Network data security alarm system based on artificial intelligence

Through the network data security alarm system based on artificial intelligence, users' operation flow data and network performance data are collected and analyzed in real time, and abnormal operation identification rules are established, which solves the problem that existing systems are difficult to accurately identify abnormal operations, and achieves efficient network data security Monitor and respond in real time.

CN119030765APending Publication Date: 2024-11-26GUANGDONG COUNTRY GARDEN VOCATIONAL COLLEGE
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411133304.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing network data security alarm system is difficult to accurately identify abnormal operations, resulting in missed and false alarms, and cannot detect and block potential threats in real time, and cannot solve network security problems in a timely manner.

Method used

A network data security alarm system based on artificial intelligence was designed. Through the data acquisition module, the user operation flow data and network performance data are collected in real time. The feature extraction module extracts key features. The data analysis module establishes abnormal operation identification rules and conducts risk assessment. The signal reception and alarm module sends alarms to monitor and warning potential threats in real time.

Benefits of technology

It significantly improves the system's identification accuracy and efficiency of abnormal operations, can monitor user behavior in real time, timely identify and warning of potential abnormal operation risks, and improves the protection ability and response speed of network data security.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119030765A_ABST
    Figure CN119030765A_ABST
Patent Text Reader

Abstract

The invention discloses a network data security alarm system based on artificial intelligence, and relates to the technical field of network security, and the system comprises the steps: collecting user operation flow data and network performance data in real time through a data collection module, and carrying out the preprocessing; the feature extraction module performs feature extraction on the preprocessed data so as to identify user abnormal operation and network abnormal fluctuation; the data analysis module comprises abnormal operation identification rule establishment, user operation flow data analysis and network performance data analysis, an abnormal operation identification rule is established through historical data, and risk assessment is carried out on real-time data; the abnormal signal receiving and alarm module receives the abnormal signal of the data analysis module, carries out corresponding network security analysis and sends an alarm; the system can timely discover abnormal operation of a user and abnormal network performance, provides an accurate alarm, helps an administrator to timely take measures to deal with network threats, and improves the security and stability of network data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of network security, and in particular to a network data security alarm system based on artificial intelligence. Background Art

[0002] With the rapid development of Internet technology, cyberspace has become an indispensable part of modern society; however, network data security issues have become increasingly prominent, with frequent hacker attacks, malware, data leaks and other incidents, posing huge threats to the information security of individuals, enterprises and even the country; traditional network security protection measures, such as firewalls, intrusion detection systems, etc., can no longer meet the current complex and ever-changing network attack needs; therefore, the network data security alarm system based on artificial intelligence technology has emerged, which provides a revolutionary solution for network data security protection with its powerful data processing capabilities and intelligent learning characteristics;

[0003] However, existing user behavior analysis technologies still have some problems in identifying and handling abnormal operations, which makes it very difficult to establish accurate abnormal operation identification rules. For example, some abnormal operations may be mistaken for normal operations, which will lead to missed reports and false alarms. In addition, traditional network data security alarm systems can usually only respond to security incidents after the fact and cannot detect and prevent potential threats in real time, which makes it impossible to solve network security problems in a timely manner and may cause greater losses. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the technical problems in the background technology, the present invention proposes a network data security alarm system based on artificial intelligence, comprising: collecting user operation flow data and network performance data in real time through a data acquisition module, and performing preprocessing; a feature extraction module extracts features from the preprocessed data to identify abnormal user operations and abnormal network fluctuations; a data analysis module includes establishing abnormal operation identification rules, user operation flow data analysis and network performance data analysis, establishing abnormal operation identification rules through historical data, and performing risk assessment on real-time data; an abnormal signal receiving and alarm module receives the abnormal signal of the data analysis module, performs corresponding network security analysis, and sends an alarm; thereby solving the technical problems recorded in the background technology.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0008] A network data security alarm system based on artificial intelligence, comprising:

[0009] The data collection module includes a user data collection unit, a network data collection unit and a preprocessing unit, which collects user operation flow data and network performance data in real time and performs corresponding preprocessing;

[0010] Feature extraction module: extracts features from pre-processed user operation flow data and network performance data;

[0011] Data analysis module: including abnormal operation identification rule establishment unit, user operation flow data analysis unit and network performance data analysis unit; establish abnormal operation identification rules based on operation flow data and user operation data in historical data; analyze the extracted network traffic characteristics and network performance characteristics respectively, and conduct network risk assessment from two aspects of user operation and network performance; among which:

[0012] The abnormal operation identification rule establishment unit analyzes different abnormal operation types and corresponding operation flow data in the historical network detection data, and determines the influencing characteristic attributes of each abnormal operation type and the weight β corresponding to each influencing characteristic attribute. i ;

[0013] The user operation flow data analysis unit combines the preprocessed real-time operation flow data with the abnormal operation identification rules, and calculates the abnormal operation risk score according to the value range of the feature value and the preset user operation risk score. j Calculate the risk index Fx of each abnormal operation j , and further analyze to obtain the risk index Fx of the current user operation;

[0014] The network performance data analysis unit processes and analyzes the network performance data. When no real-time data processing instruction is received, the unit monitors the changes of the network performance data in real time, records the duration of each network performance data exceeding the corresponding safety threshold, and obtains the average abnormal time T of the network performance in each cycle. When receiving the real-time data processing instruction, the unit performs real-time analysis on the time domain characteristics of the network performance data to obtain the real-time risk index SFx of the network performance.

