. t Let T be the length of the word sequence, and T denote the transpose. d is the scaling factor. e For the embedded dimension;

[0063] Generate the intention vector:

[0064]

[0065] Wherein, v is the semantic intention vector of the current instruction;

[0066] Calculate the prototype similarity, the formula is:

[0067]

[0068] Wherein, s k The cosine similarity of the kth device operation prototype p k , ‖·‖2 is the two norm, K is the total number of prototypes, p k The kth device operation prototype vector;

[0069] Define the intention probability distribution as:

[0070]

[0071] Wherein, P k The kth intention probability, λ is the temperature scaling coefficient;

[0072] In the formula:

[0073]

[0074] Wherein, λ is the temperature scaling coefficient, κ risk The reference coefficient set according to the instruction risk level, The similarity variance of the last period, ∈ is the fine-tuning constant;

[0075] Intention determination:

[0076]

[0077] Wherein, I * The semantic operation intention label determined.

[0078] As a preferred scheme of the console identity recognition method based on the blockchain, wherein, when the operation authority is handed over:

[0079] The decay trend of the operation behavior characteristic matching degree is predicted to predict the handover time point;

[0080] A dynamic authority label draft of the relieving operator is generated at the predicted time point;

[0081] The new and old authority label states are synchronously updated through the blockchain consensus mechanism.

[0082] The application has the advantages that: the application realizes the inattentiveness and precision of console identity recognition by the fusion of behavior characteristic modeling and blockchain technology. In the multi-person rotation scene, the operator does not need to actively perform biological verification actions, and the system converts the behavior characteristics into encrypted operation fingerprints and stores them on the chain by continuously capturing physical operation actions (such as rotary trajectory of knob) and graphic interaction logic (such as window switching path). The blockchain smart contract compares the time sequence difference between the current operation and the historical fingerprint in real time, and only when the behavior is abnormal, the biological auxiliary verification is triggered, which significantly reduces the operation interruption caused by gloves, light and other factors in the industrial environment.

[0083] The dynamic permission tag generation mechanism breaks through the limitation of static authorization. By analyzing the semantic intention of the control instruction (such as the voice command "adjust the temperature of area A"), combining the pre-set device permission topology graph, the minimum device operation range is automatically determined, and the label validity period is set according to the instruction risk level. This design completely eliminates the confusion risk in the permission transfer, ensures the strict isolation of the control domain of different operators, and meets the audit requirements of high security scenes.

[0084] The permission transfer process is intelligently predicted by the behavior attenuation trend. The system predicts the transfer time point based on the historical data on the chain, generates a draft of the successor's permission in advance, and synchronously updates it through blockchain consensus, realizing seamless switching of new and old permission states. This mechanism avoids the synchronization delay defect of centralized identity providers, and at the same time adapts to the rapid rotation demand in emergency scheduling scenarios.

[0085] The application further enhances environmental adaptability. Non-contact sensors collect gesture space trajectory and other characteristics, avoiding physical contact restrictions; the adaptive weight mechanism in the time difference measurement can suppress false positives caused by operation rhythm fluctuations, improving robustness in complex industrial environments. The blockchain sharding storage strategy considers efficiency and security, with physical characteristics stored in local terminal nodes and logical characteristics stored in cloud nodes, ensuring controllable sensitive data while supporting cross-terminal verification. BRIEF DESCRIPTION OF DRAWINGS

[0086] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0087] Figure 1 The flowchart of the console identity recognition method based on blockchain in embodiment 1. DETAILED DESCRIPTION

[0088] In order to make the above objectives, characteristics and advantages of the present application more apparent, a detailed description of the specific embodiments of the present application will be given below with reference to the accompanying drawings.

[0089] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herewith. In other instances, well-known methods have not been described in detail in order to avoid unnecessarily obscuring the present application.

[0090] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent or alternative to other embodiments.

[0091] Embodiment 1, reference Figure 1 The embodiment provides a console identity recognition method based on a blockchain, comprising the following steps:

[0092] Step S1, collecting the operation behavior characteristics of the operator through the console sensor, the operation behavior characteristics including a physical operation action sequence and a graphical interface interaction logic sequence;

[0093] In step S1, the physical operation action sequence is collected through a non-contact sensor, including gesture space trajectory characteristics and knob rotation angular velocity sequence;

[0094] The graphical interface interaction logic sequence contains a multi-window cooperative operation event stream, and the event stream is composed of window focus switching time sequence and cross-application data dragging path;

[0095] Step S2, encrypting the operation behavior characteristics to generate an operation fingerprint, and writing the hash value of the operation fingerprint into a blockchain network;

