Real-time network security threat early warning analysis method

Through real-time network security threat warning analysis method, combined with anti-quantum encryption technology and intelligent algorithms, the problem of insufficient network threat prediction in the existing technology is solved, real-time identification and prediction of emerging threats is realized, and the intelligence of network security is improved. Data security.

CN119030786AInactive Publication Date: 2024-11-26SHAANXI ZHAOTONG LIHE DIGITAL CO LTD

Patent Information

Application Number
CN202411430132.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-14
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Predictions of cyber threats in prior art rely on historical data and existing models and cannot fully cover all potential threats, especially new or advanced persistent threats.

Method used

Adopt real-time network security threat warning analysis method, combining anti-quantum encryption technology, machine learning, deep learning and generate adversarial networks to achieve real-time identification and prediction of network threats.

Benefits of technology

It improves the level of data security and intelligence of early warning systems, enhances the generalization ability and accuracy of the model, can effectively respond to new or advanced threats, and ensures data confidentiality and integrity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119030786A_ABST
    Figure CN119030786A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of network security, and discloses a real-time network security threat early warning analysis method, which comprises the optimization steps of S1, establishing a network security management system: constructing a solid network security foundation, and setting an early warning detection mechanism for different network segments; s2, information processing: receiving suspicious events and intrusion information through a regional early warning center, performing information preprocessing, determining a detection range, and binding a host IP address; s3, data encryption and real-time database analysis: carrying out data encryption by adopting an anti-quantum encryption technology and a traditional encryption technology, carrying out security early warning analysis on the received encrypted data by the real-time database, identifying potential risk data, and timely isolating and giving an alarm; and S4, designing a prediction model: establishing the prediction model to predict network security events possibly occurring in the future, continuously updating and optimizing the prediction model, and increasing data types by using the generative adversarial network.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of network security technology, and in particular to a real-time network security threat early warning analysis method. Background Art

[0002] With the rapid development of my country's Internet industry, network infrastructure has become the nervous system of social life. The rise of virtualization and mobile technology has expanded the boundaries of security defense. With the continuous deepening of the application of emerging Internet technologies such as big data, cloud computing, the Internet of Things, and the industrial Internet, the degree of enterprise informatization has become higher and higher, and the degree of dependence on information systems has reached an unprecedented level. Accurately judging the severity of current network security threats and predicting their evolution trends, and achieving timely and accurate security threat warnings are of great value to network security action decisions and emergency response.

[0003] After searching, the invention patent with Chinese patent number CN118118251A discloses a real-time network security threat early warning analysis method, which belongs to the field of network security technology. In order to solve the problem that network information attackers are no longer limited to attacking the network with a single weakness, more attackers begin to launch combined attacks, organically organize multiple weaknesses, and then cause serious security threats to the network system. It can be seen that the network security analysis method based on a single weakness can no longer meet the technical problems of the information security needs of individuals and enterprises. The specific approach includes establishing a comprehensive network security management system and real-time monitoring of the integrity of service terminal devices and client devices to respond to network security emergencies. The method is based on the detection mechanism of different network segments, establishes an early warning detection system, which is used to predict current or future possible attacks and conduct network threat assessment, and then adopt appropriate security strategies to respond to intrusion behaviors. Compared with the prior art, the invention patent with Chinese patent number CN118118251A analyzes network vulnerabilities through attack graphs to ensure the safe operation of the system. At the same time, the method has the characteristics of coarse-grained and targeted description of attacks, and establishes a weakness correlation matrix, so that weaknesses in the network can be quickly discovered and repaired.

[0004] However, during the use of the above-mentioned devices, the prediction of network threats relies on historical data and existing models, but these data and models may not fully cover all potential threats, especially new or advanced persistent threats. Therefore, a real-time network security threat early warning analysis method is proposed. Summary of the invention

[0005] The purpose of the present invention is to solve the problem that the prediction of network threats in the prior art relies on historical data and existing models, but these data and models may not fully cover all potential threats, especially new or advanced persistent threats, and propose a real-time network security threat early warning analysis method.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A real-time network security threat early warning analysis method, the optimization steps are:

[0008] S1: Establish a network security management system: Build a solid network security foundation to ensure the integrity of service terminal devices and client devices, and establish early warning detection mechanisms for different network segments. According to different network architectures, set corresponding early warning detection points to timely detect and respond to network threats;

