Deep learning-based magnetic induction smart card security access authentication method and system

Through deep learning technology, the signal data and user behavior characteristics of magnetic induction smart cards are processed, combined with communication and behavioral risk assessment, the problem of failure to effectively analyze user behavior patterns in the existing technology is solved, and more comprehensive security analysis and risk assessment are achieved, and more comprehensive security analysis and risk assessment are enhanced. The security of magnetic induction smart card.

CN118965328AInactive Publication Date: 2024-11-15SHENZHEN LIANGBIDA TECH CO LTD

Patent Information

Application Number
CN202411108306.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art ignores the analysis of user behavior patterns in the authentication process of magnetic induction smart cards, and fails to effectively combine the authentication process with user behavior patterns, resulting in the potential security risks being under-assessed.

Method used

Using a deep learning-based method, the signal data between the smart card and the card reader is processed through a convolutional neural network, delay, retransmission rate and noise information are collected, and the user's behavior characteristics and timeliness information is combined to quantify the user's security trust scores and generate Communication risk assessment and behavioral risk assessment, dynamically manage user security status.

Benefits of technology

A more comprehensive security analysis and risk assessment is achieved, and by identifying risks in the authentication process and user behavior risks, providing more accurate security trust scores, enhancing the security and dynamic management capabilities of magnetic induction smart cards.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965328A_ABST
    Figure CN118965328A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of access authentication, in particular to a magnetic induction smart card security access authentication method and system based on deep learning, and the method comprises the steps: processing signal data transmitted between a smart card and a card reader through a convolutional neural network; the method comprises the following steps: acquiring delay information, retransmission rate information and noise information generated by a successfully authenticated user in an authentication process, quantifying risks existing in the authentication process of the user, comparing behavior characteristics of the user in a database after successful authentication with behavior characteristics of other users in the database, determining similar information of the user, and performing authentication according to the similar information. And according to the historical login behavior of the user, determining the timeliness information of the user, quantifying the risk degree of the behavior information of the user, and determining the security trust score of the user, so that more comprehensive security analysis can be obtained, and comprehensive risk assessment and dynamic security management can be realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of access authentication, and in particular to a magnetic induction smart card secure access authentication method and system based on deep learning. Background Art

[0002] Magnetic induction smart cards are authenticated through communication with the card reader. However, various risks may occur in this process, such as delays, data retransmission, noise interference, etc. However, the existing technology ignores the fact that the user's behavior pattern after successful authentication may also reveal potential security risks. It is unable to combine the authentication process with the user's behavior pattern for analysis, and ignores the detailed information in the authentication process and the analysis of the user's subsequent behavior. Summary of the invention

[0003] The purpose of the present invention is to provide a magnetic induction smart card secure access authentication method and system based on deep learning to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions: S1: Process the signal data transmitted between the smart card and the card reader through a convolutional neural network, and collect the delay information, retransmission rate information, and noise information generated by the successfully authenticated user during the authentication process; S2: Quantify the risk of users in the authentication process based on the comprehensive processing of delay information, retransmission rate information and noise information; S3: Compare the user's behavior characteristics in the database after successful authentication with the behavior characteristics of other users in the database to determine the user's similar information, and determine the user's timeliness information based on the user's historical login behavior; S4: Comprehensively process the user's similarity information and timeliness information, quantify the risk level of the user's behavior information, and determine the user's security trust score based on the risks faced by the user during the authentication process.

[0005] Preferably, the delay information includes: The collected delay information is processed and represented by the delay fluctuation coefficient. The logic for obtaining the delay fluctuation coefficient is as follows: set the monitoring interval and the size of the sliding window, and mark the data in each sliding window as: , where n=1, 2, 3, ..., N, N is a positive integer, j=1, 2, 3, ..., J, J is a positive integer, n represents the number of the sliding window in the monitoring interval, and j represents the number of the data in the nth sliding window; Get the standard deviation of the data in each sliding window and mark the standard deviation of the data in each sliding window as: BZ, where , ; A reference threshold of the delay standard deviation is set, and the standard deviation calculated in each sliding window is compared with the preset reference threshold. If the standard deviation calculated in the sliding window exceeds the reference threshold, the sliding window is marked as an abnormal window, and the number of abnormal windows in the monitoring interval is obtained and marked as: , the delay fluctuation coefficient is calculated by the ratio of the number of abnormal windows in the monitoring interval to the number of sliding windows in the monitoring interval. The calculation formula of the delay fluctuation coefficient is: ,in, is the delay fluctuation coefficient.