[0096] In step S2, the blockchain network adopts a consortium chain architecture, and its consensus nodes include a console terminal, a cloud permission server and an audit terminal;

[0097] The operation fingerprint storage adopts a sharding strategy, and the physical operation characteristics are stored in the console terminal node in a sharded manner, and the logical operation characteristics are stored in the cloud permission server node in a sharded manner;

[0098] Step S3, capturing the console operation stream in real time and extracting the current operation behavior characteristics to generate a temporary verification fingerprint;

[0099] Step S4, comparing the temporary verification fingerprint with the historical operation fingerprint stored on the chain through a smart contract deployed on the blockchain;

[0100] In step S4, the smart contract performs the comparison when:

[0101] The decryption key of the historical operation fingerprint is called;

[0102] The time difference measure between the temporary verification fingerprint and the historical operation fingerprint is calculated;

[0103] Step S5 is activated when the time difference measure exceeds the adaptive threshold;

[0104] In step S4, the time difference measure is calculated in the following manner:

[0105] {P h (t),G h (t)} are linearly interpolated to a sampling frequency of 100 Hz to obtain discrete sequences of lengths N and M, respectively; c c

[0106] The local distance matrix is constructed as follows:

[0107]

[0108] where D i,j represents the local distance at row index i and column index j, w p is a 2x2 physical feature weight diagonal matrix, P h (i) is the historical physical operation feature vector, P c (j) is the current physical operation feature vector, ‖·‖2 is the Euclidean norm, w g is a 2x2 graphical feature weight diagonal matrix, G h (i) is the historical graphical interaction feature vector, G c (j) is the current graphical interaction feature vector, i∈[1,N] and j∈[1,M];

[0109] The dynamic programming cumulative distance is constructed as follows:

[0110] When i=j=1, DTW(i,j)=D 1,1 ,

[0111] In other cases, DTW(i,j)=D i,j +min{DTW(i-1,j),DTW(i,j-1),DTW(i-1,j-1)},

[0112] where DTW(i,j) is the cumulative distance matrix element, D 1,1 is the initial local distance, and min{·} represents the minimum value operation;

[0113] The normalized difference measure is constructed as follows:

[0114] ​​

[0115] wherein, Δ TS is the time difference measure, DTW(N,M) is the cumulative distance right bottom element, N is the history sequence length, M is the current sequence length, exp(·) is the natural exponential function, a is the time length mismatch amplification coefficient, |N-M| is the length difference of two sequences, max(N,M) is the larger one;

[0116] The adaptive weight is calculated as follows:

[0117] w p =diag(k s ,k r ),

[0118] w g =diag(k f ,k d ),

[0119] wherein, w p is the physical feature weight matrix, diag(·) is the diagonal matrix operator, k s is the gesture trajectory weight coefficient, k r is the knob angular velocity weight coefficient, w g is the graphical feature weight matrix, k f is the window focus switching weight coefficient, k d is the data dragging path weight coefficient,

[0120] In the formula:

[0121]

[0122] wherein, k s ,k r ,k f ,k d are the feature weight coefficients, σ s ,σ r ,σ f ,σ d are the history variances of gesture trajectory, knob angular velocity, window focus switching and data dragging path respectively, and ∈ is a fine-tuning constant 10 -6 ;

[0123] Specifically, the time difference measure maps heterogeneous operation trajectories to a unified time axis, suppresses unstable dimensions through variance-driven weights, and is sensitive to rhythm mutations through exponential compensation terms, which can capture abnormalities in real time and output stable scales while maintaining tolerance for normal rhythms.

[0124] Step S5, when the smart contract determines that the matching degree is lower than the dynamic threshold, a biometric auxiliary verification process is started.

[0125] In step S5, the biometric-assisted verification process includes:

[0126] Biometric data is collected via the local security chip in the console.

[0127] Generate biometric zero-knowledge proofs in a trusted execution environment;

[0128] Submit zero-knowledge proofs to the blockchain for on-chain verification;

[0129] Step S6: After successful verification, a dynamic permission tag is generated. This tag includes the operator's digital identity, the range of operable devices, and the tag's validity period.

[0130] In step S6, the generation of dynamic permission tags includes:

[0131] Analyze the semantic intent of the current control command;

[0132] Map the device's operating range according to semantic operational intent;

[0133] The effective duration of the label is set based on the risk level of the instruction.

[0134] Semantic operation intent parsing uses a pre-trained natural language processing model, with input sources including voice command text and graphical interface operation logs;

[0135] The mapping of device operation scope is executed based on the device permission topology map stored in the blockchain.

