A card swiping early warning method and system for an intelligent terminal transaction environment

Abstract

The application provides a card swiping early warning method and system for an intelligent terminal transaction environment, multi-source information of a transaction scene in which the intelligent terminal is located is input into a pre-trained risk identification model to obtain a preliminary risk score of the intelligent terminal transaction environment; when the preliminary risk score is greater than a risk threshold of the intelligent terminal, a sliding time window is used to continuously track a waveform change rate of a radio frequency signal reflection waveform of a specified card swiping period in the intelligent terminal, if the waveform change rate is successively increased and all exceed a baseline deviation threshold of the intelligent terminal, a card swiping risk score of the intelligent terminal transaction environment is determined through the preliminary risk score of each card swiping period and all the waveform change rates; when the card swiping risk score is greater than an alarm threshold of the intelligent terminal, a hierarchical early warning prompt of the intelligent terminal is triggered, and the number of continuous responses of a non-contact card swiping function is limited. Based on the above scheme, joint risk discrimination of the intelligent terminal transaction environment can be realized based on multi-source heterogeneous data.

Description

Technical Field

[0001] This application relates to the field of terminal transaction security technology, and more specifically, to a card swipe early warning method and system for smart terminal transaction environments. Background Technology

[0002] Smart terminal card swiping refers to the process where a user brings a contactless card or a mobile device with payment functionality close to the built-in radio frequency card reader area of ​​a smart terminal, and the terminal activates the card and completes data exchange by emitting radio frequency signals.

[0003] Current card-swiping risk detection solutions primarily rely on changes in the amplitude of radio frequency signals received by the card reader to determine the presence of anomalies. A few solutions may additionally incorporate gyroscopes or ambient sound sensors as auxiliary methods. However, data from each sensor is analyzed separately in independent channels without cross-modal fusion analysis. Logically, existing technologies generally employ a single-swipe independent decision-making model, outputting a risk assessment result immediately after each swipe. This assessment does not consider historical data from previous swipe cycles of the same terminal, nor does it track parameter change trends during multiple consecutive swipes, failing to effectively distinguish between occasional environmental interference and persistent attacks. Therefore, how to achieve joint risk assessment of smart terminal transaction environments based on multi-source heterogeneous data, thereby improving the anti-interference capability of card-swiping environment anomaly detection, has become a challenge for the industry. Summary of the Invention

[0004] This application provides a card swiping early warning method and system for smart terminal transaction environments, which can realize joint risk judgment of smart terminal transaction environments based on multi-source heterogeneous data, thereby improving the anti-interference capability of card swiping environment anomaly detection.

[0005] Firstly, this application provides a method for card swipe warning in a smart terminal transaction environment, including: The system collects environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and radio frequency signal reflection waveform during the card swiping process in the transaction scenario of the smart terminal, forming a multi-source information set. The multi-source information set is input into a pre-trained risk identification model, and the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation are extracted respectively, so as to obtain a preliminary risk score of the smart terminal transaction environment. When the preliminary risk score is greater than the risk threshold of the smart terminal, the waveform change rate of the radio frequency signal reflection waveform in the smart terminal during a specified card swiping cycle is continuously tracked by a sliding time window. If the waveform change rate increases successively and exceeds the baseline deviation threshold of the smart terminal, the card swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card swiping cycle and all waveform change rates. When the card swiping risk score exceeds the alarm threshold of the smart terminal, the smart terminal will trigger a graded warning prompt and limit the number of consecutive responses of the contactless card swiping function.

[0006] In some embodiments, the multi-source information set is input into a pre-trained risk identification model to extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation, thereby obtaining a preliminary risk score for the smart terminal transaction environment. Specifically, this includes: A pre-trained risk identification model is used to extract the joint anomaly index of the scene and the device from the environmental noise spectrum of the multi-source information set; A pre-trained risk identification model is used to extract temporal inconsistency scores of networks and behaviors from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequences of the multi-source information set; The radio frequency reflection feature deviation of the card swiping electromagnetic signal is extracted from the radio frequency signal reflection waveform of the multi-source information set using a pre-trained risk identification model. The joint anomaly index, the timing inconsistency score, and the radio frequency reflection characteristic deviation are integrated into a preliminary risk score for the smart terminal transaction environment.

[0007] In some embodiments, extracting the joint anomaly index of the scene and the device from the environmental noise spectrum of the multi-source information set using a pre-trained risk identification model specifically includes: The environmental noise spectrum of the multi-source information set is segmented according to time windows, and the power spectral density and time-frequency distribution characteristics of each time window are extracted. The pre-trained risk identification model is used to learn the noise spectrum patterns of different smart terminal devices in normal transaction scenarios, and a joint distribution feature space of devices and scenarios is established. The power spectral density and the time-frequency distribution features are mapped to the joint distribution feature space, and the probability density of the joint distribution space with the normal distribution center is calculated to obtain the joint anomaly index of the scene and the device.

[0008] In some embodiments, extracting temporal inconsistency scores of networks and behaviors from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequences of the multi-source information set using a pre-trained risk identification model specifically includes: Align the gyroscope attitude angular velocity and wireless network beacon fluctuation sequence of the multi-source information set with the same time axis to obtain dual-mode time series data; Extract the temporal correlation patterns between posture changes and beacon fluctuations in normal trading from a pre-trained risk identification model; The temporal correlation model is used to calculate the mutual information, temporal misalignment, and prediction error between the current bimodal sequences in the bimodal temporal data to obtain the temporal inconsistency score of the network and behavior.

[0009] In some embodiments, extracting the radio frequency reflection feature deviation of the card-swiping electromagnetic signal from the radio frequency signal reflection waveform of the multi-source information set using a pre-trained risk identification model specifically includes: Extract the effective reflection segment of the card swiping action from the radio frequency signal reflection waveform of the multi-source information set; The effective reflection segment is mapped to a high-dimensional feature space using a pre-trained risk identification model, and electromagnetic features including reflection intensity attenuation, phase change, and multipath distribution are extracted. The electromagnetic features are compared with those of a normal card swipe to obtain the radio frequency reflection feature deviation of the card swipe electromagnetic field.