[0006] Preferably, the retransmission rate information includes: The collected retransmission rate information is processed and represented by a retransmission rate anomaly coefficient. The acquisition logic of the retransmission rate anomaly coefficient is as follows: obtaining the retransmitted data packets within the monitoring interval, and determining the retransmitted data packet sequence according to the transmission time sequence of the retransmitted data packets; The data is stabilized, and the retransmission data packet sequence is modeled using the autoregressive moving average model. The future retransmission rate is predicted using the autoregressive moving average model. The order of the autoregressive moving average model is determined, and the processed retransmission data packet sequence is fitted with the autoregressive moving average model. The expression of the autoregressive moving average model is: ;in, is the retransmission rate at time point t, p is the order of the autoregressive part, q is the order of the moving average part, is the error term at time point t, , ,……, is the autoregressive parameter, , ,……, is the moving average parameter; Use the fitted autoregressive moving average model to predict the retransmission rate at future time points, collect the actual retransmission rate value at future time points and the predicted retransmission rate value at future time points, and calculate the deviation between the predicted value and the actual value; Set the residual threshold, compare the calculated residual with the residual threshold, if the calculated residual is greater than the residual threshold, mark the time point as an abnormal time point, if the calculated residual is less than the residual threshold, mark the time point as a normal time point, obtain the number of abnormal time points and normal time points, calculate the retransmission rate abnormal coefficient, the calculation formula is: ;in, is the retransmission rate anomaly coefficient, is the number of abnormal time points, is the number of normal time points.

[0007] Preferably, the noise information includes: The collected noise information is processed and represented by the signal-to-noise deviation coefficient. The logic for obtaining the signal-to-noise deviation coefficient is: obtain the signal strength and noise strength in the monitoring interval, and mark the signal strength and noise strength in the monitoring interval as: and , i=1, 2, 3, ..., I, I is a positive integer, i is the number of data collected within the monitoring interval; Set the signal strength threshold and the noise strength threshold, and mark the signal strength threshold and the noise strength threshold as: and ; Calculate the signal strength deviation using the formula: in, is the signal strength deviation; Calculate the noise intensity deviation using the formula: in, is the noise intensity deviation; Calculate the signal-to-noise deviation coefficient using the following formula: ;in, is the signal-to-noise deviation coefficient.

[0008] Preferably, the risks that users face during the authentication process are quantified, including: The delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient are comprehensively analyzed to establish a communication risk model and generate a communication risk assessment coefficient. The calculation formula of the communication risk assessment coefficient is: ;in, is the communication risk assessment coefficient, , , are the proportional coefficients of delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient, respectively. , , Both are greater than 0.

[0009] Preferably, the similar information includes: The user's behavioral similarity information is obtained through the user's browsing history and click records in the database after the security authentication is successful. The operation behavior characteristics of the user after entering the database are obtained. The operation behavior characteristics include click objects, browsing time, operation frequency, and access scope. A user behavior feature matrix is ​​generated to obtain the operation behavior characteristics of other normal users after entering the database. According to the click objects, browsing time, operation frequency, and access scope of other normal users, a behavior feature matrix of other normal users is generated. The similarity between the user behavior feature matrix and the other normal user behavior feature matrices is compared through similarity calculation, and the similarity results of the user behavior feature matrix and the other normal user behavior feature matrices are expressed by a similarity coefficient.