[0136] The semantic operation intent parsing process in step S6 includes:

[0137] Perform dual-source word segmentation and embedding of instructions:

[0138]

[0139] in, Let E be the i-th initial embedding vector, and E be the size |V|×d. e The word vector lookup table matrix, τ i The terms are discrete word units resulting from the fusion of speech transcription and interface logs, where |V| is the vocabulary size, and d is the word set. e For the embedded dimension;

[0140] Context encoding (bidirectional gated recurrent unit):

[0141]

[0142] Where, r i For contextual representation, BiGRU(·) is a bidirectional gated cyclic update operator;

[0143] Calculate the global attention weights:

[0144]

[0145] where β i is the i-th attention weight, u is a trainable global query vector, L t is the length of the token sequence, T represents the transpose, is a scaling factor, d e is the embedding dimension;

[0146] Generate the intent vector:

[0147]

[0148] where v is the semantic intent vector of the current instruction;

[0149] Calculate the prototype similarity, the formula is:

[0150]

[0151] where s k is the cosine similarity with the k-th device operation prototype p k , ‖·‖2 is the two-norm, K is the total number of prototypes, p k is the k-th device operation prototype vector;

[0152] Define the intent probability distribution as:

[0153]

[0154] where P k is the probability of the k-th intent, λ is the temperature scaling coefficient;

[0155] In the formula:

[0156]

[0157] where λ is the temperature scaling coefficient, κ risk is the reference coefficient set according to the risk level of the instruction, is the similarity variance of the last period, ∈ is a fine-tuning constant 10 -6 ;

[0158] Make an intent determination:

[0159]

[0160] where I * is the determined semantic operation intent label;

[0161] Specifically, the analysis process uniformly maps multi-source instructions to context vectors, and then focuses on key words with global attention to weaken the influence of modal particles and fillers. The combination of cosine similarity and temperature scaling makes the model output a sharper probability distribution in high-risk instruction scenarios, thereby facilitating the subsequent permission generation module to interpret.

[0162] Step S7, executing the control instruction within the permission boundary defined by the dynamic permission tag;

[0163] When operating the permission handover:

[0164] Predicting the handover time point based on the decay trend of the matching degree of the operation behavior characteristics;

[0165] Generating a draft of the dynamic permission tag of the relieving operator at the predicted time point;