[0009] S2: Information processing: Suspicious events and intrusion information are received through the regional early warning center. The regional early warning center, as the hub of information collection, can aggregate security information from different sources and perform information preprocessing, including filtering, deduplication, formatting and other preprocessing of the collected information. Based on the preprocessed information, the detection range is determined and the host IP address is bound;

[0010] S3: Data encryption and real-time database analysis: Data encryption is performed using quantum-resistant encryption technology and traditional encryption technology. The real-time database performs security early warning analysis on the received encrypted data, identifies potential risk data through intelligent algorithms, and isolates and issues warnings in a timely manner. The intelligent algorithms include machine learning algorithms, deep learning algorithms, and data mining algorithms. Machine learning algorithms are used to identify potential risk patterns, deep learning algorithms are used to identify abnormal patterns in network traffic or abnormal changes in user behavior, and data mining algorithms are used to discover potential security threats. Different intelligent algorithms have different advantages and limitations in different application scenarios. When selecting an intelligent algorithm, it is necessary to make comprehensive considerations based on the specific application scenarios and data characteristics.

[0011] S4: Prediction model design: Establish a prediction model to predict possible cybersecurity events in the future, continuously update and optimize the prediction model, use generative adversarial networks to increase data types, and provide a diverse network threat environment to enhance the generalization ability and accuracy of the model;

[0012] S5: Dynamic resource allocation: Dynamically adjust system resource allocation based on prediction models and suspicious events and intrusion information received through regional early warning centers, increase resource input in high-traffic situations to ensure real-time and accuracy, and optimize resource utilization in low-traffic situations to reduce energy consumption and costs;

[0013] S6: Threat prioritization: Prioritize different types of network threats, develop prioritization rules based on factors such as threat severity and impact, and give priority to high-risk threats.

[0014] The above further includes:

[0015] In S3, data is encrypted using both quantum-resistant encryption technology and traditional encryption technology:

[0016] Use quantum key distribution technology to establish key transmission channels between data centers, and use lattice-based encryption algorithms to encrypt sensitive data stored in the cloud or remote servers;

[0017] When processing large amounts of data, in scenarios where quantum security is not required (such as data transmission in an internal network), choose to encrypt the data using traditional encryption techniques such as those based on quadratic congruence.

[0018] Specific steps to establish a key transmission channel between data centers using quantum key distribution technology:

[0019] Initialization: Select a quantum key distribution protocol, such as BB84 or E91, and prepare a quantum channel, including optical fiber or other quantum communication media;

[0020] Quantum channel establishment: Initialize the quantum channel to ensure that the quantum states at both ends of the channel are consistent. The sender and receiver send and receive quantum bits through the quantum channel.

[0021] Key generation: The sender and receiver measure the received qubits respectively, record the measurement results, and use the same random number generator to generate random numbers corresponding to the measurement results. These random numbers are exchanged through a secure channel to establish the key.

[0022] Key transmission: Since the measurement results in the quantum key distribution process are uncertain, any third-party monitoring will change the measurement results and thus be detected. Both parties transmit the measurement results through a secure channel to establish the key;

[0023] Key verification: Use the key established by both parties for encrypted communication and key verification to ensure the integrity and security of the key;

[0024] Key update: Update keys regularly to improve the security of communications;

[0025] The specific steps of encryption using the lattice-based encryption algorithm are:

[0026] Parameter selection: Select the size and dimension of the lattice. The lattice is a discrete structure defined by a polynomial. The size and dimension of the lattice determine the security strength of the encryption. Select a generator to generate elements in the lattice.

[0027] Key generation: Choose a private key, usually an element in the lattice, and use the private key to generate a public key, which is also an element in the lattice;

[0028] Encryption process: The message to be encrypted is converted into an element in the lattice. This is achieved by selecting a message vector and expanding it into the lattice. The encrypted ciphertext is calculated using the public key and the message vector:

[0029] Assuming the message vector is m and the public key is g, the encryption process can be expressed as:

[0030] c=g·m

[0031] Among them, c is the encrypted ciphertext;

[0032] Decryption process: Use the private key and encrypted ciphertext to calculate the decrypted message. Assuming the private key is s and the encrypted ciphertext is c, the decryption process can be expressed as:

[0033] m=c·s -1

[0034] Among them, m is the decrypted message;

[0035] Security considerations: The security of lattice-based cryptographic methods depends on the choice of lattice size and dimensionality, as well as the generation of private and public keys and their distribution over secure channels.