[0015] Abnormal signal receiving and alarm module: Receives abnormal signals transmitted from the data analysis module, performs corresponding network security analysis based on the type of signal, and sends corresponding alarms to user networks with abnormal network security analysis results.

[0016] Specifically, the user operation flow data is the flow data generated by the user operation; the extracted network traffic features are saved in the feature set T1, T1 = {A1, A2, ..., A n}, where A irepresents the ith network traffic feature, and n represents the number of network traffic features. Feature selection is performed from network performance data, including four network performance features: packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter.

[0017] Furthermore, all abnormal user operations in the historical data are counted and a user abnormal operation set T2 is established, where T2 = {X1, X2, …, X m}, X i represents the ith abnormal operation, including the abnormal operation name and the characteristic attribute set t generated when the user performs the abnormal operation, m is the number of abnormal operation types appearing in the historical data, and the characteristic attribute set t is initially an empty set;

[0018] The data with the same abnormal operation type in the historical data are gathered together, and these different types of abnormal operations are analyzed one by one. The mutual information value I(A) between each network traffic feature and each abnormal operation is calculated by the symmetric uncertainty algorithm. i ;X j ) and the mutual information value I(A) between each network traffic feature and other network traffic features i ; A k );

[0019] According to the mutual information value I(A i ;X j ) and the mutual information value I(A) between each network traffic feature and other features i ; A k ) Calculate the characteristic estimation index Gj of each network traffic feature i , specifically:

[0020]

[0021] in, represents the average mutual information value between each network traffic feature and the abnormal operation used, that is, It represents the average mutual information value between each network traffic feature and other network traffic features, that is,

[0022] A feature that is more strongly correlated with each abnormal operation and less correlated with other features is selected as the correlated feature of each abnormal operation; and a corresponding weight is set for the correlated feature of each abnormal operation according to the mutual information value between each correlated feature and each abnormal operation.

[0023] Furthermore, an associated feature set is set for each abnormal operation, and the calculated maximum feature estimation index Gj is iThe corresponding network traffic features are put into the associated feature set of the corresponding abnormal operation; the next feature estimation index Gj is selected from the remaining network traffic features j The largest network traffic features are placed into the associated feature set;

[0024] Calculate the characteristic estimation index value Gj of two network traffic characteristics ij , if the calculated characteristic estimation index value Gj ij Less than the original feature estimation index comprehensive value Gj of the newly added feature j , then delete this indicator from the associated feature set; compare each network traffic feature in turn until the feature combination with the largest comprehensive value of the feature estimation indicator is found, which is used as the associated feature set of this abnormal operation.

[0025] Furthermore, the calculated mutual information value between each associated feature and the corresponding abnormal operation is taken out, and the mean of the mutual information values ​​of all associated features in the associated feature set is calculated. And variance σ, and normalize it, the result is the weight index β of each associated feature i , the expression is:

[0026]

[0027] Among them, I i Represents the mutual information value between the i-th associated feature in the associated feature set and the corresponding abnormal operation.

[0028] Furthermore, all the associated features in the associated feature set of all abnormal operations are taken out, and their value ranges when judged as abnormal user operations in historical data are counted respectively; the value range of the feature is divided into multiple intervals of equal size, and the frequency of all feature values ​​appearing in each interval is counted as the frequency P of each interval. i ;

[0029] Start judging from the associated features with the largest mutual information value in the associated feature set of each abnormal operation, find the features of the same type as the associated features with the largest mutual information value from the features acquired in real time, and record the values ​​of these features. i As the influence of this feature on the user's abnormal operation, Yx i ;

[0030] The influence of all the associated features of each user's abnormal operation on the corresponding user's abnormal operation is combined with the weight index of each associated feature to form the prediction index Yc of each user's abnormal operation. j , the expression is:

[0031]

[0032] Among them, Yc j represents the j-th user's abnormal operation, and N represents the total number of associated features;

[0033] The predicted index Yc of each user's abnormal operation j is combined with the preset risk score Score of each user's abnormal operation j to form the risk degree index Fx of each user's abnormal operation j , and the maximum value of the risk degree index among the user's abnormal operations is used as the risk degree index Fx of the current user's operation, that is where j is a positive integer not greater than m;

[0034] Set the risk degree index threshold Fx0. If Fx < Fx0, it means that the user's operation is normal and no abnormal operation has occurred, and no signal is sent; if Fx ≥ Fx0, it means that the user is suspected of having an abnormal operation. At this time, a user suspected abnormal operation signal is sent to the abnormal signal receiving and processing module and the control center; after the system receives the signal, a real-time data processing instruction is sent to the network performance data analysis unit.

