A method and apparatus for detecting transaction risks based on multi-electromagnetic wave characteristics

Abstract

This invention discloses a method and apparatus for transaction risk detection based on multi-electromagnetic wave features. First, the raw radio frequency (RF) signal generated during NFC card reading is acquired. A Fast Fourier Transform (FFT) is performed on the raw RF signal to generate a power spectral density map. Feature extraction is then performed on the power spectral density map to obtain multi-electromagnetic wave feature vectors. These feature vectors are then input into a transaction risk detection model. Based on the model, an anomaly score is determined. Finally, the transaction risk detection result is determined based on the anomaly score and a preset anomaly score threshold. This invention addresses the issue at the physical layer, using FFT to convert the raw RF signal into a power spectral density map that is easy to analyze and has distinct features, thereby revealing the fundamental flaws in abnormal transactions at the physical layer. Compared to existing transaction risk detection schemes, this method reduces the limitations of existing schemes and improves the accuracy of transaction risk detection.

Claims

1. A method for detecting transaction risks based on multi-electromagnetic wave characteristics, characterized in that the method includes: The raw radio frequency (RF) signal generated during NFC card reading is acquired, and a Fast Fourier Transform (FFT) is performed on the raw RF signal to generate a power spectral density map corresponding to the raw RF signal. Feature extraction is performed on the power spectral density map to obtain a multivariate electromagnetic wave feature vector. This multivariate electromagnetic wave feature vector includes a spectral entropy feature vector, a spectral skewness feature vector, a spectral kurtosis feature vector, and an out-of-band spurious power feature vector. This multivariate electromagnetic wave feature vector is input into a transaction risk detection model, and based on this model, a transaction anomaly score is determined. The transaction risk detection result is determined based on the transaction anomaly score and a preset anomaly score threshold.

2. The transaction risk detection method as described in claim 1, wherein, After acquiring the raw radio frequency signal generated during the NFC card reading process, before performing a fast Fourier transform on the raw radio frequency signal, the method further includes: performing noise reduction filtering and signal alignment processing on the raw radio frequency signal.

3. The transaction risk detection method as described in claim 1, wherein, The multivariate electromagnetic wave feature vector is input into the transaction risk detection model. Based on the transaction risk detection model, the transaction anomaly score is determined by: acquiring transaction data, inputting the transaction data and the multivariate electromagnetic wave feature vector into the transaction risk detection model, and determining the transaction anomaly score based on the transaction risk detection model; wherein, the transaction data includes transaction identifier, transaction amount, transaction time, and the identifier of the POS device involved in the transaction.

4. The transaction risk detection method as described in claim 1, wherein, Based on the transaction anomaly score and a preset anomaly score threshold, the transaction risk detection result is determined as follows: If the transaction anomaly score is greater than the preset first anomaly score threshold, the transaction risk detection result is determined to be an abnormal transaction; If the transaction anomaly score is less than the preset second anomaly score threshold, the transaction risk detection result is determined to be a normal transaction; If the transaction anomaly score is not less than the preset second anomaly score threshold and not greater than the preset first anomaly score threshold, the transaction risk detection result is determined to be a risky transaction.

5. The transaction risk detection method as described in claim 4, wherein, If the transaction risk detection result is determined to be a risky transaction, the method further includes: outputting transaction authentication prompt information, which includes at least one of password input prompt information, SMS verification prompt information, and biometric authentication prompt information; receiving the user's transaction authentication information, and executing the corresponding transaction behavior based on the transaction authentication information.

6. The transaction risk detection method as described in claim 4, wherein, If the transaction anomaly score is not less than the preset second anomaly score threshold and not greater than the first anomaly score threshold, determining the transaction risk detection result as a risky transaction includes: obtaining the transaction amount; if the transaction amount is greater than a preset amount threshold, and the transaction anomaly score is not less than the preset second anomaly score threshold and not greater than the first anomaly score threshold, determining the transaction risk detection result as a risky transaction.

7. The transaction risk detection method as described in claim 6, wherein, If the transaction amount is not greater than a preset amount threshold, the method further includes: determining that the transaction risk detection result is a normal transaction.

8. The transaction risk detection method as described in claim 3, wherein, The training of the transaction risk detection model includes: acquiring sample transaction data and sample multivariate electromagnetic wave feature vectors from the training set. These feature vectors are obtained by performing a Fast Fourier Transform on the original radio frequency signal generated during the NFC card reading process to generate a sample power spectral density map, and then extracting features from this power spectral density map. The sample transaction data, the sample multivariate electromagnetic wave feature vectors, and the transaction risk label are then input into the transaction risk detection model to be trained. Based on this model, a predicted transaction anomaly score is determined. A loss value is determined based on the predicted transaction anomaly score and the transaction risk label. Finally, the transaction risk detection model is trained based on this loss value.