[0036] In S3, members regularly submit threat intelligence they discover to the real-time database, including attack methods, malware samples, attacker IP addresses, domain names and other information. The real-time database organizes, analyzes and verifies the threat intelligence.

[0037] In S4, a prediction model is established to predict possible cybersecurity incidents in the future, and the prediction model is continuously updated and optimized;

[0038] In the process of building the prediction model, a convolutional neural network is selected as the basic architecture, and malware and benign software samples are used as training data. The dataset generated by the generative adversarial network is input into the neural network model for model training, which includes the following steps:

[0039] initialization:

[0040] Build a model: Build a defined neural network model architecture;

[0041] Initialization parameters: The weights and biases of the model are initialized to small random numbers based on Gaussian distribution;

[0042] Set the optimizer: Select the Adam optimizer to automatically adjust the learning rate;

[0043] Training loop:

[0044] Iterate over training data: Divide the training dataset into batches, each batch contains a fixed number of samples, and for each batch, perform the following steps:

[0045] Forward propagation: Calculate the predicted output for each sample, y pred =f(X,θ), where X is the input data, θ is the model parameter, f is the model function, and y pred is the predicted output;

[0046] Calculate loss: Use the loss function to calculate the error between the predicted output and the true label. Where N is the number of samples in the batch and l is the loss function;

[0047] Back propagation: Calculate the gradient of the loss function with respect to the model parameters through the chain rule, in, is a vector containing the loss function L for all parameters θ i The partial derivative of

[0048] Update parameters: Use the Adam optimizer to update model parameters according to the gradient. Among them, α is the learning rate, and AdamUpdate is the update rule of the Adam optimizer;

[0049] Apply early stopping:

[0050] Monitor validation set performance: At the end of each epoch, use the validation dataset to evaluate the performance of the model;

[0051] Check performance improvement: If the validation performance of the current epoch fails to improve within the predetermined patience period, stop training;

[0052] Save the best model: During the training process, retain the model parameters that perform best on the validation set;

[0053] Output:

[0054] The final output is the model that performs best on the validation set.

[0055] In the update parameters, the update rule of the Adam optimizer is:

[0056] Adam uses two exponential moving averages to estimate the mean and uncentered variance of the gradient, and then updates the parameters based on these estimates:

[0057] m t =β1m t-1 +(1-β1)g t ;

[0058]

[0059] Among them, g t is the current gradient, β1 and β2 are the decay rates;

[0060] The adjusted update rule is:

[0061]

[0062] in, and is m t and t is a bias-corrected version of , where ψ is a small constant that prevents division by zero.

[0063] In S4, the specific steps of using the generative adversarial network to increase the data type are:

[0064] Data preparation:

[0065] Data collection: receiving malware and benign software samples as training data;

[0066] Data preprocessing: perform necessary preprocessing on the collected training data;

[0067] Define the GAN network structure:

[0068] Generator: responsible for generating new data samples similar to real data. During the training process, the goal of the generator is to generate real data to deceive the discriminator.

[0069] Input: random noise vector z;

[0070] Output: synthesized data G(z);

[0071] Discriminator: responsible for distinguishing whether the input data is real data or fake data generated by the generator. During the training process, the goal of the discriminator is to accurately distinguish between real data and generated data;

[0072] Input: real data x or data generated by the generator G(z);

[0073] Output: A probability value D(x), indicating the probability that the input data is true data;

[0074] Define the loss function:

[0075] The training goal of the generative adversarial network is to minimize the loss of the generator while maximizing the loss of the discriminator, which is expressed by the following formula:

[0076]

[0077] Among them, Pdata(x) is the distribution of real data, p z(z) is the distribution of noise, D(x) is the judgment of the discriminator on the real data x, and D(G(z)) is the judgment of the discriminator on the generated data G(z);

[0078] Training process:

[0079] Initialization: Randomly initialize the parameters of the generator and discriminator;

[0080] Iterative training:

[0081] Training the discriminator: fix the parameters of the generator and update the parameters of the discriminator by optimizing the loss function of the discriminator;

[0082] Train the generator: fix the parameters of the discriminator and update the parameters of the generator by optimizing the loss function of the generator;

[0083] Repeat iteration: Repeat the above iterative training until the preset number of training rounds is reached or other stopping conditions are met;

[0084] Synthetic Data:

[0085] Use the trained generator, input the random noise vector z, and generate new data G(z);

[0086] Data Merge:

[0087] The synthesized data G(z) is merged with other training data to form the final data set.