[0010] In some embodiments, determining the card-swiping risk score of the smart terminal transaction environment through the preliminary risk score of each swipe cycle and all waveform change rates specifically includes: Collect the preliminary risk score and the waveform change rate of the corresponding radio frequency signal reflection waveform for each card swipe cycle; Risk weights for each card swiping cycle are assessed based on the rate of change of each waveform to obtain the risk weight for each card swiping cycle. By using various risk weights, the initial risk scores for each card-swiping cycle are fused and evaluated according to time sequence and weighted waveform change rate to obtain the card-swiping risk score for the smart terminal transaction environment.

[0011] In some embodiments, the baseline deviation threshold is a critical discrimination value that distinguishes between normal card swipe fluctuations and abnormal incremental fluctuations.

[0012] Secondly, this application provides a card-swiping early warning system for smart terminal transaction environments, including a hierarchical early warning unit, wherein the hierarchical early warning unit includes: The data acquisition module is used to collect the environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and radio frequency signal reflection waveform during the card swiping process of the smart terminal in the transaction scenario, forming a multi-source information set; The processing module is used to input the multi-source information set into a pre-trained risk identification model, extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation, and then obtain a preliminary risk score of the smart terminal transaction environment. The processing module is also used to continuously track the waveform change rate of the radio frequency signal reflection waveform in the smart terminal for a specified card swiping cycle using a sliding time window when the preliminary risk score is greater than the risk threshold of the smart terminal. If the waveform change rate increases successively and exceeds the baseline deviation threshold of the smart terminal, the card swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card swiping cycle and all waveform change rates. The execution module is used to trigger a graded early warning prompt on the smart terminal and limit the number of consecutive responses of the contactless card swiping function when the card swiping risk score is greater than the alarm threshold of the smart terminal.

[0013] Thirdly, this application provides a computer device, which includes a memory and a processor. The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the above-described card swipe warning method for a smart terminal transaction environment.

[0014] Fourthly, this application provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the aforementioned card-swiping warning method for a smart terminal transaction environment.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: This application provides a card-swiping early warning method and system for smart terminal transaction environments. The method collects environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and radio frequency signal reflection waveform during the card-swiping process from the transaction scene where the smart terminal is located, forming a multi-source information set. This multi-source information set is input into a pre-trained risk identification model to extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection characteristic deviation, thereby obtaining a preliminary risk score for the smart terminal transaction environment. When the preliminary risk score is greater than the risk threshold of the smart terminal, the waveform change rate of the radio frequency signal reflection waveform in a specified card-swiping cycle is continuously tracked using a sliding time window. If the waveform change rate increases successively and all exceed the baseline deviation threshold of the smart terminal, the card-swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card-swiping cycle and all waveform change rates. When the card-swiping risk score is greater than the alarm threshold of the smart terminal, a graded early warning prompt is triggered on the smart terminal, and the number of consecutive responses of the contactless card-swiping function is limited.

[0016] Therefore, in this application, when the card-swiping risk score exceeds the alarm threshold of the smart terminal, a graded early warning prompt is triggered on the smart terminal, and the number of consecutive responses of the contactless card-swiping function is limited. First, by determining the preliminary risk score, the joint anomaly representation of multi-source heterogeneous data within a single card-swiping cycle can be obtained. This allows for the fusion of complementary information from four dimensions—ambient acoustics, device motion, network fluctuations, and radio frequency electromagnetics—within a single time segment. The ambient noise spectrum reflects the physical spatial attributes of the transaction scenario, the gyroscope attitude angular velocity characterizes the terminal's operational behavior, the wireless network beacon fluctuation sequence reveals the stability of the surrounding signal environment, and the radio frequency signal reflection waveform directly responds to the electromagnetic characteristics of the card-swiping object. After inputting the four heterogeneous data into the pre-trained model, the model extracts the degree of anomaly deviation in each dimension through joint distribution feature space, temporal correlation pattern, and electromagnetic feature template, and then weighted fusion to form a comprehensive score. This score can effectively suppress misjudgments caused by single-dimensional noise or occasional interference, thereby improving the anti-interference capability of preliminary anomaly detection. Then, by determining the card swiping risk score, a quantitative measure of the risk evolution trend over multiple consecutive card swiping cycles can be obtained. This allows for further filtering of transient fluctuations and identification of persistent abnormal patterns with malicious intent through temporal consistency checks. Data in any dimension of a single card swiping cycle may fluctuate within a normal range due to differences in user operating habits or sudden environmental changes, causing the initial risk score to occasionally exceed the risk threshold. A sliding time window is used to track the progressively increasing characteristics of the waveform change rate. If the change rate of the radio frequency reflection waveform in multiple consecutive card swiping cycles exceeds the baseline deviation threshold and shows a monotonically increasing trend, it indicates that the abnormal behavior is repetitive and deteriorating, rather than an isolated noise event. This approach has a natural ability to suppress single-point-of-flight interference, thus effectively eliminating non-steady-state noise and occasional interference in the time dimension, enhancing the robustness of anomaly detection in long-term transaction environments. In summary, based on the above scheme, joint risk discrimination of the smart terminal transaction environment can be achieved based on multi-source heterogeneous data, thereby improving the anti-interference capability of anomaly detection in the card swiping environment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an exemplary flowchart of a card swiping warning method for a smart terminal transaction environment according to some embodiments of this application; Figure 2 This is a schematic diagram of a smart terminal card swiping transaction process according to some embodiments of this application; Figure 3 This is a schematic diagram of the process for determining a credit card risk score according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a hierarchical early warning unit according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing a card swiping warning method in a smart terminal transaction environment, according to some embodiments of this application. Detailed Implementation

[0019] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] refer to Figure 1 The figure is an exemplary flowchart of a card swipe warning method for a smart terminal transaction environment according to some embodiments of this application. The card swipe warning method for a smart terminal transaction environment mainly includes the following steps: In step 101, the environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and radio frequency signal reflection waveform during the card swiping process of the smart terminal in the transaction scenario are collected to form a multi-source information set.

[0021] It should be noted that, in this application, the environmental noise spectrum is a dataset characterizing the energy distribution of sound signals at different frequencies in the transaction scenario where the smart terminal is located; the gyroscope attitude angular velocity is a sequence of physical quantities used to describe the speed of rotation of the smart terminal around various axes in space during the card swiping process; the wireless network beacon fluctuation sequence is an ordered array used to record the changes in the signal strength of surrounding wireless access points received by the smart terminal over time; the radio frequency signal reflection waveform is waveform data reflecting the change in amplitude of the signal returned after the radio frequency signal emitted by the smart terminal encounters the target object during card swiping over time; and the multi-source information set is an input dataset used to fuse data from different types of sensors to comprehensively judge the risk of the transaction environment.