[0010] Preferably, the timeliness information includes: The user's timeliness information is represented by the timeliness coefficient. The logic for obtaining the timeliness coefficient is as follows: according to the user's authentication record, the time interval between each user's login is obtained, and the time interval between each user's login is marked as: , where k=1, 2, 3, ..., K, K is a positive integer, and k is the number of the time interval for each user login; Calculate the standard deviation of the time interval between each user login. The calculation formula is: ;in, is the standard deviation of the time interval between each user login, is the average time interval between each user login. ; Get the time interval between the user's last authentication login and the last authentication login time interval as: , calculate the aging coefficient, the calculation formula is: ;in, is the time efficiency coefficient.

[0011] Preferably, quantifying the risk level of user behavior information and determining the user's security trust score includes: The similarity coefficient and timeliness coefficient are comprehensively analyzed to establish a behavior risk model and generate a user behavior risk assessment coefficient. The calculation formula for the user behavior risk assessment coefficient is: ;in, is the user behavior risk assessment coefficient, is the similarity coefficient, , is the proportional coefficient of similarity coefficient and time efficiency coefficient, , Both are greater than 0.

[0012] The communication risk assessment coefficient and the user behavior risk assessment coefficient are calculated by weighted summation to generate a security trust score. The expression of the security trust score is: ;in, Score the security trust score, , is the ratio coefficient of communication risk assessment coefficient and user behavior risk assessment coefficient, , Both are less than 0.

[0013] Set a security trust score threshold. Each time a user passes authentication through a magnetic induction smart card, obtain the user's security trust score, and compare the user's security trust score with the security trust score threshold. If the security trust score is greater than the security trust score threshold, a security signal is generated. If the security trust score is less than the security trust score threshold, a warning signal is generated.

[0014] Preferably, a magnetic induction smart card secure access authentication system based on deep learning comprises a data acquisition module, an authentication module, a behavior analysis module, a risk assessment module and a scoring module, and the modules are signal-connected; The data collection module is used to collect the data generated during the authentication process between the smart card and the card reader, and collect the user's behavior data in the database; The authentication module is used to handle the authentication process between the smart card and the card reader, use the deep learning model to verify the transmitted signal data, generate a preliminary result of authentication success or failure, and collect relevant data for subsequent analysis; Behavior analysis module, used to record user operation behaviors in the database and extract user behavior characteristics; The risk assessment module is used to quantify the risk of the authentication process and the risk of user behavior through the communication risk model and the behavior risk model respectively; The scoring module is used to integrate the authentication process risks and behavioral risks to generate a user's security trust score, and compare it with the threshold to determine the user's security status.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention collects corresponding data from the process of user authentication and identification using a magnetic induction smart card, quantifies the risk level of the authentication and identification process, quantifies the risk level of user behavior from the perspective of user behavior, and comprehensively analyzes and determines the user's security score, which helps to obtain a more comprehensive security analysis and achieve comprehensive risk assessment and dynamic security management. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 It is a schematic diagram of the structure of the present invention; Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose and technical solution of the present invention clearly and completely described, and the advantages more clearly understood, the embodiments of the present invention are further described in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of the embodiments, and are only used to explain the embodiments of the present invention, and are not used to limit 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.

[0017] For example, see Figure 1 The present invention provides a magnetic induction smart card secure access authentication method based on deep learning, which specifically includes the following steps: S1: Process the signal data transmitted between the smart card and the card reader through a convolutional neural network, and collect the delay information, retransmission rate information, and noise information generated by the successfully authenticated user during the authentication process; S2: Quantify the risk of users in the authentication process based on the comprehensive processing of delay information, retransmission rate information and noise information; S3: Compare the user's behavior characteristics in the database after successful authentication with the behavior characteristics of other users in the database to determine the user's similar information, and determine the user's timeliness information based on the user's historical login behavior; S4: Comprehensively process the user's similarity information and timeliness information, quantify the risk level of the user's behavior information, and determine the user's security trust score based on the risks faced by the user during the authentication process.

[0018] The signal data transmitted between the smart card and the card reader is processed by a convolutional neural network (CNN). After verification, the user can access the database. The original data extracted from the electromagnetic signal between the smart card and the card reader includes frequency characteristics, signal strength, signal waveform, and noise characteristics.