[0088] In S6, different types of cyber threats are prioritized, including the following steps:

[0089] Threat identification: Collect and analyze alerts and logs from various security devices and systems to identify potential threats;

[0090] Threat classification: Threats are classified according to factors such as threat type, source, and impact scope. For example, they can be classified into categories such as viruses, hacker attacks, internal threats, and system vulnerabilities.

[0091] Threat assessment: Conduct risk assessment on each threat, considering its potential impact and likelihood, using quantitative indicators, including the potential impact of the threat (such as data leakage, system paralysis, etc.), the likelihood of the threat occurring (such as known vulnerabilities, historical attacks, etc.), and the complexity of the threat (such as the complexity of the attack technology, using a weighted score method: Threat score = potential impact score × likelihood score × complexity score;

[0092] Prioritization: Based on the threat assessment results, threats are prioritized using a sorting algorithm, such as bubble sort and quick sort, to prioritize threats with the highest scores and the greatest impact.

[0093] Resource allocation: Allocate appropriate resources, such as security experts, protective equipment, response time, etc., based on the priority of the threat to ensure that high-priority threats can be responded to in a timely and adequate manner.

[0094] The present invention has the following beneficial effects:

[0095] 1. In the present invention, dual encryption technology is used in the process of data transmission and storage, including quantum-resistant encryption technology and traditional encryption technology. This dual encryption mechanism greatly improves the security of data and ensures the confidentiality and integrity of data even in the face of new threats such as quantum computing.

[0096] 2. In the present invention, the prediction model can be continuously updated and optimized, the intelligence level and prediction accuracy of the early warning system can be improved, and the generalization ability and accuracy of the model can be enhanced by generating adversarial networks to simulate diverse network threat environments, thereby achieving adaptive learning and intelligent defense against network threats. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 A method step diagram of a real-time network security threat early warning analysis method proposed by the present invention. DETAILED DESCRIPTION

[0098] 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.

[0099] See also Figure 1 As shown, the present invention is a real-time network security threat early warning analysis method, and the optimization steps are:

[0100] S1: Establish a network security management system: Build a solid network security foundation to ensure the integrity of service terminal devices and client devices, and establish early warning detection mechanisms for different network segments. According to different network architectures, set corresponding early warning detection points to timely detect and respond to network threats;

[0101] S2: Information processing: Suspicious events and intrusion information are received through the regional early warning center. The regional early warning center, as the hub of information collection, can aggregate security information from different sources and perform information preprocessing, including filtering, deduplication, formatting and other preprocessing of the collected information. Based on the preprocessed information, the detection range is determined and the host IP address is bound;

[0102] S3: Data encryption and real-time database analysis: Data encryption is performed using quantum-resistant encryption technology and traditional encryption technology. The real-time database performs security early warning analysis on the received encrypted data, identifies potential risk data through intelligent algorithms, and isolates and issues warnings in a timely manner. The intelligent algorithms include machine learning algorithms, deep learning algorithms, and data mining algorithms. Machine learning algorithms are used to identify potential risk patterns, deep learning algorithms are used to identify abnormal patterns in network traffic or abnormal changes in user behavior, and data mining algorithms are used to discover potential security threats. Different intelligent algorithms have different advantages and limitations in different application scenarios. When selecting an intelligent algorithm, it is necessary to make comprehensive considerations based on the specific application scenarios and data characteristics.

[0103] S4: Prediction model design: Establish a prediction model to predict possible cybersecurity events in the future, continuously update and optimize the prediction model, use generative adversarial networks to increase data types, and provide a diverse network threat environment to enhance the generalization ability and accuracy of the model;

[0104] S5: Dynamic resource allocation: Dynamically adjust system resource allocation based on prediction models and suspicious events and intrusion information received through regional early warning centers, increase resource input in high-traffic situations to ensure real-time and accuracy, and optimize resource utilization in low-traffic situations to reduce energy consumption and costs;

[0105] S6: Threat prioritization: Prioritize different types of network threats, develop prioritization rules based on factors such as threat severity and impact, and give priority to high-risk threats.