[0035] Furthermore, when the network performance data analysis unit does not receive the real-time data processing instruction, it continuously monitors the specific values of the network performance data and compares them with the corresponding security thresholds. When it is found that the value of a certain network performance data exceeds the corresponding security threshold, the abnormal duration of the data is counted, and the abnormal durations of each network performance data are combined in real time to obtain the average abnormal time T of the current network performance. The expression is:

[0036] FFx = α1 * t ab + α2 * t sy + α3 * t dk + α4 * t sy

[0037] where α1, α2, α3, and α4 are the preset weight coefficients of the packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter amount respectively, and α2 > α1 > α3 > α4 > 0, and α1 + α2 + α3 + α4 = 1.

[0038] Further, compare the calculated average abnormal time T of the network performance with the network performance abnormal time threshold T0. If T = 0, it indicates that the current network performance has no abnormality. If T < T0 and T ≠ 0, it indicates that the current network performance may have an abnormality and needs further analysis and confirmation. Specifically, record the average abnormal time T1 calculated this time in the cache stack. Within the preset time, if it is detected again that T2 < T0 and T2 ≠ 0, take out T1 in the cache stack and add it to the current T2 to form a new average abnormal time T, and then compare it with the network performance abnormal time threshold T0 again to make the same judgment as above. If T ≥ T0, it indicates that the network performance has an abnormality in this period, and send the network performance data abnormal signal to the abnormal signal receiving and processing module.

[0039] Further, when the network performance data analysis unit receives a real-time data processing instruction, analyze the time-domain characteristics of the real-time monitored network performance data. Specifically, starting from receiving the real-time data processing instruction, record the real-time packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter amount data, and record the real-time values in four plane rectangular coordinate systems in real time to form four irregular rays, with the abscissa being time. Use a straight line to fit these irregular rays in real time and record the slope of the fitted straight line in real time to obtain the packet loss rate straight line slope K1, the server resource utilization rate straight line slope K2, the bandwidth occupancy rate data straight line slope K3, and the delay jitter amount straight line slope K4. Combine the four slopes to obtain the real-time risk degree index SFx of the network performance. The expression is:

[0040]

[0041] Compare the calculated real-time risk degree index SFx of the network performance with the network performance real-time risk degree index threshold SFx0. If SFx < SFx0, it indicates that no abnormality is found in the change of the network performance, and no signal is sent. If SFx ≥ SFx0, it indicates that an abnormality is found in the change of the network performance, and send the real-time risk degree index abnormal signal to the abnormal signal receiving and processing module.

[0042] Specifically, if no abnormal signal is received, it indicates that no abnormality is found in the website under the user's operation, and no alarm and signal are sent.

[0043] If a user suspected abnormal operation signal is received and a real-time risk degree index abnormal signal is received within the preset determination time, it indicates that the user has indeed performed the corresponding abnormal operation. Obtain the most likely abnormal operation of the user from the user operation flow data analysis unit and pop up the corresponding abnormal operation alarm on the user's network interface.

[0044] If a user's suspected abnormal operation signal is received, but no real-time risk index abnormal signal is received within the preset judgment time, it means that the user's suspected abnormal operation has not caused abnormal network performance. The monitoring results are recorded and the user operation is continuously tracked. If multiple abnormal operation signal false alarms occur, a signal for manual detection is sent to the control center.

[0045] If a network performance data abnormality signal is received but no user suspected abnormal operation signal is received, it means that the abnormal operation attribute performed by the user does not appear in the historical data, and an alarm of low operation security is sent to the user and a signal for manual detection is sent to the control center.

[0046] (III) Beneficial effects

[0047] The present invention provides a network data security alarm system based on artificial intelligence, which has the following beneficial effects:

[0048] 1. The data acquisition module collects and pre-processes user operation flow data and network performance data in real time, extracts key features to identify abnormal user operations and abnormal network fluctuations, and is the basis for effectively improving the system's monitoring and alarm capabilities for potential network data security threats;

[0049] 2. By establishing an abnormal operation identification rule unit, we conduct in-depth analysis of abnormal operations and operation flow data in historical network detection data, use the symmetric uncertainty algorithm to calculate the mutual information value, effectively identify the characteristic attributes that are strongly related to each abnormal operation and relatively independent, and set reasonable weights for these characteristic attributes; this significantly improves the system's recognition accuracy and efficiency for abnormal operations, and provides a more reliable guarantee for network data security;