[0019] After cleaning the data, key features related to identity authentication are extracted, such as time domain or frequency domain features (e.g. spectrum data after Fourier transform), and the signal data is normalized to adapt it to the input requirements of the CNN model to ensure the stability and accuracy of model training.

[0020] The CNN model is integrated with the existing database management system to ensure that users can access the database smoothly after the signal authentication is passed. After the user passes the signal authentication, the corresponding database access rights are granted according to the user's identity and permissions.

[0021] No further measures are taken for users who fail verification. For users who succeed in verification, that is, users who are granted corresponding database access rights, the data generated during the verification process is collected through the collaborative work of the card reader and the system.

[0022] During the process of users using smart cards for verification, the delay information, retransmission rate information and noise information are collected and processed to obtain the delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient. A communication risk model is established based on the delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient. Through the communication risk model, the risk level in the process of users using smart cards for verification is quantified.

[0023] Among them, the delay information generated when the user uses the smart card for verification refers to the delay between the card reader and the smart card. The communication between the magnetic induction smart card and the card reader is carried out through wireless signals. The smart card needs to process the reader's request and return a response. If a malicious attack is received during the verification process, the delay between the card reader and the smart card may increase.

[0024] The collected delay information is processed and represented by the delay fluctuation coefficient. The logic for obtaining the delay fluctuation coefficient is as follows: set the monitoring interval and the size of the sliding window, and mark the data in each sliding window as: , where n=1, 2, 3, ..., N, N is a positive integer, j=1, 2, 3, ..., J, J is a positive integer, n represents the number of the sliding window in the monitoring interval, and j represents the number of the data in the nth sliding window; It should be noted that the monitoring interval and sliding window are a specific time range. The monitoring interval usually contains multiple sliding windows. The monitoring interval provides a broad perspective to help identify long-term trends and periodic patterns. The sliding window helps detect immediate anomalies and changes. The monitoring interval and sliding window are set by professional staff. Usually the monitoring interval is from the placement of the smart card, through the recognition of the card reader, to the system completing the authentication and confirmation of the data. The sliding window is used to monitor and analyze each stage of the authentication process in real time.

[0025] Get the standard deviation of the data in each sliding window and mark the standard deviation of the data in each sliding window as: BZ, where , ; According to the normal operation data and historical records of the system, a reference threshold of the delay standard deviation is set, and the standard deviation calculated in each sliding window is compared with the preset reference threshold. If the standard deviation calculated in the sliding window exceeds the reference threshold, the sliding window is marked as an abnormal window, and the number of abnormal windows in the monitoring interval is obtained and marked as: , the delay fluctuation coefficient is calculated by the ratio of the number of abnormal windows in the monitoring interval to the number of sliding windows in the monitoring interval. The calculation formula of the delay fluctuation coefficient is: ,in, is the delay fluctuation coefficient.

[0026] It can be seen from the formula that the larger the delay fluctuation coefficient is, the more unstable the communication delay is during the authentication process. A large change in the communication delay may indicate that an abnormal situation has occurred during the authentication process. Conversely, the smaller the delay fluctuation coefficient is, the more stable the communication delay is during the authentication process.

[0027] The retransmission rate information generated during the user's use of smart cards for verification is a key indicator for measuring the communication quality during the smart card verification process. It reflects the frequency of data packet transmission failure and the need for retransmission due to various reasons (such as signal interference, network congestion, etc.). A high retransmission rate usually indicates unstable communication, which may affect the verification speed and success rate.