[0106] In one embodiment, for the above S3, quantum-resistant encryption technology and traditional encryption technology are used to encrypt data in S3:

[0107] Use quantum key distribution technology to establish key transmission channels between data centers, and use lattice-based encryption algorithms to encrypt sensitive data stored in the cloud or remote servers;

[0108] When processing large amounts of data, in scenarios where quantum security is not required (such as data transmission in an internal network), choose to encrypt the data using traditional encryption techniques such as those based on quadratic congruence.

[0109] In one embodiment, for the above quantum key distribution technology, the specific steps of establishing a key transmission channel between data centers using the quantum key distribution technology are:

[0110] Initialization: Select a quantum key distribution protocol, such as BB84 or E91, and prepare a quantum channel, including optical fiber or other quantum communication media;

[0111] Quantum channel establishment: Initialize the quantum channel to ensure that the quantum states at both ends of the channel are consistent. The sender and receiver send and receive quantum bits through the quantum channel.

[0112] Key generation: The sender and receiver measure the received qubits respectively, record the measurement results, and use the same random number generator to generate random numbers corresponding to the measurement results. These random numbers are exchanged through a secure channel to establish the key.

[0113] Key transmission: Since the measurement results in the quantum key distribution process are uncertain, any third-party monitoring will change the measurement results and thus be detected. Both parties transmit the measurement results through a secure channel to establish the key;

[0114] Key verification: Use the key established by both parties for encrypted communication and key verification to ensure the integrity and security of the key;

[0115] Key update: Update keys regularly to improve the security of communications;

[0116] The specific steps of encryption using the lattice-based encryption algorithm are:

[0117] Parameter selection: Select the size and dimension of the lattice. The lattice is a discrete structure defined by a polynomial. The size and dimension of the lattice determine the security strength of the encryption. Select a generator to generate elements in the lattice.

[0118] Key generation: Choose a private key, usually an element in the lattice, and use the private key to generate a public key, which is also an element in the lattice;

[0119] Encryption process: The message to be encrypted is converted into an element in the lattice. This is achieved by selecting a message vector and expanding it into the lattice. The encrypted ciphertext is calculated using the public key and the message vector:

[0120] Assuming the message vector is m and the public key is g, the encryption process can be expressed as:

[0121] c=g·m

[0122] Among them, c is the encrypted ciphertext;

[0123] Decryption process: Use the private key and encrypted ciphertext to calculate the decrypted message. Assuming the private key is s and the encrypted ciphertext is c, the decryption process can be expressed as:

[0124] m=c·s -1

[0125] Among them, m is the decrypted message;

[0126] Security considerations: The security of lattice-based cryptographic methods depends on the choice of lattice size and dimensionality, as well as the generation of private and public keys and their distribution over secure channels.

[0127] In one embodiment, for the above-mentioned S3, in S3, members regularly submit threat intelligence discovered by themselves to the real-time database, including attack methods, malware samples, attacker IP addresses, domain names and other information, and the real-time database organizes, analyzes and verifies the threat intelligence.

[0128] In one embodiment, for the above S4, in S4, a prediction model is established to predict possible network security events in the future, and the prediction model is continuously updated and optimized;

[0129] In the process of building the prediction model, a convolutional neural network is selected as the basic architecture, and malware and benign software samples are used as training data. The dataset generated by the generative adversarial network is input into the neural network model for model training, which includes the following steps:

[0130] initialization:

[0131] Build a model: Build a defined neural network model architecture;

[0132] Initialization parameters: The weights and biases of the model are initialized to small random numbers based on Gaussian distribution;

[0133] Set the optimizer: Select the Adam optimizer to automatically adjust the learning rate;

[0134] Training loop:

[0135] Iterate over training data: Divide the training dataset into batches, each batch contains a fixed number of samples, and for each batch, perform the following steps:

[0136] Forward propagation: Calculate the predicted output for each sample, y pred =f(X,θ), where X is the input data, θ is the model parameter, f is the model function, and y pred is the predicted output;

[0137] Calculate loss: Use the loss function to calculate the error between the predicted output and the true label. Where N is the number of samples in the batch and l is the loss function;

[0138] Back propagation: Calculate the gradient of the loss function with respect to the model parameters through the chain rule, in, is a vector containing the loss function L for all parameters θ i The partial derivative of

[0139] Update parameters: Use the Adam optimizer to update model parameters according to the gradient. Among them, α is the learning rate, and AdamUpdate is the update rule of the Adam optimizer;

[0140] Apply early stopping:

[0141] Monitor validation set performance: At the end of each epoch, use the validation dataset to evaluate the performance of the model;

[0142] Check performance improvement: If the validation performance of the current epoch fails to improve within the predetermined patience period, stop training;

[0143] Save the best model: During the training process, retain the model parameters that perform best on the validation set;

[0144] Output:

[0145] The final output is the model that performs best on the validation set.