[0022] In practice, firstly, as the smart terminal initiates a card transaction, the ambient sound acquisition function is activated. The built-in microphone continuously records the sounds of the transaction scene. This sound signal is processed in frames at fixed time intervals. A Fast Fourier Transform is performed on each frame to obtain the energy value of each frequency component. All energy values ​​corresponding to all frequencies are arranged from low to high frequency, forming a curve showing the energy change with frequency; this curve represents the ambient noise spectrum. Secondly, the gyroscope sensor in the smart terminal is activated simultaneously to continuously read the terminal's rotational angular velocity in the horizontal, vertical, and forward / backward directions, corresponding to the speed of rotation around the three spatial axes. Each reading yields three values, which are arranged chronologically to form three sequences that change over time. The set of all sequences is taken as the gyroscope's attitude angular velocity. Thirdly, the smart terminal's wireless network scanning function is activated simultaneously to periodically detect nearby connectable wireless access points. For each detected access point, its physical address and received signal strength are recorded. The signal strength of the same access point is arranged in chronological order of detection time to form a signal strength fluctuation sequence for that access point. If multiple access points exist, the fluctuation sequences of multiple access points are aligned by time and merged to form a wireless network beacon fluctuation sequence. Then, the RF card reader module of the smart terminal is activated simultaneously. During the card swiping process, the amplitude of the RF signal emitted by the transmitting antenna after reflection by the card-swiping object and back to the receiving antenna is continuously collected. The relationship between this amplitude and time is recorded to form a waveform curve with amplitude fluctuation over time. This curve is the RF signal reflection waveform. Finally, the four data sets collected in the above steps—environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and RF signal reflection waveform—are aligned according to the timestamp of the same transaction event and merged and stored into a complete data packet. This data packet is the multi-source information set.

[0023] It should be noted that in this application, Figure 2The flowchart illustrates the process of a smart terminal card transaction. The cardholder initiates a transaction request to the merchant and smart terminal by swiping their card (Step 1). The merchant terminal uploads the transaction notification to the acquiring bank (Step 2). The acquiring bank then uploads the transaction notification to the clearing institution (Step 3). The clearing institution forwards the transaction notification to the issuing bank (Step 4). The issuing bank sends a deduction notification to the cardholder and returns the deduction result to the clearing institution (Step 5). The clearing institution updates and replies to the acquiring bank with the transaction data (Step 6). The acquiring bank then updates and replies to the merchant terminal (Step 7). The merchant terminal prints a transaction receipt to the cardholder to complete the online authorization of the transaction (Step 8). At the end of the day, the clearing institution completes the transaction information clearing and initiates a fund clearing instruction through the central bank system (Step 9). The central bank system completes the interbank fund clearing for the issuing bank (Step 10) and transfers the cleared funds to the acquiring bank (Step 11). Finally, the acquiring bank and the merchant complete the fund transfer in their settlement accounts (generally T+1). (Step 12) The entire process realizes the separation of information flow to complete transaction authorization and data interaction, and fund flow to complete cross-institutional fund settlement through the central bank clearing system at the end of the day.

[0024] In step 102, the multi-source information set is input into the pre-trained risk identification model to extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation, thereby obtaining a preliminary risk score of the smart terminal transaction environment.

[0025] In some embodiments, the multi-source information set is input into a pre-trained risk identification model to extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation, thereby obtaining a preliminary risk score of the smart terminal transaction environment. This can be achieved through the following steps: A pre-trained risk identification model is used to extract the joint anomaly index of the scene and the device from the environmental noise spectrum of the multi-source information set; A pre-trained risk identification model is used to extract temporal inconsistency scores of networks and behaviors from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequences of the multi-source information set; The radio frequency reflection feature deviation of the card swiping electromagnetic signal is extracted from the radio frequency signal reflection waveform of the multi-source information set using a pre-trained risk identification model. The joint anomaly index, the timing inconsistency score, and the radio frequency reflection characteristic deviation are integrated into a preliminary risk score for the smart terminal transaction environment.

[0026] It should be noted that, in this application, the joint anomaly index is a risk indicator that quantifies the degree to which the combination of scene and device in the current transaction environment deviates from the normal noise pattern; the timing inconsistency score is a comprehensive indicator that quantifies the degree to which the timing coordination between the smart terminal's posture behavior and the wireless network environment deviates from the normal in the current transaction; the radio frequency reflection characteristic deviation is a numerical indicator that quantifies the degree of difference between the electromagnetic characteristics of the current card swiping process and the normal electromagnetic characteristics; and the preliminary risk score is a single numerical indicator that measures the overall card swiping risk level faced by the smart terminal in the current transaction scenario. The higher the preliminary risk score, the more abnormal the current transaction environment is, and the greater the possibility of card swiping fraud or illegal attacks.

[0027] In specific implementation, firstly, a pre-trained risk identification model is used to extract the joint anomaly index of the scene and device from the environmental noise spectrum of the multi-source information set; secondly, the pre-trained risk identification model is used to extract the temporal inconsistency score of the network and behavior from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequence of the multi-source information set; then, the pre-trained risk identification model is used to extract the radio frequency reflection characteristic deviation of the card swiping electromagnetic signal from the radio frequency signal reflection waveform of the multi-source information set; finally, the joint anomaly index, temporal inconsistency score, and radio frequency reflection characteristic deviation are normalized respectively. The parameters are mapped to a uniform value range of zero to one to eliminate the influence of different dimensions and value ranges. After normalization, the fusion weight coefficient of each value is obtained from the parameter library of the smart terminal. The magnitude of the weight coefficient is updated in the parameter library at regular intervals according to the contribution of that dimension to risk judgment in historical data. The sum of the three weight coefficients is one. The normalized joint anomaly index is multiplied by its weight, the normalized temporal inconsistency score is multiplied by its weight, and the normalized radio frequency reflection characteristic deviation is multiplied by its weight. The result of the sum of the three products is used as the preliminary risk score.