[0050] 3. Through the user operation flow data analysis unit, the real-time operation flow data is deeply analyzed, and combined with the abnormal operation identification rules, the risk index of each abnormal operation is effectively calculated, and the risk index of the current user operation is further obtained; it can monitor user behavior in real time, identify and warn potential abnormal operation risks in a timely manner, and significantly improve the system's protection ability and response speed for network data security;

[0051] 4. Comprehensively monitor and analyze network performance data through the network performance data analysis unit, detect network performance changes in real time, and effectively calculate the average abnormal time and real-time risk index of network performance; it can timely discover and warn of network performance anomalies, provide the system with reliable network status monitoring and security protection, and significantly improve the maintenance capability and response efficiency of network data security. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1This is a schematic diagram of the structure of the artificial intelligence-based network data security alarm system provided by the present invention. DETAILED DESCRIPTION

[0053] 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.

[0054] refer to Figure 1 The present invention provides a network data security alarm system based on artificial intelligence, comprising:

[0055] Data collection module: including user data collection unit, network data collection unit and preprocessing unit, collects user operation flow data and network performance data in real time, and performs corresponding preprocessing;

[0056] User operation flow data refers to the real-time behavior flow data of users, specifically the flow data generated when users operate on the website, including the amount of data transmitted during each operation, the speed of data transmission, the number of data packets, the size of data packets, and the time interval between data packet arrivals, etc., reflecting the status and behavior of network traffic. Through further analysis, it can be determined whether the user has performed abnormal operations, and then the risk value of the user operation can be obtained; network performance data refers to the network performance data during user operations, including data such as delay jitter, network packet loss rate, and server resource utilization rate, reflecting the performance and status of the network, so as to facilitate timely measures to deal with possible network attacks;

[0057] When the control center detects that a user is operating, a data collection instruction is generated and sent to the user data collection unit and the network data collection unit. After receiving the data collection instruction, the user data collection unit collects the user operation flow data generated by the user operation in real time through the network traffic analysis tool NetFlow. After receiving the data collection instruction, the network data collection unit collects the network performance data in real time through the network monitoring tool Wireshark and the server monitoring tool Zabbix.

[0058] The collected raw data often contains noise and redundant information, and needs to be cleaned and preprocessed, using regular expressions, string processing functions and other methods to organize the data into the required format and structure.

[0059] Feature extraction module: extracts features from pre-processed user operation flow data and network performance data;

[0060] The user operation flow data mainly includes network traffic features such as traffic size and rate, number of packets, length, and time interval; these features are used to identify whether the user has performed abnormal operations; the extracted network traffic features are saved in the feature set T1, T1 = {A1, A2, ..., A n}, where A i represents the ith network traffic feature, and n represents the number of network traffic features;

[0061] The network performance data mainly includes packet loss rate, server resource utilization rate, bandwidth occupancy rate and delay jitter; these characteristics are used to determine whether the network has abnormal fluctuations.

[0062] Data analysis module: including abnormal operation identification rule establishment unit, user operation flow data analysis unit and network performance data analysis unit; establish abnormal operation identification rules based on operation flow data and user operation data in historical data; analyze the extracted network traffic characteristics and network performance characteristics respectively, and conduct network risk assessment from two aspects of user operation and network performance;

[0063] The abnormal operation identification rule establishment unit analyzes different abnormal operation types and corresponding operation flow data in the historical network detection data, and determines the influencing characteristic attributes of each abnormal operation type and the weight β corresponding to each influencing characteristic attribute. i , including the following steps:

[0064] S11. Count all abnormal user operations in historical data and establish a user abnormal operation set T2, where T2 = {X1, X2, ..., X m}, X i represents the ith abnormal operation, including the abnormal operation name and the characteristic attribute set t generated when the user performs the abnormal operation, m is the number of abnormal operation types appearing in the historical data, and the characteristic attribute set t is initially an empty set;

[0065] S12, gather the data with the same abnormal operation type in the historical data, analyze these different types of abnormal operations one by one, and calculate the mutual information value I (A) between each network traffic feature and each abnormal operation through the symmetric uncertainty algorithm i ;X j ), the expression is:

[0066]

[0067] Among them, p(A i ) represents the network traffic feature A i The probability of occurrence in historical data, p(X j ) indicates abnormal operation X jThe probability of occurrence in historical data, p(A i ,X j ) represents the network traffic feature A in historical data i With abnormal operation X j The joint probability of

[0068] And the mutual information value I(A i ; A k ), the expression is:

[0069]

[0070] Among them, p(A i ,A k ) represents the network traffic feature A in historical data i With network traffic characteristics A k The joint probability of

[0071] Mutual information is a measure of the degree of mutual dependence between features and label classes, that is, the correlation index between each feature and each abnormal operation. The larger the mutual information value, the greater the degree of mutual dependence between the feature and the label class.