[0028] The collected retransmission rate information is processed and represented by a retransmission rate anomaly coefficient. The acquisition logic of the retransmission rate anomaly coefficient is as follows: obtaining the retransmitted data packets within the monitoring interval, and determining the retransmitted data packet sequence according to the transmission time sequence of the retransmitted data packets; The data is processed in a stable manner (such as differential, logarithmic transformation, etc.), and the retransmission data packet sequence is modeled using the autoregressive moving average model. The future retransmission rate is predicted using the autoregressive moving average model, and the order of the autoregressive moving average model is determined. The processed retransmission data packet sequence is fitted with the autoregressive moving average model. The expression of the autoregressive moving average model is: ;in, is the retransmission rate at time point t, p is the order of the autoregressive part, q is the order of the moving average part, is the error term at time point t, , ,……, is the autoregressive parameter, , ,……, is the moving average parameter; Use the fitted autoregressive moving average model to predict the retransmission rate at future time points, collect the actual retransmission rate value at future time points and the predicted retransmission rate value at future time points, and calculate the deviation between the predicted value and the actual value; According to the historical data and the performance of the model, the residual threshold is set, and the calculated residual is compared with the residual threshold. If the calculated residual is greater than the residual threshold, the time point is marked as an abnormal time point. If the calculated residual is less than the residual threshold, the time point is marked as a normal time point. The number of abnormal time points and normal time points is obtained, and the abnormal coefficient of the retransmission rate is calculated. The calculation formula is: ;in, is the retransmission rate anomaly coefficient, is the number of abnormal time points, is the number of normal time points.

[0029] It can be seen from the formula that the larger the retransmission rate anomaly coefficient is, the greater the deviation of the retransmission rate from the normal value during the authentication process is, indicating that there may be an abnormality in the authentication process.

[0030] The noise information generated when users use smart cards for authentication refers to various irrelevant or interfering signals in addition to useful signals during the communication between the smart card and the card reader.

[0031] The collected noise information is processed and represented by the signal-to-noise deviation coefficient. The logic for obtaining the signal-to-noise deviation coefficient is: obtain the signal strength and noise strength in the monitoring interval, and mark the signal strength and noise strength in the monitoring interval as: and , i=1, 2, 3, ..., I, I is a positive integer, i is the number of data collected within the monitoring interval; Set the signal strength threshold and the noise strength threshold, and mark the signal strength threshold and the noise strength threshold as: and ; Calculate the signal strength deviation using the formula: in, is the signal strength deviation; Calculate the noise intensity deviation using the formula: in, is the noise intensity deviation; Calculate the signal-to-noise deviation coefficient using the following formula: ;in, is the signal-to-noise deviation coefficient.

[0032] It can be seen from the formula that the larger the signal-to-noise deviation coefficient is, the better the current communication environment is, and the signal can effectively resist noise interference, which means that the authentication process is safer. Conversely, the smaller the signal-to-noise deviation coefficient is, the worse the current communication environment is, which means that the risk is higher during the authentication process.

[0033] The delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient are comprehensively analyzed to establish a communication risk model and generate a communication risk assessment coefficient. The calculation formula of the communication risk assessment coefficient is: ;in, is the communication risk assessment coefficient, , , are the proportional coefficients of delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient, respectively. , , Both are greater than 0.

[0034] It should be noted that the larger the communication risk assessment coefficient, the more risks or unstable factors may exist in the current communication environment. This means that even if the user succeeds in verification using a smart card, the entire verification process may experience more delays, retransmissions or signal interference. These unstable factors may cause the system to have less trust in the verification. If the communication risk assessment coefficients are high during multiple verification processes, the system may put the user on a monitoring list to prevent potential security threats.

[0035] Embodiment 2, based on embodiment 1, further improves the security of the system and the trust in users by combining the user's behavior information, and can identify potential security threats, wherein the user's behavior information includes behavior similarity information and timeliness information, analyzes and processes the user's behavior similarity information and timeliness information to obtain the similarity coefficient and timeliness coefficient, and establishes a behavior risk model to quantify the risks brought by the user's behavior.

[0036] Among them, the user's behavioral similarity information is obtained through the user's browsing history and click records in the database after the security authentication is successful, and the operation behavior characteristics of the user after entering the database are obtained. The operation behavior characteristics include click objects, browsing time, operation frequency, access scope, etc., and a user behavior feature matrix is ​​generated to obtain the operation behavior characteristics of other normal users after entering the database. According to the click objects, browsing time, operation frequency, access scope and other information of other normal users, a behavior feature matrix of other normal users is generated, and the similarity of the user behavior feature matrix and the other normal user behavior feature matrix is ​​compared by similarity calculation, and the similarity results of the user behavior feature matrix and the other normal user behavior feature matrix are expressed by a similarity coefficient.