[0028] Preferably, in the above embodiments, extracting the joint anomaly index of the scene and the device from the environmental noise spectrum of the multi-source information set using a pre-trained risk identification model can be achieved through the following steps: The environmental noise spectrum of the multi-source information set is segmented according to time windows, and the power spectral density and time-frequency distribution characteristics of each time window are extracted. The pre-trained risk identification model is used to learn the noise spectrum patterns of different smart terminal devices in normal transaction scenarios, and a joint distribution feature space of devices and scenarios is established. The power spectral density and the time-frequency distribution features are mapped to the joint distribution feature space, and the probability density of the joint distribution space with the normal distribution center is calculated to obtain the joint anomaly index of the scene and the device.

[0029] It should be noted that, in this application, power spectral density is a physical quantity that measures the amount of signal power contained per unit frequency in the ambient noise spectrum; time-frequency distribution characteristics are a set of values ​​used to describe the energy distribution variation law of the ambient noise spectrum at different times and frequencies; joint distribution feature space is a multi-dimensional mathematical space used to represent the correspondence between the noise spectrum patterns of different smart terminal devices and their normal transaction scenarios; normal distribution center is a position vector used to represent the geometric center of the clustering region of all normal transaction samples in the joint distribution feature space; probability density is a value that measures the degree of closeness between the current ambient noise spectrum characteristics and the normal distribution center in the joint distribution feature space.

[0030] In practice, firstly, the environmental noise spectrum is extracted from the multi-source information set. This environmental noise spectrum is then continuously segmented according to fixed-length time windows, with some overlap between adjacent time windows. For the noise signal within each time window, a Fast Fourier Transform (FFT) is used to convert it from a time waveform to a frequency distribution. The square of the signal amplitude at each frequency point is calculated and then divided by the total duration of the signal to obtain the relationship between the power at that frequency and the frequency. This relationship is the power spectral density. The power spectral densities calculated for each time window are arranged in chronological order to form a two-dimensional grid consisting of a time axis and a frequency axis. Each... Each grid cell represents the power value at a specific frequency at a given moment. The entire two-dimensional grid represents the time-frequency distribution feature. For example, if a one-second noise signal is divided into ten time windows, and the power values ​​for each window are calculated at one hundred frequency points from low to high frequencies, a table of ten rows and one hundred columns is obtained. This table represents the time-frequency distribution feature; thus, the power spectral density and time-frequency distribution feature can be obtained. Then, environmental noise spectra collected from different smart terminal devices under normal trading scenarios within a specified time period (defaulting to the most recent month) are collected. The power spectral density and time-frequency distribution feature of each normal sample are extracted using the method described in the first step. A pre-trained [system / framework] is then constructed. The risk identification model contains a feature extraction network that maps the features of input normal samples to a high-dimensional mathematical space. Each dimension of this space represents a latent pattern resulting from the combination of device model and scene type. Within this high-dimensional space, feature points of all normal transaction samples cluster in a specific region, and the center vector of this region is labeled as the normal distribution center. The high-dimensional space itself, along with the defined normal distribution centers, constitutes a joint distribution feature space. Finally, the power spectral density and time-frequency distribution features of the current transaction are input into the pre-trained risk identification model, which uses the same... The feature extraction network maps the current feature to a joint distribution feature space, thus obtaining a corresponding location point. In this joint distribution feature space, the Mahalanobis distance between the current feature point and the center of the normal distribution is calculated. This Mahalanobis distance is then substituted into the probability density function formula of the Gaussian distribution. This formula uses the natural constant as the base and half the square of the negative Mahalanobis distance as the exponent for exponential operation. The result is then divided by the product of the square root of twice pi and the determinant of the covariance matrix to obtain the probability density of the current feature point. The smaller the probability density value, the farther the current feature point deviates from the center of the normal distribution. The reciprocal of the probability density is used as the joint anomaly index of the scene and the device.

[0031] Preferably, in the above embodiments, the extraction of network and behavior temporal inconsistency scores from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequences of the multi-source information set using a pre-trained risk identification model can be achieved through the following steps: Align the gyroscope attitude angular velocity and wireless network beacon fluctuation sequence of the multi-source information set with the same time axis to obtain dual-mode time series data; Extract the temporal correlation patterns between posture changes and beacon fluctuations in normal trading from a pre-trained risk identification model; The temporal correlation model is used to calculate the mutual information, temporal misalignment, and prediction error between the current bimodal sequences in the bimodal temporal data to obtain the temporal inconsistency score of the network and behavior.

[0032] It should be noted that, in this application, the dual-modal time series data refers to two synchronized time series data formed by aligning the gyroscope attitude angular velocity and the wireless network beacon fluctuation sequence on the same time coordinate; the time series correlation mode is a mathematical structure used to describe the correlation between the smart terminal attitude change and the wireless network beacon fluctuation over time during normal transactions; mutual information is a numerical value that measures the amount of information shared between the gyroscope attitude angular velocity sequence and the wireless network beacon fluctuation sequence; the time series misalignment is a numerical value that characterizes the degree of relative offset between the two time series signals on the time axis; and the prediction error is a numerical value that reflects the magnitude of the deviation generated when predicting the current value of another signal based on the historical value of one signal.