[0072] S13, according to the mutual information value I(A) between each network traffic feature and each abnormal operation i ;X j ) and the mutual information value I(A) between each network traffic feature and other features i ; A k ) Calculate the characteristic estimation index Gj of each network traffic feature i , specifically:

[0073]

[0074] in, represents the average mutual information value between each network traffic feature and the abnormal operation used, that is, It represents the average mutual information value between each network traffic feature and other network traffic features, that is,

[0075] S14, select the feature that is more strongly correlated with each abnormal operation and less correlated with other features as the relevant feature of each abnormal operation; set a related feature set for each abnormal operation, and calculate the maximum feature estimation index Gj i The corresponding network traffic features are put into the associated feature set of the corresponding abnormal operation; the next feature estimation index Gj is selected from the remaining network traffic features j The largest network traffic feature is placed in the associated feature set; the feature estimation index value Gj of the two network traffic features is calculated ij, if the calculated characteristic estimation index value Gj ij Less than the original feature estimation index comprehensive value Gj of the newly added feature j , then delete this indicator from the associated feature set; compare each network traffic feature in turn until the feature combination with the largest comprehensive value of the feature estimation indicator is found, which is used as the associated feature set of this abnormal operation;

[0076] S15, according to the mutual information value between each associated feature and each abnormal operation, set a corresponding weight for the associated feature of each abnormal operation; take out the mutual information value between each associated feature and the corresponding abnormal operation calculated in step S12, and calculate the mean of the mutual information values ​​of all associated features in the associated feature set And variance σ, and normalize it, the result is the weight index β of each associated feature i , the expression is:

[0077]

[0078] Among them, I i Represents the mutual information value between the i-th associated feature in the associated feature set and the corresponding abnormal operation.

[0079] The user operation flow data analysis unit combines the preprocessed real-time operation flow data with the abnormal operation identification rules, and calculates the abnormal operation risk score according to the value range of the feature value and the preset user operation risk score. j Calculate the risk index Fx of each abnormal operation j , and further analyze to obtain the risk index Fx of the current user operation. The specific steps are as follows:

[0080] S21, extract all the associated features from the associated feature set of all abnormal operations, and count the value ranges of the features when they are judged as abnormal user operations in the historical data; divide the value range of the feature into multiple intervals of equal size, and count the frequencies of all feature values ​​appearing in each interval as the frequency P of each interval. i , the frequency P of each interval i As a basis for determining whether the user has performed corresponding abnormal operations based on the associated features;

[0081] S22, starting from the associated features with the largest mutual information value in the associated feature set of each abnormal operation, find out the features of the same type as the associated features with the largest mutual information value from the features acquired in real time, and record the values ​​of these features, and calculate the frequency P of the interval where the actual values ​​of these features are located. i As the influence of this feature on the user's abnormal operation, Yx i ;

[0082] S23. Combine the influence degree of the value of all associated features of each user's abnormal operation on the corresponding user's abnormal operation with the weight index of each associated feature to form the prediction index Yc of each user's abnormal operation. j , the expression is:

[0083]

[0084] Among them, Yc j represents the jth user's abnormal operation, and N represents the total number of associated features;

[0085] S24. Combine the prediction index Yc of each user's abnormal operation j with the preset risk score Score of each user's abnormal operation j to form the risk degree index Fx of each user's abnormal operation j . Take the maximum value of the risk degree index in the user's abnormal operation as the risk degree index Fx of the current user's operation, that is where j is a positive integer not greater than m;

[0086] S25. Set the risk degree index threshold Fx0. If Fx < Fx0, it means that the user's operation is normal and no abnormal operation has occurred, and no signal is sent; if Fx ≥ Fx0, it means that the user is suspected of having an abnormal operation. At this time, send a user suspected abnormal operation signal to the abnormal signal receiving and processing module and the control center; after the system receives the signal, send a real-time data processing instruction to the network performance data analysis unit.

[0087] The network performance data analysis unit processes and analyzes the network performance data. When it does not receive a real-time data processing instruction, it monitors the changes in the network performance data in real time, records the duration of each network performance data exceeding the corresponding security threshold, and obtains the average abnormal time T of the network performance in each cycle; when it receives a real-time data processing instruction, it performs real-time analysis on the time-domain characteristics of the network performance data to obtain the real-time risk degree index SFx of the network performance.

[0088] The packet loss rate refers to the proportion of data packets lost due to various reasons in network transmission; in a security system, the packet loss rate may affect the integrity and timeliness of data, and thus affect the security and availability of the system.

[0089] The server resource utilization rate refers to the occupancy of server hardware resources (such as CPU, memory, disk, etc.); if the resource utilization rate of the server is too high, it may lead to a decline in system performance, and even cause the system to crash or be denied service.

[0090] The bandwidth occupancy rate refers to the proportion of the bandwidth occupied by data traffic in network transmission; a high bandwidth occupancy rate may affect other applications and performance of the network, such as increased network latency, slowed data transmission speed, etc.; in a security system, the bandwidth occupancy rate may affect the system's defense ability against external attacks, such as DDoS attacks, etc.