[0146] When a new software sample appears and is identified as malware by the model, we can further analyze the behavior and characteristics of the malware to predict the threats and impacts it may cause. For example, if the malware has characteristics such as fast propagation, wide infection range, and the ability to steal sensitive information, then we can predict that large-scale data leaks or cyber attacks may occur in the future.

[0147] In one embodiment, for the above update parameters, in the update parameters, the update rule of the Adam optimizer is:

[0148] Adam uses two exponential moving averages to estimate the mean and uncentered variance of the gradient, and then updates the parameters based on these estimates:

[0149] m t =β1m t-1 +(1-β1)g t ;

[0150]

[0151] Among them, g t is the current gradient, β1 and β2 are the decay rates;

[0152] The adjusted update rule is:

[0153]

[0154] in, and is m t and tis a bias-corrected version of , where ψ is a small constant that prevents division by zero.

[0155] In one embodiment, for the above S4, in S4, the specific steps of increasing the data type by using the generative adversarial network are:

[0156] Data preparation:

[0157] Data collection: receiving malware and benign software samples as training data;

[0158] Data preprocessing: perform necessary preprocessing on the collected training data;

[0159] Define the GAN network structure:

[0160] Generator: Responsible for generating new data samples similar to real data. During the training process, the goal of the generator is to generate real data to deceive the discriminator.

[0161] Input: random noise vector z;

[0162] Output: synthesized data G(z);

[0163] Discriminator: Responsible for distinguishing whether the input data is real data or fake data generated by the generator. During the training process, the goal of the discriminator is to accurately distinguish between real data and generated data;

[0164] Input: real data x or data generated by the generator G(z);

[0165] Output: A probability value D(x), indicating the probability that the input data is true data;

[0166] Define the loss function:

[0167] The training goal of the generative adversarial network is to minimize the loss of the generator while maximizing the loss of the discriminator, which is expressed by the following formula:

[0168]

[0169] Among them, Pdata(x) is the distribution of real data, p z (z) is the distribution of noise, D(x) is the judgment of the discriminator on the real data x, and D(G(z)) is the judgment of the discriminator on the generated data G(z);

[0170] Training process:

[0171] Initialization: Randomly initialize the parameters of the generator and discriminator;

[0172] Iterative training:

[0173] Training the discriminator: fix the parameters of the generator and update the parameters of the discriminator by optimizing the loss function of the discriminator;

[0174] Train the generator: fix the parameters of the discriminator and update the parameters of the generator by optimizing the loss function of the generator;

[0175] Repeat iteration: Repeat the above iterative training until the preset number of training rounds is reached or other stopping conditions are met;

[0176] Synthetic Data:

[0177] Use the trained generator, input the random noise vector z, and generate new data G(z);

[0178] Data Merge:

[0179] The synthesized data G(z) is merged with other training data to form the final data set.

[0180] In one embodiment, for the above S6, in S6, different types of network threats are prioritized, including the following steps:

[0181] Threat identification: Collect and analyze alerts and logs from various security devices and systems to identify potential threats;

[0182] Threat classification: Threats are classified according to factors such as threat type, source, and impact scope. For example, they can be classified into categories such as viruses, hacker attacks, internal threats, and system vulnerabilities.

[0183] Threat assessment: Conduct risk assessment on each threat, considering its potential impact and likelihood, using quantitative indicators, including the potential impact of the threat (such as data leakage, system paralysis, etc.), the likelihood of the threat occurring (such as known vulnerabilities, historical attacks, etc.), and the complexity of the threat (such as the complexity of the attack technology, using a weighted score method: Threat score = potential impact score × likelihood score × complexity score;

[0184] Prioritization: Based on the threat assessment results, threats are prioritized using a sorting algorithm, such as bubble sort and quick sort, to prioritize threats with the highest scores and the greatest impact.

[0185] Resource allocation: Allocate appropriate resources, such as security experts, protective equipment, response time, etc., based on the priority of the threat to ensure that high-priority threats can be responded to in a timely and adequate manner.