[0033] In practice, firstly, the gyroscope attitude angular velocity and the wireless network beacon fluctuation sequence are extracted from the multi-source information set. The gyroscope attitude angular velocity is a set of values ​​recorded in chronological order, with each value corresponding to the speed of rotation of the smart terminal around a certain axis at a sampling moment. The wireless network beacon fluctuation sequence is another set of values ​​recorded in chronological order, with each value corresponding to the signal strength received from a certain wireless access point at a sampling moment. The start times and sampling intervals of the two sets of data may be different. To align them, a common start time point is selected, usually the moment when the card swiping action begins as time zero. The gyroscope attitude angular velocity sequence is then interpolated and aligned according to the sampling moments of the wireless network beacon fluctuation sequence. The interpolation method is to connect two adjacent gyroscope sampling points with a straight line and calculate the gyroscope value corresponding to each wireless network beacon sampling moment. After alignment, each moment simultaneously possesses a gyroscope attitude angular velocity value and a wireless network beacon fluctuation value. Numerical data is generated by arranging paired values ​​in chronological order as bimodal time-series data. Then, bimodal time-series data samples collected within a specified time period (defaulting to the most recent month) under normal transaction scenarios are gathered. Each sample corresponds to a gyroscope attitude change and wireless network beacon fluctuation during a normal card swipe process. All samples are input into a pre-trained risk identification model. This model contains a time-series correlation learning module. This module uses a sliding window approach to divide each sample into multiple segments according to a fixed-length time window. For each segment, the model learns the mapping relationship from the gyroscope attitude angular velocity sequence to the wireless network beacon fluctuation sequence, and vice versa. Specifically, the model learns that in normal transactions, when the gyroscope attitude angular velocity exhibits a certain upward or downward pattern, the wireless network beacon fluctuation will typically show a corresponding enhancement or weakening pattern after a certain delay.These mapping relationships and the accompanying time delay range together constitute the time-series correlation pattern, which is stored in the pre-trained risk identification model in the form of a set of mathematical parameters. Finally, the bimodal time-series data of the current transaction is input into the pre-trained risk identification model, and the time-series correlation pattern is used to evaluate the data. The evaluation process consists of three parallel computations. The first computation is mutual information, which involves dividing the gyroscope attitude angular velocity sequence and the wireless network beacon fluctuation sequence into several value intervals, counting the frequency of the two sequences falling into each interval combination, calculating the joint entropy and the marginal entropy of the two sequences based on these frequencies, and subtracting the sum of the two marginal entropies from the joint entropy to obtain the mutual information value. The larger the mutual information, the stronger the correlation between the two. The second computation is the time-series misalignment, which uses the cross-correlation analysis method to calculate the time-series misalignment. The beacon wave sequence of the wireless network is gradually shifted to the left or right relative to the gyroscope attitude angular velocity sequence. After each shift, the correlation coefficient between the two sequences is calculated, and the shift that maximizes the correlation coefficient is recorded. This shift is the temporal misalignment, representing the average delay or advance time of the beacon wave relative to the attitude change. The third calculation is the prediction error, which uses the mapping parameters in the temporal correlation model to predict the beacon wave value of the current wireless network based on the gyroscope attitude angular velocity values ​​several times prior to the current time. The absolute difference between the predicted value and the actual acquired value is then calculated; this difference is the prediction error. The mutual information, temporal misalignment, and prediction error are normalized to fall between zero and one, and then weighted and summed according to preset weights to obtain the temporal inconsistency score of the network and behavior. A higher score is obtained by lower mutual information, a larger absolute value of temporal misalignment, and a larger prediction error.

[0034] Preferably, in the above embodiments, the extraction of the radio frequency reflection characteristic deviation of the card-swiping electromagnetic signal from the radio frequency signal reflection waveform of the multi-source information set using a pre-trained risk identification model can be achieved by the following steps: Extract the effective reflection segment of the card swiping action from the radio frequency signal reflection waveform of the multi-source information set; The effective reflection segment is mapped to a high-dimensional feature space using a pre-trained risk identification model, and electromagnetic features including reflection intensity attenuation, phase change, and multipath distribution are extracted. The electromagnetic features are compared with those of a normal card swipe to obtain the radio frequency reflection feature deviation of the card swipe electromagnetic field.

[0035] It should be noted that, in this application, the effective reflection segment is a data interval used to separate the waveform segment that actually corresponds to the card swiping action from the complete radio frequency signal reflection waveform; the high-dimensional feature space is a mathematical space used to expand and map the electromagnetic information in the original radio frequency reflection waveform to a combination of multiple dimensions; the reflection intensity attenuation is a numerical value that measures the degree to which the amplitude of the returned signal after encountering the card-swiping object decreases relative to the amplitude of the transmitted signal; the phase change is a numerical value used to describe the magnitude of the shift in the starting angle of the radio frequency signal during reflection; the multipath distribution is a set of numerical values ​​reflecting the distribution of the signal strength of the radio frequency signal reaching the receiving antenna through multiple different paths on different propagation paths; the electromagnetic feature is a numerical vector characterizing the electromagnetic response characteristics generated by the interaction between the radio frequency signal and the card-swiping object during the card swiping process; and the electromagnetic feature template is a set of electromagnetic feature standards used as a reference for judging whether the current card swiping is normal.

[0036] In practice, firstly, the radio frequency signal reflection waveform is extracted from the multi-source information set. This waveform is a complete amplitude-time curve, containing signals from the three stages of before, during, and after the card swipe. To extract the waveform that truly corresponds to the card swipe action, the start and end times of the action need to be detected. The method for detecting the start time is as follows: calculate the short-time energy of the waveform amplitude. The short-time energy is calculated by taking a fixed-length time window, squaring the amplitude value at each moment within the window, and summing the results to obtain an energy value. This energy value is then calculated by sliding it along the time axis. When the energy value suddenly rises from a stable state and exceeds a preset energy threshold, that moment is marked as the start of the card swipe. The method for detecting the end time is as follows: starting from the card swipe start time, continue sliding backward to calculate the short-term energy. When the energy value drops from a high level and remains below a preset energy threshold, this time is marked as the card swipe end time. The waveform segment between the start time and the end time is the effective reflection segment. Then, the effective reflection segment is input into a pre-trained risk identification model. This model contains a feature encoder, which maps the time series data of the effective reflection segment to a high-dimensional feature space. In the high-dimensional feature space, every subtle change in the original waveform is expanded and expressed in different dimensions. The specific extracted electromagnetic features include three aspects; the first is the reflection intensity. The first characteristic is attenuation, obtained by calculating the ratio between the maximum amplitude of the signal in the effective reflection segment and the original amplitude of the transmitted signal; a smaller ratio indicates more severe attenuation. The second characteristic is phase change, obtained by analyzing the changes in the zero-crossing intervals of the signal in the effective reflection segment. Specifically, this involves recording the moment the waveform changes from a negative to a positive value or from a positive to a negative value, calculating the time interval between adjacent zero-crossings, and using the difference between this event interval and the reference interval of the transmitted signal as the phase offset. The third characteristic is multipath distribution, obtained by performing autocorrelation analysis on the effective reflection segment. Autocorrelation analysis involves multiplying the effective reflection segment point-by-point with its replicas at different time delays and then calculating the result. The autocorrelation function is obtained, and the multiple peaks in this function reflect the time difference of the signal after reflection through different length paths. The positions and heights of these peaks are combined to form the multipath distribution. The three values ​​of reflection intensity attenuation, phase change and multipath distribution are concatenated into a numerical vector in a fixed order. This vector is the electromagnetic feature. Finally, the electromagnetic feature template of normal card swiping is read from the pre-trained risk identification model. This template is a standard vector obtained by collecting effective reflection segments in normal card swiping processes within a specified time period (default is the most recent month) and extracting electromagnetic features according to the method in step two. The average or median of these features is then taken.The electromagnetic characteristics of the current transaction are compared dimension-by-dimensionally with the electromagnetic characteristic template of a normal card swipe. The comparison method involves calculating the weighted Euclidean distance between the two vectors. Specifically, the value of each dimension in the electromagnetic characteristic is subtracted from the corresponding value in the normal template to obtain the difference. This difference is squared and multiplied by the weight coefficient corresponding to that dimension. The results of summing the weighted squares of all dimensions are then calculated, and the square root of the sum is taken to obtain the weighted Euclidean distance. The weight coefficients are pre-set based on the importance of each dimension in distinguishing between normal and abnormal card swipes. This weighted Euclidean distance can be used as the deviation of the radio frequency reflection characteristics of the card swipe electromagnetic signal.