[0091] Jitter is the variation in the delay of different data packets in the same traffic flow; although jitter mainly reflects network performance, an abnormal increase in them may imply potential security issues in the network; for example, the network being maliciously occupied, the routing being tampered with, or network devices being attacked may all lead to an increase in delay and jitter.

[0092] When the network performance data analysis unit does not receive real-time data processing instructions, it monitors the specific values of network performance data in real time and compares them with the corresponding security thresholds. When it is found that the value of a certain network performance data exceeds the corresponding security threshold, the abnormal duration of this data is counted, and the abnormal duration of each network performance data is combined in real time to obtain the average abnormal time T of the current network performance. The expression is:

[0093] FFx = α1 * t db + α2 * t sy + α3 * t dk + α4 * t sy

[0094] Where α1, α2, α3, and α4 are the preset weight coefficients of the packet loss rate, server resource utilization rate, bandwidth occupancy rate, and jitter amount respectively, and α2 > α1 > α3 > α4 > 0, α1 + α2 + α3 + α4 = 1;

[0095] The calculated average abnormal time T of network performance is compared with the network performance abnormal time threshold T0; if T = 0, it means that the current network performance has not shown abnormalities; if T < T0 and T ≠ 0, it means that the current network performance may have abnormalities and needs further analysis and confirmation. Specifically, the average abnormal time T1 calculated this time is recorded in the cache stack. Within the preset time, if it is detected again that T2 < T0 and T2 ≠ 0, then T1 in the cache stack is taken out and added to the current T2 to form a new average abnormal time T, and it is compared with the network performance abnormal time threshold T0 again for the same judgment as above; if T ≥ T0, it means that network performance abnormalities are found in this cycle, and the network performance data abnormal signal is sent to the abnormal signal receiving and processing module.

[0096] When the network performance data analysis unit receives a real-time data processing instruction, it analyzes the time-domain characteristics of the real-time monitored network performance data. Specifically, starting from receiving the real-time data processing instruction, it records the real-time packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter amount data, and records the real-time values in four plane rectangular coordinate systems in real time, forming four irregular rays, with the abscissa being time; it uses a straight line to fit these irregular rays in real time and records the slope of the fitted straight line in real time to obtain the packet loss rate straight line slope K1, the server resource utilization rate straight line slope K2, the bandwidth occupancy rate data straight line slope K3, and the delay jitter amount straight line slope K4; combining the four slopes, it obtains the real-time risk degree index SFx of the network performance, and the expression is:

[0097]

[0098] Compare the calculated real-time risk degree index SFx of the network performance with the network performance real-time risk degree index threshold SFx0. If SFx < SFx0, it means that no abnormality is found in the change of the network performance, and no signal is sent; if SFx ≥ SFx0, it means that an abnormality is found in the change of the network performance, and the real-time risk degree index abnormal signal is sent to the abnormal signal receiving and processing module.

[0099] Abnormal signal receiving and alarm module: Receives the abnormal signal passed in from the data analysis module, conducts corresponding network security analysis according to the type of the signal, and sends corresponding alarms to the user network with abnormal network security analysis results.

[0100] If no abnormal signal is received, it means that no abnormality is found in the website under the user's operation, and no alarm or signal is sent;

[0101] If a user suspected abnormal operation signal is received and a real-time risk degree index abnormal signal is received within the preset determination time, it means that the user has indeed performed the corresponding abnormal operation. Obtain the most likely abnormal operation of the user from the user operation flow data analysis unit, and pop up a corresponding abnormal operation alarm on the user's network interface;

[0102] If a user suspected abnormal operation signal is received, but a real-time risk degree index abnormal signal is not received within the preset determination time, it means that the suspected abnormal operation performed by the user does not cause an abnormality in the network performance. Record the results monitored this time and conduct continuous tracking of the user's operation; if there are multiple false alarms of abnormal operation signals, send a signal for manual detection to the control center;

[0103] If a network performance data abnormality signal is received but no user suspected abnormal operation signal is received, it means that the abnormal operation attribute performed by the user does not appear in the historical data, and an alarm of low operation security is sent to the user and a signal for manual detection is sent to the control center.