[0037] The larger the similarity coefficient, the more similar the user's operation behavior is to that of other normal users, which means that the user's operation behavior is closer to the normal range. On the contrary, the smaller the similarity coefficient, the greater the difference between the user's operation behavior and that of other normal users, and there may be abnormalities or risks.

[0038] It should be noted that similarity calculation methods include Euclidean distance similarity calculation, cosine similarity calculation and other similarity calculation methods, which will not be described in detail here.

[0039] For example, after a user obtains access rights to a database, the user's operation behavior characteristics in the database are obtained, including clicked objects, browsing time, operation frequency, and access scope, and the user behavior feature matrix A is generated. , and obtain other normal user behavior feature matrix B, marked as: , using the cosine similarity calculation method, the similarity coefficient is marked as: XS; like , , then the similarity coefficient of the user is: 0.96, among which, is the similarity coefficient, and the similarity coefficient of this user is 0.96.

[0040] The user's timeliness information refers to the performance of the user's authentication behavior at the time level. The time of each user's login is determined based on the user's authentication record, and the user's timeliness information is represented by the timeliness coefficient. The logic for obtaining the timeliness coefficient is: according to the user's authentication record, the time interval between each user's login is obtained, and the time interval between each user's login is marked as: , where k=1, 2, 3, ..., K, K is a positive integer, and k is the number of the time interval for each user login; Calculate the standard deviation of the time interval between each user login. The calculation formula is: ;in, is the standard deviation of the time interval between each user login, is the average time interval between each user login. ; Get the time interval between the user's last authentication login and the last authentication login time interval as: , calculate the aging coefficient, the calculation formula is: ;in, is the time efficiency coefficient.

[0041] It can be seen from the formula that the larger the timeliness coefficient is, the lower the user's activity is, which means that the trust in the user may be lower. Conversely, the smaller the timeliness coefficient is, the higher the user's activity is, which means that the trust in the user may be higher.

[0042] The similarity coefficient and timeliness coefficient are comprehensively analyzed to establish a behavior risk model and generate a user behavior risk assessment coefficient. The calculation formula for the user behavior risk assessment coefficient is: ;in, is the user behavior risk assessment coefficient, , is the proportional coefficient of similarity coefficient and time efficiency coefficient, , Both are greater than 0.

[0043] The communication risk assessment coefficient obtained from the user authentication process and the user behavior risk assessment coefficient obtained from the user behavior information are weighted and summed to generate a security trust score. The expression of the security trust score is: ;in, Score the security trust score, , is the ratio coefficient of communication risk assessment coefficient and user behavior risk assessment coefficient, , Both are less than 0.

[0044] Set a security trust score threshold. Each time a user passes authentication through a magnetic induction smart card, obtain the user's security trust score and compare it with the security trust score threshold. If the security trust score is greater than the security trust score threshold, a security signal is generated, indicating that the user's login is relatively safe. If the security trust score is less than the security trust score threshold, a warning signal is generated, indicating that there are potential risks or abnormalities in the user's authentication process, which requires further review and marking of the user for further verification by staff.

[0045] The present invention collects corresponding data from the process of user authentication and identification using a magnetic induction smart card, quantifies the risk level of the authentication and identification process, quantifies the risk level of user behavior from the perspective of user behavior, and comprehensively analyzes and determines the user's security score, which helps to obtain a more comprehensive security analysis and achieve comprehensive risk assessment and dynamic security management.

[0046] For example 3, please refer to Figure 2 , the present invention provides a magnetic induction smart card security access authentication system based on deep learning, including a data acquisition module, an authentication module, a behavior analysis module, a risk assessment module and a scoring module, and the modules are signal connected; The data collection module is used to collect the data generated during the authentication process between the smart card and the card reader, and collect the user's behavior data in the database; The authentication module is used to handle the authentication process between the smart card and the card reader, use the deep learning model to verify the transmitted signal data, generate a preliminary result of authentication success or failure, and collect relevant data for subsequent analysis; Behavior analysis module, used to record user operation behaviors in the database and extract user behavior characteristics; The risk assessment module is used to quantify the risk of the authentication process and the risk of user behavior through the communication risk model and the behavior risk model respectively; The scoring module is used to integrate the authentication process risks and behavioral risks to generate a user's security trust score, and compare it with the threshold to determine the user's security status.