[0037] In step 103, when the preliminary risk score is greater than the risk threshold of the smart terminal, the waveform change rate of the radio frequency signal reflection waveform in the smart terminal during a specified card swiping cycle is continuously tracked using a sliding time window. If the waveform change rate increases successively and all exceed the baseline deviation threshold of the smart terminal, the card swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card swiping cycle and all waveform change rates.

[0038] It should be noted that, in this application, the baseline deviation threshold is a critical discrimination value for distinguishing between normal card swiping fluctuations and abnormal incremental fluctuations. Based on the monitoring of the dynamic change trend of the radio frequency reflection waveform during multiple consecutive card swiping processes, it is determined whether there is any risk escalation behavior. In normal transactions, the radio frequency reflection waveform of the same terminal swiping multiple times in a short period of time should remain relatively stable, and the waveform change rate should not show a systematic incremental increase. However, malicious attacks or illegal card copying operations often lead to increasingly larger changes in the reflection waveform with each successive swipe.

[0039] In practical implementation, firstly, the risk threshold of the smart terminal can be obtained from the parameter library of the smart terminal. When the initial risk score exceeds the risk threshold of the smart terminal, a sliding time window of fixed length is started. This window moves forward step by step on the time axis. After each move, the window contains several recent specified card swiping cycles. For each card swipe within the window, the amplitude sequence of the radio frequency signal reflection waveform during that card swipe is extracted. The absolute value of the difference between the amplitudes of two adjacent sampling points is calculated and summed, then divided by the total duration of the waveform to obtain the waveform change rate of that card swipe. According to the order of card swipes, the waveform change rates of two adjacent card swipes are compared sequentially to determine whether the latter is greater than the former, that is, the change rate is progressively increased. The process involves comparing each waveform change rate with a pre-stored baseline deviation threshold on the smart terminal. This threshold is based on the upper bound of the statistical distribution of waveform change rates measured from multiple swipes in historical normal transactions. A risk trend is confirmed only when the waveform change rate of each swipe within the window simultaneously meets two conditions: it is increasing relative to the previous swipe and its own value exceeds the baseline deviation threshold. Given that the initial risk score has already triggered an alert, by tracking the temporal evolution of the waveform change rate, occasional fluctuations or single anomalies are filtered out, accurately identifying high-risk swipe behaviors with continuously deteriorating characteristics. This avoids false alarms and provides a reliable basis for subsequent tiered early warnings and limiting the number of responses.

[0040] In some embodiments, the card-swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each swipe cycle and all waveform change rates, with reference to... Figure 3 The diagram is a flowchart illustrating the process of determining a credit card swipe risk score in some embodiments of this application. In this embodiment, determining the credit card swipe risk score can be achieved through the following steps: In step 1031, the preliminary risk score and the waveform change rate of the corresponding radio frequency signal reflection waveform are collected for each card swiping cycle; In step 1032, risk weights are assessed for each card-swiping cycle based on the rate of change of each waveform to obtain the risk weight for each card-swiping cycle. In step 1033, the preliminary risk scores of each card-swiping cycle are evaluated by risk fusion according to time sequence and weighted waveform change rate using various risk weights to obtain the card-swiping risk score of the smart terminal transaction environment.

[0041] It should be noted that, in this application, the waveform change rate is a numerical indicator used to describe the degree of fluctuation of the radio frequency signal reflection waveform over time within a single card swiping cycle; the risk weight is a proportional coefficient used to represent the contribution of a single card swiping cycle to the final risk among all card swiping cycles within the current sliding time window; and the card swiping risk score is a numerical indicator used to evaluate the overall card swiping risk level of a smart terminal within multiple consecutive card swiping cycles.

[0042] In practice, firstly, within the sliding time window, the multi-source information set corresponding to each card-swiping cycle is retrieved sequentially according to the order of card swiping. For each card-swiping cycle, the score value for that cycle is read from the previously calculated preliminary risk score result. The waveform change rate is calculated from the radio frequency signal reflection waveform of that cycle. The calculation method is as follows: arrange the waveform amplitude values ​​in chronological order, calculate the absolute value of the amplitude difference between two adjacent sampling points, sum all the absolute values ​​of adjacent differences, and then divide by the total duration of the waveform from the start time to the end time to obtain the waveform change rate for that cycle. The preliminary risk score and waveform change rate of each card-swiping cycle are paired according to the card-swiping order and recorded in the list of the sliding time window. For example, if the sliding time window contains three card-swiping cycles, the score and waveform change rate of the first cycle, the score and waveform change rate of the second cycle, and the score and waveform change rate of the third cycle are recorded respectively. The two values ​​of each card-swiping cycle are paired and recorded as the correspondence between the cycle score and the waveform change rate. Then, all the waveforms are... The waveform change rates are arranged into a numerical sequence according to the card swiping order. The sum of all waveform change rates in this sequence is calculated by adding the waveform change rate values ​​of each cycle to obtain a total value. For each card swiping cycle, the waveform change rate of that cycle is divided by this total value to obtain a ratio, which is the risk weight of that cycle. According to this calculation method, the cycle with a larger waveform change rate has a higher risk weight, indicating that the cycle has a greater impact on the overall risk. The sum of the risk weights of all cycles equals one. Finally, for each card swiping cycle, the preliminary risk score of that cycle is multiplied by the risk weight of that cycle to obtain a weighted score. The waveform change rate of that cycle is also multiplied by the risk weight of that cycle to obtain a weighted waveform change rate. The weighted scores of all cycles are added together according to the card swiping time order to obtain a weighted score sum. The weighted waveform change rates of all cycles are also added together to obtain a weighted waveform change rate sum. The result of multiplying the weighted score sum by the weighted waveform change rate sum again is the card swiping risk score of the smart terminal transaction environment.