9. The transaction risk detection method as described in claim 8, wherein, The method also includes: adding the newly labeled sample data to the training set according to a preset model optimization cycle, or when the number of newly labeled sample data reaches a preset threshold, and optimizing the training of the transaction risk detection model based on the updated training set.

10. A transaction risk detection device based on multi-electromagnetic wave characteristics, characterized in that the device comprises: A feature vector extraction module is used to acquire the raw radio frequency signal generated during NFC card reading, perform a Fast Fourier Transform on the raw radio frequency signal to generate a power spectral density map corresponding to the raw radio frequency signal, and extract features from the power spectral density map to obtain a multivariate electromagnetic wave feature vector. This multivariate electromagnetic wave feature vector includes a spectral entropy feature vector, a spectral skewness feature vector, a spectral kurtosis feature vector, and an out-of-band spurious power feature vector. A transaction risk detection module is used to input this multivariate electromagnetic wave feature vector into a transaction risk detection model, determine a transaction anomaly score based on the transaction risk detection model, and determine the transaction risk detection result based on the transaction anomaly score and a preset anomaly score threshold.

11. The transaction risk detection device as described in claim 10, wherein, This feature vector extraction module is also used to perform noise reduction filtering and signal alignment processing on the original radio frequency signal.

12. The transaction risk detection device as described in claim 10, wherein, This transaction risk detection module is specifically used to acquire transaction data, input the transaction data and the multivariate electromagnetic wave feature vector into the transaction risk detection model, and determine the transaction anomaly score based on the transaction risk detection model; wherein, the transaction data includes transaction identifier, transaction amount, transaction time, and the identifier of the POS device involved in the transaction.

13. The transaction risk detection device as described in claim 10, wherein, The transaction risk detection module is specifically used to determine that the transaction risk detection result is an abnormal transaction if the transaction abnormality score is greater than a preset first abnormality score threshold; to determine that the transaction risk detection result is a normal transaction if the transaction abnormality score is less than a preset second abnormality score threshold; and to determine that the transaction risk detection result is a risky transaction if the transaction abnormality score is not less than the preset second abnormality score threshold and not greater than the first abnormality score threshold.

14. The transaction risk detection device as described in claim 13, wherein, The transaction risk detection module is also used to output transaction authentication prompt information if the transaction risk detection result is determined to be a risky transaction. The transaction authentication prompt information includes at least one of the following: password input prompt information, SMS verification prompt information, and biometric authentication prompt information; receive the user's transaction authentication information, and execute the corresponding transaction behavior based on the transaction authentication information.

15. The transaction risk detection device as described in claim 13, wherein, The transaction risk detection module is also used to obtain the transaction amount. If the transaction amount is greater than a preset amount threshold, the transaction anomaly score is not less than a preset second anomaly score threshold and not greater than the first anomaly score threshold, and the transaction risk detection result is determined to be a risky transaction.

16. The transaction risk detection device as described in claim 15, wherein, The transaction risk detection module is also used to determine that the transaction risk detection result is a normal transaction if the transaction amount is not greater than a preset amount threshold.

17. The transaction risk detection device as described in claim 12, wherein, The device further includes: a training module for acquiring sample transaction data and sample multivariate electromagnetic wave feature vectors in the training set, wherein the sample multivariate electromagnetic wave feature vectors are obtained by performing a fast Fourier transform on the original radio frequency signal generated during the sample NFC card reading process to generate a sample power spectral density map, and extracting features from the sample power spectral density map; inputting the sample transaction data, the sample multivariate electromagnetic wave feature vectors, and transaction risk labels into the transaction risk detection model to be trained, determining a predicted transaction anomaly score based on the transaction risk detection model; determining a loss value based on the predicted transaction anomaly score and the transaction risk label; and training the transaction risk detection model based on the loss value.

18. The transaction risk detection device as described in claim 17, wherein, The training module is also used to add the newly labeled sample data to the training set according to the preset model optimization cycle or when the number of newly labeled sample data reaches the preset threshold, and to optimize and train the transaction risk detection model based on the updated training set.

19. An electronic device, wherein, The device includes a processor, a communication interface, memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other through the communication bus; the memory is used to store computer programs; and the processor, when executing the programs stored in the memory, implements the transaction risk detection method as described in any one of claims 1 to 9.

20. A computer-readable storage medium, wherein, The computer-readable storage medium contains a computer program that, when executed by a processor, implements the transaction risk detection method as described in any one of claims 1 to 9.

21. A computer program product, wherein, The computer program product includes an executable program that is executed by a processor to implement the transaction risk detection method as described in any one of claims 1 to 9.

Images