[0047] 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 magnetic induction smart card secure access authentication method based on deep learning, characterized in that: The following steps are involved: S1: Process the signal data transmitted between the smart card and the card reader through a convolutional neural network, and collect the delay information, retransmission rate information, and noise information generated by the successfully authenticated user during the authentication process; S2: Quantify the risk of users in the authentication process based on the comprehensive processing of delay information, retransmission rate information and noise information; S3: Compare the user's behavior characteristics in the database after successful authentication with the behavior characteristics of other users in the database to determine the user's similar information, and determine the user's timeliness information based on the user's historical login behavior; S4: Comprehensively process the user's similarity information and timeliness information, quantify the risk level of the user's behavior information, and determine the user's security trust score based on the risks faced by the user during the authentication process. The collected delay information is processed and represented by the delay fluctuation coefficient. The logic for obtaining the delay fluctuation coefficient is as follows: set the monitoring interval and the size of the sliding window, and mark the data in each sliding window as: , where n=1, 2, 3, ..., N, N is a positive integer, j=1, 2, 3, ..., J, J is a positive integer, n represents the number of the sliding window in the monitoring interval, and j represents the number of the data in the nth sliding window; Get the standard deviation of the data in each sliding window and mark the standard deviation of the data in each sliding window as: BZ, where , ; A reference threshold of the delay standard deviation is set, and the standard deviation calculated in each sliding window is compared with the preset reference threshold. If the standard deviation calculated in the sliding window exceeds the reference threshold, the sliding window is marked as an abnormal window, and the number of abnormal windows in the monitoring interval is obtained and marked as: , the delay fluctuation coefficient is calculated by the ratio of the number of abnormal windows in the monitoring interval to the number of sliding windows in the monitoring interval. The calculation formula of the delay fluctuation coefficient is: ,in, is the delay fluctuation coefficient. The collected retransmission rate information is processed and represented by a retransmission rate anomaly coefficient. The acquisition logic of the retransmission rate anomaly coefficient is as follows: obtaining the retransmitted data packets within the monitoring interval, and determining the retransmitted data packet sequence according to the transmission time sequence of the retransmitted data packets; The data is stabilized, and the retransmission data packet sequence is modeled using the autoregressive moving average model. The future retransmission rate is predicted using the autoregressive moving average model. The order of the autoregressive moving average model is determined, and the processed retransmission data packet sequence is fitted with the autoregressive moving average model. The expression of the autoregressive moving average model is: ;in, is the retransmission rate at time point t, p is the order of the autoregressive part, q is the order of the moving average part, is the error term at time point t, , ,……, is the autoregressive parameter, , ,……, is the moving average parameter; Use the fitted autoregressive moving average model to predict the retransmission rate at future time points, collect the actual retransmission rate value at future time points and the predicted retransmission rate value at future time points, and calculate the deviation between the predicted value and the actual value; Set the residual threshold, compare the calculated residual with the residual threshold, if the calculated residual is greater than the residual threshold, mark the time point as an abnormal time point, if the calculated residual is less than the residual threshold, mark the time point as a normal time point, obtain the number of abnormal time points and normal time points, calculate the retransmission rate abnormal coefficient, the calculation formula is: ;in, is the retransmission rate anomaly coefficient, is the number of abnormal time points, is the number of normal time points. The collected noise information is processed and represented by the signal-to-noise deviation coefficient. The logic for obtaining the signal-to-noise deviation coefficient is: obtain the signal strength and noise strength in the monitoring interval, and mark the signal strength and noise strength in the monitoring interval as: and , i=1, 2, 3, ..., I, I is a positive integer, i is the number of data collected within the monitoring interval; Set the signal strength threshold and the noise strength threshold, and mark the signal strength threshold and the noise strength threshold as: and ; Calculate the signal strength deviation using the formula: in, is the signal strength deviation; Calculate the noise intensity deviation using the formula: in, is the noise intensity deviation; Calculate the signal-to-noise deviation coefficient using the following formula: ;in, is the signal-to-noise deviation coefficient. The delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient are comprehensively analyzed to establish a communication risk model and generate a communication risk assessment coefficient. The calculation formula of the communication risk assessment coefficient is: ;in, is the communication risk assessment coefficient, , , are the proportional coefficients of delay fluctuation coefficient, retransmission rate anomaly coefficient and signal-to-noise deviation coefficient, respectively. , , Both are greater than 0. The user's behavioral similarity information is obtained through the user's browsing history and click records in the database after the security authentication is successful. The operation behavior characteristics of the user after entering the database are obtained. The operation behavior characteristics include click objects, browsing time, operation frequency, and access scope. A user behavior feature matrix is ​​generated to obtain the operation behavior characteristics of other normal users after entering the database. According to the click objects, browsing time, operation frequency, and access scope of other normal users, a behavior feature matrix of other normal users is generated. The similarity between the user behavior feature matrix and the other normal user behavior feature matrices is compared through similarity calculation, and the similarity results of the user behavior feature matrix and the other normal user behavior feature matrices are expressed by a similarity coefficient. The user's timeliness information is represented by the timeliness coefficient. The logic for obtaining the timeliness coefficient is as follows: according to the user's authentication record, the time interval between each user's login is obtained, and the time interval between each user's login is marked as: , where k=1, 2, 3, ..., K, K is a positive integer, and k is the number of the time interval for each user login; Calculate the standard deviation of the time interval between each user login. The calculation formula is: ;in, is the standard deviation of the time interval between each user login, is the average time interval between each user login. ; Get the time interval between the user's last authentication login and the last authentication login time interval as: , calculate the aging coefficient, the calculation formula is: ;in, is the time efficiency coefficient. The similarity coefficient and timeliness coefficient are comprehensively analyzed to establish a behavior risk model and generate a user behavior risk assessment coefficient. The calculation formula for the user behavior risk assessment coefficient is: ;in, is the user behavior risk assessment coefficient, is the similarity coefficient, , is the proportional coefficient of similarity coefficient and time efficiency coefficient, , Both are greater than 0. The communication risk assessment coefficient and the user behavior risk assessment coefficient are calculated by weighted summation to generate a security trust score. The expression of the security trust score is: ;in, Score the security trust score. , is the ratio coefficient of communication risk assessment coefficient and user behavior risk assessment coefficient, , Both are less than 0. Set a security trust score threshold. Each time a user passes authentication through a magnetic induction smart card, obtain the user's security trust score, and compare the user's security trust score with the security trust score threshold. If the security trust score is greater than the security trust score threshold, a security signal is generated. If the security trust score is less than the security trust score threshold, a warning signal is generated. The data collection module is used to collect the data generated during the authentication process between the smart card and the card reader, and collect the user's behavior data in the database; The authentication module is used to handle the authentication process between the smart card and the card reader, use the deep learning model to verify the transmitted signal data, generate a preliminary result of authentication success or failure, and collect relevant data for subsequent analysis; Behavior analysis module, used to record user operation behaviors in the database and extract user behavior characteristics; The risk assessment module is used to quantify the risk of the authentication process and the risk of user behavior through the communication risk model and the behavior risk model respectively; The scoring module is used to integrate the authentication process risks and behavioral risks to generate a user's security trust score, and compare it with the threshold to determine the user's security status.

Citation Information

Patent Citations

  • Smart card information processing method and device

    CN104573765A

  • Intelligent card and COS post-update security method thereof

    CN111190631A

  • Account abnormality detection method, device and system and storage medium

    CN112714093A

  • Outpatient service information query reservation management system based on big data

    CN116644825A

  • Mining vertical shaft cage emergency communication management system based on data analysis

    CN116927885A

Cited By

  • Method, device and system for carrying out trust degree evaluation on side end equipment node

    CN119155126A

  • Distribution network area intelligent equipment authentication method and configuration terminal

    CN119580682A

  • Industrial data acquisition and analysis system and method

    CN120276398A