[0043] In step 104, when the card swiping risk score is greater than the alarm threshold of the smart terminal, the smart terminal is triggered to issue a graded warning and the number of consecutive responses of the contactless card swiping function is limited.

[0044] It should be noted that in this application, based on the comparison results of the card swiping risk score and the preset alarm threshold, the warning response level of the smart terminal and the availability of subsequent card swiping functions are dynamically adjusted, thereby realizing gradual intervention in high-risk transaction behaviors. The higher the card swiping risk score, the more obvious the abnormal characteristics are and the worsening trend is during multiple consecutive card swiping processes. At this time, it is necessary to take different levels of warning measures according to the severity of the risk, and block potential repeated attack behaviors by limiting the number of consecutive responses of the contactless card swiping function.

[0045] In practice, the system first retrieves the alarm threshold of the smart terminal from its parameter library and compares the card-swiping risk score with this threshold. The alarm thresholds are set to two different values ​​by default: a Level 1 alarm threshold and a Level 2 alarm threshold, with the Level 2 threshold being higher than the Level 1 threshold. When the card-swiping risk score is greater than the Level 1 alarm threshold but less than or equal to the Level 2 alarm threshold, a Level 1 warning is triggered, indicated by a medium-frequency beep or a flashing yellow light, alerting the operator without interrupting the transaction process. When the card-swiping risk score is greater than the Level 2 alarm threshold, a Level 2 warning is triggered, indicated by a high-frequency beep or a rapidly flashing red light. Simultaneously, the system automatically records this abnormal event and uploads it to the backend server. After triggering any Level 1 warning, a restriction mechanism for contactless card swiping is immediately activated: a continuous response counter is set, with an initial value set to a preset upper limit of five times by default. Each time a user attempts to perform a contactless card swipe, the counter value decreases by one. When the counter value reaches zero, the smart terminal temporarily disables the contactless card swipe function. The user must complete the transaction through other methods such as contact card insertion or entering a password, or wait for a preset cooldown period after which the counter automatically returns to its initial value. The system uses tiered alerts to provide differentiated responses to different levels of risk, avoiding the desensitization of users by using the same intensity of alarms for all anomalies. At the same time, by limiting the number of consecutive responses, it effectively prevents multiple replay attacks or card duplication attempts that exploit contactless card swipe vulnerabilities in a short period of time, ensuring transaction security while also ensuring a smooth transition for the user experience.

[0046] Furthermore, in another aspect of this application, in some embodiments, this application provides a card swipe warning system for a smart terminal transaction environment. This card swipe warning system for a smart terminal transaction environment includes a tiered warning unit, as referenced. Figure 4 The figure is a schematic diagram of the structure of a hierarchical early warning unit according to some embodiments of this application. The hierarchical early warning unit includes: a data acquisition module 201, a processing module 202, and an execution module 203, which are described below: The acquisition module 201 in this application is mainly used to acquire the environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence and radio frequency signal reflection waveform during the transaction scenario of the smart terminal, forming a multi-source information set; Processing module 202, in this application, is used to input the multi-source information set into a pre-trained risk identification model, extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation, and then obtain a preliminary risk score of the smart terminal transaction environment. It should be noted that the processing module 202 is also used to continuously track the waveform change rate of the radio frequency signal reflection waveform in the smart terminal for a specified card swiping cycle by using a sliding time window when the preliminary risk score is greater than the risk threshold of the smart terminal. If the waveform change rate increases successively and exceeds the baseline deviation threshold of the smart terminal, the card swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card swiping cycle and all waveform change rates. The execution module 203 in this application is mainly used to trigger the smart terminal's graded early warning prompt when the card swiping risk score is greater than the alarm threshold of the smart terminal, and to limit the number of consecutive responses of the contactless card swiping function.

[0047] The foregoing has detailed examples of the card-swiping early warning method and system for smart terminal transaction environments provided in the embodiments of this application. It is understood that the corresponding apparatus, in order to achieve the above functions, includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0048] In some embodiments, this application also provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device executes the above-described card swipe warning method for a smart terminal transaction environment.

[0049] In some embodiments, reference Figure 5The dashed lines in the figure indicate that the unit or module is optional. This figure is a structural schematic diagram of a computer device for implementing a card-swiping warning method for a smart terminal transaction environment according to an embodiment of this application. The card-swiping warning method for a smart terminal transaction environment described in the above embodiments can... Figure 5 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a memory 302 and at least one communication unit 305. The computer device may be a terminal device, a server or a chip.

[0050] Processor 301 can be a general-purpose processor or a special-purpose processor. For example, processor 301 can be a central processing unit (CPU), which can be used to control computer devices, execute software programs, and process data from software programs. The computer device may also include a communication unit 305 for inputting (receiving) and outputting (transmitting) signals.

[0051] For example, the computer device may be a chip, and the communication unit 305 may be the input and / or output circuit of the chip, or the communication unit 305 may be the communication interface of the chip, which may be a component of a terminal device, network device or other device.

[0052] For example, the computer device may be a terminal device or a server, and the communication unit 305 may be a transceiver of the terminal device or the server, or the communication unit 305 may be a transceiver circuit of the terminal device or the server.

[0053] The computer device may include one or more memories 302 storing a program 304. The program 304 can be executed by a processor 301 to generate instructions 303, causing the processor 301 to execute the method described in the above method embodiments according to the instructions 303. Optionally, the memory 302 may also store data (such as a target audit model). Optionally, the processor 301 may also read data stored in the memory 302, which may be stored at the same storage address as the program 304, or it may be stored at a different storage address than the program 304.

[0054] The processor 301 and memory 302 can be configured separately or integrated together, for example, integrated on the system on chip (SOC) of the terminal device.