[0166] Synchronously updating the new and old permission tag states through a blockchain consensus mechanism.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1.A blockchain-based console identity recognition method, characterized in that, The method comprises the following steps: Step S1, collecting the operation behavior characteristics of the operator through the console sensor, the operation behavior characteristics including a physical operation action sequence and a graphical interface interaction logic sequence; Step S2, encrypting the operation behavior characteristics to generate an operation fingerprint, and writing the hash value of the operation fingerprint into a blockchain network; Step S3, capturing the console operation flow in real time and extracting the current operation behavior characteristics to generate a temporary verification fingerprint; Step S4, comparing the temporary verification fingerprint with the historical operation fingerprint stored on the chain through a smart contract deployed on the blockchain; Step S5, when the smart contract determines that the matching degree is lower than a dynamic threshold, starting a biological feature auxiliary verification process; Step S6, generating a dynamic permission tag after successful verification, the tag including an operator digital identity, an operable device range and a tag valid time length; Step S7, executing a control instruction within the permission boundary defined by the dynamic permission tag. 2.The console identity identification method based on blockchain of claim 1, wherein, In step S1, the physical operation action sequence is collected by a non-contact sensor, including gesture space trajectory characteristics and knob rotation angular velocity sequence; The graphical interface interaction logic sequence includes a multi-window cooperative operation event stream, and the event stream is composed of window focus switching time sequence and cross-application data dragging path. 3.The console identity identification method based on blockchain of claim 1, wherein, In step S2, the blockchain network adopts a consortium chain architecture, and its consensus nodes include a console terminal, a cloud permission server and an audit terminal; The operation fingerprint storage adopts a sharding strategy, and the physical operation characteristics are stored in the console terminal node in a sharded manner, and the logical operation characteristics are stored in the cloud permission server node in a sharded manner. 4.The console identity identification method based on blockchain of claim 1, wherein, In step S4, when the smart contract performs comparison: The decryption key of the historical operation fingerprint is called; The time sequence difference measure of the temporary verification fingerprint and the historical operation fingerprint is calculated; When the time sequence difference measure exceeds an adaptive threshold, step S5 is activated. 5.The console identity identification method based on blockchain of claim 4, wherein, In step S4, the calculation method of the time sequence difference measure is as follows: {P h (t),G h (t)} and {P c (t),G c (t)} are linearly interpolated to a 100 Hz sampling frequency, resulting in discrete sequences of length N and M, respectively; Construct a local distance matrix: where D i,j denotes the local distance at row index i and column index j, w p is a 2x2 physical feature weight diagonal matrix, P h (i) is the historical physical operation feature vector, P c (j) is the current physical operation feature vector, ||·||2 is the Euclidean norm, w g is a 2x2 graphical feature weight diagonal matrix, G h (i) is the historical graphical interaction feature vector, G c (j) is the current graphical interaction feature vector, i∈[1,N], j∈[1,M]; Dynamically program the accumulated distance: DTW(i, j) = D when i = j = 1 1,1 , Otherwise, DTW(i,j) = D i,j + min{DTW(i-1,j), DTW(i,j-1), DTW(i-1,j-1)}, where DTW(i,j) is the cumulative distance matrix element, D 1,1 is the initial local distance, and min{•} denotes the minimum operation. Normalize the difference measure: where Δ TS is the time difference measure, DTW(N,M) is the cumulative distance bottom-right element, N is the history sequence length, M is the current sequence length, exp(·) is the natural exponential function, a is the time length mismatch amplification coefficient, |N-M| is the length difference of two sequences, and max(N,M) is the larger one. Calculate the adaptive weight: w p = diag(k s ,k r ), w g = diag(k f ,k d ), wherein w p is a physical feature weight matrix, diag(·) is a diagonal matrix generator operator, k s is a gesture trajectory weight coefficient, k r is a knob angular velocity weight coefficient, w g is a graphical feature weight matrix, k f is a window focus switching weight coefficient, k d is a data dragging path weight coefficient, In step S5, the biological feature auxiliary verification process includes: where k s ,k r ,k f ,k d are feature weight coefficients, σ s ,σ r ,σ f ,σ d are history variances of gesture trajectory, knob angular velocity, window focus switching, and data dragging path, respectively, ∈ is a fine-tuning constant, and is a pre-defined constant. 6.The console identity identification method based on blockchain of claim 1, wherein, Collecting biological feature data through a local security chip of the console; Generating a biological feature zero-knowledge proof in a trusted execution environment; Submitting the zero-knowledge proof to the blockchain for on-chain verification. In step S6, the generation of the dynamic permission tag includes: 7.The console identity identification method based on blockchain of claim 1, wherein, Analyzing the semantic operation intention of the current control instruction; Mapping the device operation range according to the semantic operation intention; Setting the tag valid time length based on the instruction risk level. The semantic operation intention analysis adopts a pre-trained natural language processing model, and the input sources include voice instruction text and graphical interface operation log; 8.The console identity identification method based on blockchain of claim 7, wherein, The device operation range mapping is performed according to the device permission topology graph stored in the blockchain. The semantic operation intention analysis process in step S6 includes: 9.The console identity identification method based on blockchain of claim 8, wherein, Performing instruction double-source word segmentation and embedding: Performing context encoding: wherein, is the i-th initial embedding vector, E is a word embedding lookup matrix of size |V| x d e is the i-th initial embedding vector, E is a word embedding lookup matrix of size |V| x d i is the i-th initial embedding vector, E is a word embedding lookup matrix of size |V| x d e is the i-th initial embedding vector, E is a word embedding lookup matrix of size |V| x d Calculating global attention weight: wherein r i is a contextual representation, BiGRU(·) is a bidirectional gated recurrent update operator; Generating an intention vector: where β i is the i-th attention weight, u is a trainable global query vector, L t is the length of the token sequence, T denotes the transpose, is a scaling factor, d e is the embedding dimension; Wherein, v is the semantic intention vector of the current instruction; Calculating the prototype similarity, the formula is: Defining the intention probability distribution as: where s k is the cosine similarity with the kth device operation prototype p k , ||·||2 is the two-norm, K is the total number of prototypes, p k is the kth device operation prototype vector; The formula is: where P k is the probability of the kth intent, and λ is a temperature scaling factor. Performing intention determination: where λ is a temperature scaling factor, κ risk is a reference coefficient set according to the instruction risk level, is a variance of the similarity of the previous period, and ∈ is a fine-tuning constant. When the operation permission is transferred: wherein I * is the determined semantic operation intent label. 10.The console identity identification method based on blockchain of claim 1, wherein, Based on the decay trend prediction of the operation behavior characteristic matching degree, the transfer time point is predicted; ​ Generating a dynamic permission tag draft for the relieving operator at the prediction time point; Synchronously updating the new and old permission tag states through a blockchain consensus mechanism.

Images