[0186] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A real-time network security threat early warning analysis method, It is characterized in that The optimization steps are: S1: Establish a network security management system: Build a solid network security foundation and set up early warning detection mechanisms for different network segments; S2: Information processing: Receive suspicious events and intrusion information through the regional early warning center, perform information preprocessing, determine the detection range and bind the host IP address; S3: Data encryption and real-time database analysis: Data encryption is performed using quantum-resistant encryption technology and traditional encryption technology. The real-time database performs security early warning analysis on the received encrypted data, identifies potential risk data, and isolates and issues warnings in a timely manner. S4: Prediction model design: Receive risk data, build a prediction model to predict possible cybersecurity events in the future, continuously update and optimize the prediction model, and use generative adversarial networks to increase the data type; S5: Dynamic resource allocation: Dynamically adjust system resource allocation based on prediction models and suspicious events and intrusion information received through regional early warning centers; S6: Threat prioritization: Prioritize different types of cyber threats, develop prioritization rules, and prioritize high-risk threats.

2. According to the real-time network security threat early warning analysis method of claim 1, It is characterized in that In S3, data is encrypted using both quantum-resistant encryption technology and traditional encryption technology: Use quantum key distribution technology to establish key transmission channels between data centers, and use lattice-based encryption algorithms to encrypt sensitive data stored in the cloud or remote servers; When processing large amounts of data, in scenarios where quantum security is not required, choose to encrypt the data using traditional encryption technology.

3. A real-time network security threat early warning analysis method according to claim 2, It is characterized in that Specific steps to establish a key transmission channel between data centers using quantum key distribution technology: Initialization: Select the quantum key distribution protocol and prepare the quantum channel; Quantum channel establishment: Initialize the quantum channel to ensure that the quantum states at both ends of the channel are consistent. The sender and receiver send and receive quantum bits through the quantum channel. Key generation: The sender and receiver measure the received qubits respectively, record the measurement results, and use the same random number generator to generate random numbers corresponding to the measurement results. These random numbers are exchanged through a secure channel to establish the key. Key transmission: Both parties transmit measurement results through a secure channel to establish a key; Key verification: Use the key established by both parties to encrypt the communication and perform key verification; Key update: update keys regularly; The specific steps of encryption using the lattice-based encryption algorithm are: Parameter selection: select the size and dimension of the grid, and select a generator to generate the elements in the grid; Key generation: Select a private key and use it to generate a public key; Encryption process: The message to be encrypted is converted into an element in the lattice. This is achieved by selecting a message vector and expanding it into the lattice. The encrypted ciphertext is calculated using the public key and the message vector: Assuming the message vector is m and the public key is g, the encryption process can be expressed as: c=g·m Among them, c is the encrypted ciphertext; Decryption process: Use the private key and encrypted ciphertext to calculate the decrypted message. Assuming the private key is s and the encrypted ciphertext is c, the decryption process can be expressed as: m=c·s -1 Among them, m is the decrypted message; Security considerations: The security of lattice-based cryptographic methods depends on the choice of lattice size and dimensionality, as well as the generation of private and public keys and their distribution over secure channels.

4. According to the real-time network security threat early warning analysis method of claim 3, It is characterized in that In S3, members regularly submit threat intelligence they discover to the real-time database, which organizes, analyzes and verifies the threat intelligence.

5. A real-time network security threat early warning analysis method according to claim 1, It is characterized in that In S4, a prediction model is established to predict possible cybersecurity incidents in the future, and the prediction model is continuously updated and optimized; In the process of building the prediction model, a convolutional neural network is selected as the basic architecture, and malware and benign software samples are used as training data. The dataset generated by the generative adversarial network is input into the neural network model for model training, which includes the following steps: initialization: Build a model: Build a defined neural network model architecture; Initialization parameters: The weights and biases of the model are initialized to small random numbers based on Gaussian distribution; Set the optimizer: Select the Adam optimizer to automatically adjust the learning rate; Training loop: Iterate over training data: Divide the training dataset into batches, each batch contains a fixed number of samples, and for each batch, perform the following steps: Forward propagation: Calculate the predicted output for each sample, y pred =f(X,θ), where X is the input data, θ is the model parameter, f is the model function, and y pred is the predicted output; Calculate loss: Use the loss function to calculate the error between the predicted output and the true label. Where M is the number of output samples, N is the number of samples in the batch, is the loss of a single sample, y true is the actual output; Back propagation: Calculate the gradient of the loss function with respect to the model parameters through the chain rule, in, is a vector containing the loss function L for all parameters θ i The partial derivative of Update parameters: Use the Adam optimizer to update model parameters according to the gradient. Among them, α is the learning rate, and AdamUpdate is the update rule of the Adam optimizer; Apply early stopping: Monitor validation set performance: At the end of each epoch, use the validation dataset to evaluate the performance of the model; Check performance improvement: If the validation performance of the current epoch fails to improve within the predetermined patience period, stop training; Save the best model: During the training process, retain the model parameters that perform best on the validation set; Output: The final output is the model that performs best on the validation set.