[0055] It should be understood that each step of the above method embodiment can be completed by hardware logic circuits or software instructions in the processor 301. The processor 301 can be a CPU, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, such as discrete gate, transistor logic devices, or discrete hardware components.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0057] For example, in some embodiments, this application also provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the above-described card-swiping warning method for a smart terminal transaction environment.

[0058] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0059] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for card swipe early warning in a smart terminal transaction environment, characterized in that, Includes the following steps: The system collects environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and radio frequency signal reflection waveform during the card swiping process in the transaction scenario of the smart terminal, forming a multi-source information set. The multi-source information set is input into a pre-trained risk identification model, and the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation are extracted respectively, so as to obtain a preliminary risk score of the smart terminal transaction environment. When the preliminary risk score is greater than the risk threshold of the smart terminal, the waveform change rate of the radio frequency signal reflection waveform in the smart terminal during a specified card swiping cycle is continuously tracked by a sliding time window. If the waveform change rate increases successively and exceeds the baseline deviation threshold of the smart terminal, the card swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card swiping cycle and all waveform change rates. When the card swiping risk score exceeds the alarm threshold of the smart terminal, the smart terminal will trigger a graded warning prompt and limit the number of consecutive responses of the contactless card swiping function.

2. The method as described in claim 1, characterized in that, The multi-source information set is input into a pre-trained risk identification model, and the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation are extracted respectively, thereby obtaining a preliminary risk score of the smart terminal transaction environment, specifically including: A pre-trained risk identification model is used to extract the joint anomaly index of the scene and the device from the environmental noise spectrum of the multi-source information set; A pre-trained risk identification model is used to extract temporal inconsistency scores of networks and behaviors from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequences of the multi-source information set; The radio frequency reflection feature deviation of the card swiping electromagnetic signal is extracted from the radio frequency signal reflection waveform of the multi-source information set using a pre-trained risk identification model. The joint anomaly index, the timing inconsistency score, and the radio frequency reflection characteristic deviation are integrated into a preliminary risk score for the smart terminal transaction environment.

3. The method as described in claim 2, characterized in that, The extraction of the joint anomaly index of the scene and the device from the environmental noise spectrum of the multi-source information set using a pre-trained risk identification model specifically includes: The environmental noise spectrum of the multi-source information set is segmented according to time windows, and the power spectral density and time-frequency distribution characteristics of each time window are extracted. The pre-trained risk identification model is used to learn the noise spectrum patterns of different smart terminal devices in normal transaction scenarios, and a joint distribution feature space of devices and scenarios is established. The power spectral density and the time-frequency distribution features are mapped to the joint distribution feature space, and the probability density of the joint distribution space with the normal distribution center is calculated to obtain the joint anomaly index of the scene and the device.

4. The method as described in claim 2, characterized in that, The temporal inconsistency score of network and behavior is extracted from the gyroscope attitude angular velocity and wireless network beacon fluctuation sequences of the multi-source information set using a pre-trained risk identification model. Specifically, this includes: Align the gyroscope attitude angular velocity and wireless network beacon fluctuation sequence of the multi-source information set with the same time axis to obtain dual-mode time series data; Extract the temporal correlation patterns between posture changes and beacon fluctuations in normal trading from a pre-trained risk identification model; The temporal correlation model is used to calculate the mutual information, temporal misalignment, and prediction error between the current bimodal sequences in the bimodal temporal data to obtain the temporal inconsistency score of the network and behavior.

5. The method as described in claim 2, characterized in that, The specific deviations in the radio frequency reflection characteristics of card-swiping electromagnetic signals extracted from the radio frequency signal reflection waveform of the multi-source information set using the pre-trained risk identification model include: Extract the effective reflection segment of the card swiping action from the radio frequency signal reflection waveform of the multi-source information set; The effective reflection segment is mapped to a high-dimensional feature space using a pre-trained risk identification model, and electromagnetic features including reflection intensity attenuation, phase change, and multipath distribution are extracted. The electromagnetic features are compared with those of a normal card swipe to obtain the radio frequency reflection feature deviation of the card swipe electromagnetic field.

6. The method as described in claim 1, characterized in that, The card-swiping risk score for the smart terminal transaction environment is determined by the preliminary risk score for each swipe cycle and all waveform change rates. Specifically, this includes: Collect the preliminary risk score and the waveform change rate of the corresponding radio frequency signal reflection waveform for each card swipe cycle; Risk weights for each card swiping cycle are assessed based on the rate of change of each waveform to obtain the risk weight for each card swiping cycle. By using various risk weights, the initial risk scores for each card-swiping cycle are fused and evaluated according to time sequence and weighted waveform change rate to obtain the card-swiping risk score of the smart terminal transaction environment.

7. The method as described in claim 1, characterized in that, The baseline deviation threshold is the critical discrimination value that distinguishes between normal card swipe fluctuations and abnormal, progressively increasing fluctuations.

8. A card swipe early warning system for a smart terminal transaction environment, the card swipe early warning system for a smart terminal transaction environment comprising a hierarchical early warning unit, characterized in that, The hierarchical early warning unit includes: The data acquisition module is used to collect the environmental noise spectrum, gyroscope attitude angular velocity, wireless network beacon fluctuation sequence, and radio frequency signal reflection waveform during the card swiping process of the smart terminal in the transaction scenario, forming a multi-source information set; The processing module is used to input the multi-source information set into a pre-trained risk identification model, extract the joint anomaly index of the scene and device, the temporal inconsistency score of the network and behavior, and the radio frequency reflection feature deviation, and then obtain a preliminary risk score of the smart terminal transaction environment. The processing module is also used to continuously track the waveform change rate of the radio frequency signal reflection waveform in the smart terminal for a specified card swiping cycle using a sliding time window when the preliminary risk score is greater than the risk threshold of the smart terminal. If the waveform change rate increases successively and exceeds the baseline deviation threshold of the smart terminal, the card swiping risk score of the smart terminal transaction environment is determined by the preliminary risk score of each card swiping cycle and all waveform change rates. The execution module is used to trigger a graded early warning prompt on the smart terminal and limit the number of consecutive responses of the contactless card swiping function when the card swiping risk score is greater than the alarm threshold of the smart terminal.

9. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to call and run the computer programs from the memory, so that the computer device executes the card swiping early warning method for the smart terminal transaction environment according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions or code that, when executed on a computer, cause the computer to implement the card-swiping early warning method for a smart terminal transaction environment as described in any one of claims 1 to 7.

Images