[0104] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer storage medium or transmitted through a computer storage medium. The computer instructions can be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer storage medium can be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

[0105] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A network data security alarm system based on artificial intelligence, characterized by: including: A data acquisition module, including a user data acquisition unit, a network data acquisition unit, and a preprocessing unit, which collects user operation flow data and network performance data in real time and performs corresponding preprocessing; A feature extraction module: extracts features from the preprocessed user operation flow data and network performance data; A data analysis module: including an abnormal operation recognition rule establishment unit, a user operation flow data analysis unit, and a network performance data analysis unit; establishes an abnormal operation recognition rule according to the operation flow data and user operation data in historical data; analyzes the extracted network traffic features and network performance features respectively, and conducts network risk assessment from two aspects of user operation and network performance; where: The abnormal operation identification rule establishment unit analyzes different abnormal operation types and corresponding operation flow data in the historical network detection data, and determines the influencing characteristic attributes of each abnormal operation type and the weight β corresponding to each influencing characteristic attribute. i ; The user operation flow data analysis unit combines the preprocessed real-time operation flow data with the abnormal operation identification rules, and calculates the abnormal operation risk score according to the value range of the feature value and the preset user operation risk score. j Calculate the risk index Fx of each abnormal operation j , and further analyze to obtain the risk index Fx of the current user operation; The network performance data analysis unit, by processing and analyzing network performance data, when not receiving a real-time data processing instruction, monitors the change of network performance data in real time, records the duration of each network performance data exceeding the corresponding security threshold, and obtains the average abnormal time T of network performance for each cycle; when receiving a real-time data processing instruction, conducts real-time analysis on the time-domain features of network performance data to obtain the real-time risk degree index SFx of network performance; An abnormal signal receiving and alarm module: receives the abnormal signals transmitted in the data analysis module, conducts corresponding network security analysis according to the types of the signals, and sends corresponding alarms to the user networks with abnormal results in network security analysis.

2. The network data security alarm system based on artificial intelligence according to claim 1, characterized in that: The user operation flow data is the flow data generated by the user operation; the extracted network traffic features are saved in the feature set T1, T1 = {A1, A2, ..., A n }, where A i represents the ith network traffic feature, and n represents the number of network traffic features. Feature selection is performed from network performance data, including four network performance features: packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter.

3. The network data security alarm system based on artificial intelligence according to claim 2, characterized in that: All abnormal user operations in historical data are counted, and a user abnormal operation set T2 is established, where T2 = {X1, X2, ..., X m }, X i represents the ith abnormal operation, including the abnormal operation name and the characteristic attribute set t generated when the user performs the abnormal operation, m is the number of abnormal operation types appearing in the historical data, and the characteristic attribute set t is initially an empty set; The data with the same abnormal operation type in the historical data are gathered together, and these different types of abnormal operations are analyzed one by one. The mutual information value I(A) between each network traffic feature and each abnormal operation is calculated by the symmetric uncertainty algorithm. i ;X j ) and the mutual information value I(A) between each network traffic feature and other network traffic features i ; A k ); According to the mutual information value I(A i ;X j ) and the mutual information value I(A) between each network traffic feature and other features i ; A k ) Calculate the characteristic estimation index Gj of each network traffic feature i , specifically: in, represents the average mutual information value between each network traffic feature and the abnormal operation used, that is, It represents the average mutual information value between each network traffic feature and other network traffic features, that is, Select the features that are more relevant to each abnormal operation and less relevant to other features as the relevant features of each abnormal operation; set corresponding weights for the relevant features of each abnormal operation according to the mutual information value between each associated feature and each abnormal operation.

4. The network data security alarm system based on artificial intelligence according to claim 3, characterized in that: Set the associated feature set for each abnormal operation and calculate the maximum feature estimation index Gj i The corresponding network traffic features are put into the associated feature set of the corresponding abnormal operation; the next feature estimation index Gj is selected from the remaining network traffic features j The largest network traffic features are placed into the associated feature set; Calculate the characteristic estimation index value Gj of two network traffic characteristics ij , if the calculated characteristic estimation index value Gj ij Less than the original feature estimation index comprehensive value Gj of the newly added feature j , then delete this indicator from the associated feature set; Compare each network traffic feature in turn until the feature combination with the largest comprehensive value of the feature estimation index is found as the associated feature set of this abnormal operation.

5. The network data security alarm system based on artificial intelligence according to claim 4, characterized in that: Take out the calculated mutual information value of each associated feature and the corresponding abnormal operation, and calculate the mean of the mutual information values ​​of all associated features in the associated feature set. And variance σ, and normalize it, the result is the weight index β of each associated feature i , the expression is: Among them, I i Represents the mutual information value between the i-th associated feature in the associated feature set and the corresponding abnormal operation.