6. A real-time network security threat early warning analysis method according to claim 5, It is characterized in that In the update parameters, the update rule of the Adam optimizer is: Adam uses two exponential moving averages to estimate the mean and uncentered variance of the gradient, and then updates the parameters based on these estimates: m t =b 1 m t-1 +(1-β 1 )g t ; Among them, g t is the current gradient, β 1 and β 2 is the decay rate; The adjusted update rule is: in, and is m t and t The bias-corrected version of t and t are the mean estimate and variance estimate respectively, ψ is a small constant to prevent division by zero, and α is the learning rate.

7. A real-time network security threat early warning analysis method according to claim 1, It is characterized in that In S4, the specific steps of using the generative adversarial network to increase the data type are: Data preparation: Data collection: receiving malware and benign software samples as training data; Data preprocessing: perform necessary preprocessing on the collected training data; Define the GAN network structure: Generator: responsible for generating new data samples similar to real data. During the training process, the goal of the generator is to generate real data to deceive the discriminator. Input: random noise vector z; Output: synthesized data G(z); Discriminator: responsible for distinguishing whether the input data is real data or fake data generated by the generator. During the training process, the goal of the discriminator is to accurately distinguish between real data and generated data; Input: real data x or data synthesized by the generator G(z); Output: A probability value D(x), indicating the probability that the input data is true data; Define the loss function: The training goal of the generative adversarial network is to minimize the loss of the generator while maximizing the loss of the discriminator, which is expressed by the following formula: Among them, min C max D V(D,G), E is the mathematical expectation, Pdata(x) is the distribution of the real data, p z (z) is the distribution of noise, D(x) is the judgment of the discriminator on the real data x, and D(G(z)) is the judgment of the discriminator on the synthetic data G(z); Training process: Initialization: Randomly initialize the parameters of the generator and discriminator; Iterative training: Training the discriminator: fix the parameters of the generator and update the parameters of the discriminator by optimizing the loss function of the discriminator; Train the generator: fix the parameters of the discriminator and update the parameters of the generator by optimizing the loss function of the generator; Repeat iteration: Repeat the above iterative training until the preset number of training rounds is reached or other stopping conditions are met; Synthetic Data: Use the trained generator, input the random noise vector z, and generate new synthetic data G(z); Data Merge: The synthesized data G(z) is merged with other training data to form the final data set.

8. A real-time network security threat early warning analysis method according to claim 1, It is characterized in that In S6, different types of cyber threats are prioritized, including the following steps: Threat identification: Collect and analyze alerts and logs from various security devices and systems to identify potential threats; Threat classification: Categorize threats; Threat Assessment: Conduct a risk assessment for each threat, considering its potential impact and likelihood, using quantitative indicators, including the potential impact of the threat, the likelihood of the threat occurring, and the complexity of the threat, using a weighted score approach: Threat score = potential impact score × likelihood score × complexity score; Prioritization: Based on the threat assessment results, a sorting algorithm is used to prioritize threats, giving priority to those with the highest scores and the greatest impact; Resource allocation: Allocate appropriate resources based on threat priority to ensure that high-priority threats can be responded to in a timely and adequate manner.

Citation Information

Patent Citations

  • Generative adversarial network and computer-generated holography-based chip defect detection method

    CN113554636A

  • Enhancement method and device for code vulnerability detection data, equipment and storage medium

    CN117521078A

  • Motion posture recognition data processing method and device based on deep learning model

    CN117727094A

  • Real-time network security threat early warning analysis method

    CN118118251A

  • Systems and methods for post-quantum cryptography optimization

    US11336462B1

Cited By

  • 5G network real-time encryption technology based on artificial intelligence

    CN119421152A

  • Unmanned aerial vehicle navigation signal collaborative blocking system based on quantum key distribution

    CN120614081A

  • Cooperative blocking system for drone navigation signals based on quantum key distribution

    CN120614081B