6. The network data security alarm system based on artificial intelligence according to claim 5, characterized in that: Take out all the associated features from the associated feature set of all abnormal operations, and count their value ranges when they are judged as abnormal user operations in historical data; divide the value range of the feature into multiple intervals of equal size, and count the frequency of all feature values ​​appearing in each interval as the frequency P of each interval i ; Start judging from the associated features with the largest mutual information value in the associated feature set of each abnormal operation, find the features of the same type as the associated features with the largest mutual information value from the features acquired in real time, and record the values ​​of these features. i As the influence of this feature on the user's abnormal operation, Yx i ; The influence of all the associated features of each user's abnormal operation on the corresponding user's abnormal operation is combined with the weight index of each associated feature to form the prediction index Yc of each user's abnormal operation. j , the expression is: Among them, Yc j represents the jth user’s abnormal operation, and N represents the total number of associated features; The prediction index Yc of each user's abnormal operation j The risk score preset for each user's abnormal operation j Combined to form the risk index Fx of each user's abnormal operation j , the maximum value of the risk index in the user's abnormal operation is taken as the risk index Fx of the current user operation, which is Wherein j is a positive integer not greater than m; Set a risk degree index threshold Fx0. If Fx < Fx0, it means that the user operation is normal and no abnormal operation occurs, and no signal is sent; if Fx ≥ Fx0, it means that the user is suspected of having an abnormal operation. At this time, send a user suspected abnormal operation signal to the abnormal signal receiving and processing module and the control center; after the system receives the signal, send a real-time data processing instruction to the network performance data analysis unit.

7. The network data security alarm system based on artificial intelligence according to claim 6, characterized in that: When the network performance data analysis unit does not receive a real-time data processing instruction, it monitors the specific values of network performance data in real time and compares them with the corresponding security thresholds. When it is found that the value of a certain network performance data exceeds the corresponding security threshold, the abnormal duration of this data is counted. The average abnormal time T of the current network performance is obtained by combining the abnormal durations of each network performance data in real time. The expression is: FFx=α1*t db +α2*t sy +α3*t dk +α4*t sy Among them, α1, α2, α3, and α4 are the preset weight coefficients of the packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter amount respectively, and α2 > α1 > α3 > α4 > 0, α1 + α2 + α3 + α4 = 1.

8. A network data security alarm system based on artificial intelligence according to claim 7, characterized in that: The calculated average abnormal time T of the network performance is compared with the network performance abnormal time threshold T0; if T = 0, it means that the current network performance does not show any abnormality; if T < T0 and T ≠ 0, it means that the current network performance may be abnormal and needs further analysis and confirmation. Specifically, the average abnormal time T1 calculated this time is recorded in the cache stack. Within the preset time, if it is detected again that T2 < T0 and T2 ≠ 0, then T1 in the cache stack is taken out and added to the current T2 to form a new average abnormal time T, and then compared with the network performance abnormal time threshold T0 again, and the same judgment as above is made; if T ≥ T0, it means that the network performance is found to be abnormal during this period, and the network performance data abnormal signal is sent to the abnormal signal receiving and processing module.

9. A network data security alarm system based on artificial intelligence according to claim 8, characterized in that: When the network performance data analysis unit receives a real-time data processing instruction, it analyzes the time-domain characteristics of the real-time monitored network performance data. Specifically, starting from the receipt of the real-time data processing instruction, it records the real-time packet loss rate, server resource utilization rate, bandwidth occupancy rate, and delay jitter amount data, and records the real-time values in four plane rectangular coordinate systems in real time to form four irregular rays, with the abscissa being time; Use a straight line to fit these irregular rays in real time and record the slope of the fitted straight line in real time to obtain the packet loss rate straight line slope K1, the server resource utilization rate straight line slope K2, the bandwidth occupancy rate data straight line slope K3, and the delay jitter amount branch line slope K4; combine the four slopes to obtain the real-time risk degree index SFx of the network performance. The expression is: The calculated real-time risk degree index SFx of the network performance is compared with the network performance real-time risk degree index threshold SFx0. If SFx < SFx0, it means that no abnormality is found in the change of the network performance, and no signal is sent; if SFx ≥ SFx0, it means that an abnormality is found in the change of the network performance, and the real-time risk degree index abnormal signal is sent to the abnormal signal receiving and processing module.

10. A network data security alarm system based on artificial intelligence according to claim 9, characterized in that: If no abnormal signal is received, it means that no abnormality is found in the website under the user's operation, and no alarm or signal is sent. If a suspected abnormal operation signal from the user is received, and a real-time risk index abnormal signal is received within the preset judgment time, it means that the user has indeed performed the corresponding abnormal operation, and the abnormal operation most likely to be performed by the user is obtained from the user operation flow data analysis unit, and the corresponding abnormal operation alarm is popped up in the network interface of the user; If a user's suspected abnormal operation signal is received, but no real-time risk index abnormal signal is received within the preset judgment time, it means that the user's suspected abnormal operation has not caused abnormal network performance. The monitoring results are recorded and the user operation is continuously tracked; If there are multiple abnormal operation signal false alarms, a signal for manual detection will be sent to the control center; If a network performance data abnormality signal is received but no user suspected abnormal operation signal is received, it means that the abnormal operation attribute performed by the user does not appear in the historical data, and an alarm of low operation security is sent to the user and a signal for manual detection is sent to the control center.

Citation Information

Cited By

  • Network security abnormity intelligent monitoring method and system

    CN120263561A

  • Network security data analysis method based on artificial intelligence

    CN120354418A

  • Network security situation contrastive analysis method and device based on artificial intelligence, and medium

    CN121000477A