A low-data-loss-rate online lending data privacy protection method based on inhibition strategy

Abstract

The application discloses to the network information security technical field, specifically for the low data loss rate of the network loan data privacy protection method based on the inhibition strategy, including the following operation steps: S1: the information quantity of each attribute needing to be protected is calculated through the entropy algorithm, S2: the privacy attribute protection level of the user is learned and predicted according to the demand and individualization of the user, so that the dynamic property of the privacy protection model is improved, S3: then according to the different scene, the sensitive attribute and the insensitive attribute are distinguished, the classification protection is carried out, S4: five parameters, sensitivity, specificity, data accuracy, false positive and false negative, are introduced, the application can prove the strength of the sensitive attribute of the privacy, facilitate the timely protection of the privacy of the network loan user, and achieve the classification and grading protection of the sensitive attribute of the privacy, increase the individualization level of the privacy protection, and prevent the privacy leakage.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network information security, in particular to a P2P data privacy protection method based on a low data loss rate and a suppression strategy. BACKGROUND

[0002] In recent years, a lending platform will require a borrower to provide as much information as possible about personal credit. The more information is rich and open, the more accurate the lender's measurement of the borrower's repayment ability, and the more stable the platform's trading activities. With the extension of information collection in the breadth and depth direction, the range of borrower information and even personal privacy disclosure is becoming wider and wider. Once an information leakage event occurs, the consequences will be disastrous. In order to prevent others from illegally obtaining the privacy information of others, we need to protect their personal information, but the method of data encryption will reduce the data mining efficiency.

[0003] The existing P2P (peer-to-peer) network protection is a privacy protection technology based on the k-anonymity model. From the existing research, the basic idea of k-anonymity and its extension model is to define the attributes related to the attacker's background knowledge in the data set as quasi-identifier. However, the security of this defense mechanism is related to the background knowledge possessed by the attacker, so it cannot provide an effective and strict method to prove the level of privacy protection. Therefore, we propose a P2P data privacy protection method based on a low data loss rate and a suppression strategy. SUMMARY

[0004] In view of the above and / or existing problems in the P2P data privacy protection method based on a low data loss rate and a suppression strategy, the present application is proposed.

[0005] Therefore, the purpose of the present application is to provide a P2P data privacy protection method based on a low data loss rate and a suppression strategy. The amount of information of each attribute that needs to be protected is calculated by an entropy algorithm. Then, according to the user's needs and individual needs, the privacy attribute protection level is learned and predicted in depth. Then, according to different scenarios, the classification protection is selected, which can solve the above-mentioned existing problems.

[0006] To solve the above technical problems, according to one aspect of the present application, the present application provides the following technical scheme:

[0007] The P2P data privacy protection method based on a low data loss rate and a suppression strategy comprises the following operation steps:

[0008] S1: calculating the amount of information of each attribute that needs to be protected by an entropy algorithm;

[0009] S2: learning and predicting the privacy attribute protection level in depth according to the user's needs and individual needs.

[0010] S3: Then, based on different scenarios, select and distinguish between sensitive and non-sensitive attributes for categorized protection;

[0011] S4: Introduce 5 parameters: sensitivity, specificity, data precision, false positives, and false negatives;

[0012] S5: Calculate the sensitivity of the data = t_pos / pos;

[0013] S6: Calculate data specificity = t_neg / neg;

[0014] S7: Calculate the error negatives of the data = 1 - t_pos / pos;

[0015] S8: Error positives in technical data = 1 - t_neg / neg;

[0016] S9: Calculate data precision = t_pos(t_pos+f_pos);

[0017] S10: Determine the final accuracy.

[0018] As a preferred embodiment of the online lending data privacy protection method with low data loss rate based on suppression strategy described in this invention, wherein: in S1, the entropy algorithm yields the largest amount of effective information and has the highest sensitivity, and the suppression strategy is used to suppress highly sensitive data.

[0019] As a preferred embodiment of the online lending data privacy protection method with low data loss rate based on suppression strategy described in this invention, in step S3, when pl = 1, ε = γ. Conversely, the lower the privacy level pl, the larger ε is, and the lower the privacy protection strength is. When pl is close to 0, ε is close to ∞.

[0020] As a preferred embodiment of the online lending data privacy protection method with low data loss rate based on suppression strategy described in this invention, wherein: in S5, t_pos is the actual number of samples and pos is the number of positive samples.

[0021] As a preferred embodiment of the online lending data privacy protection method with low data loss rate based on suppression strategy described in this invention, wherein: in S6, t_neg is the number of truly negative samples and neg is the number of negative samples.

[0022] As a preferred embodiment of the online lending data privacy protection method with low data loss rate based on suppression strategy described in this invention, wherein: in S9, f_pos is the number of false positive samples.

[0023] As a preferred embodiment of the online lending data privacy protection method with low data loss rate based on suppression strategy described in this invention, wherein: the final accuracy rate in S10 is:

[0024] A=(t_pos / pos)*pos / (pos+neg)+(t_neg / neg)*neg / (pos+neg).

[0025] Compared with existing technologies: In order to obtain the sensitive attributes of privacy, the entropy algorithm is used to calculate the amount of information of each attribute that needs to be protected. The entropy algorithm yields the largest amount of effective information and the highest sensitivity, thus proving the strength of the sensitive attributes of privacy, facilitating timely protection of privacy, and ensuring that users' sensitive privacy is not leaked.

[0026] By using deep learning to predict the level of privacy protection based on user needs and personalization requirements, and then selecting different protection levels according to different scenarios, sensitive and non-sensitive attributes are distinguished and classified for protection, thereby increasing the level of privacy protection.

[0027] Given γ, the higher the privacy level pl, the smaller the allocated privacy protection budget ε, and the stronger the privacy protection. When pl = 1, ε = γ. Conversely, the lower the privacy level pl, the larger ε, and the weaker the privacy protection. When pl is close to 0, ε is close to ∞, thus classifying and classifying sensitive privacy attributes to prevent privacy leakage.

[0028] Since the purpose of data publishers releasing personal privacy data is for researchers to conduct data mining and meet the needs of multiple researchers to complete different data mining tasks, it is necessary to consider how to improve data utility while ensuring user privacy. It is necessary to judge the usability of the data. However, in many data mining applications, it is inappropriate to describe the performance of the model by measuring the number of global error rates. To describe the quality of the model, more complex and global metrics are required. Therefore, it is necessary to introduce five parameters: sensitivity, specificity, data precision, false positives, and false negatives. These metrics are defined as follows: data sensitivity = t_pos / pos, data specificity = t_neg / ne, false negatives = 1 - t_pos / pos, false positives = 1 - t_neg / neg, and data precision = t_pos(t_pos + f_pos). This allows for the calculation of the data loss rate, thereby enabling timely processing and compensation of privacy leaks.

[0029] Then, the final accuracy is calculated as A = (t_pos / pos)*pos / (pos+neg) + (t_neg / neg)*neg / (pos+neg).

[0030] This allows for the calculation of the accuracy rate of privacy information, thus facilitating timely processing and remediation of leaked privacy information. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0032] This invention provides a low-data-loss-rate method for protecting the privacy of online lending data based on a suppression strategy. It has the advantages of classifying and classifying sensitive privacy attributes for protection, preventing privacy leaks, and promptly processing and remedying leaked information. The method includes the following steps:

[0033] S1: Calculate the amount of information for each attribute that needs to be protected using the entropy algorithm;

[0034] Among them, the entropy algorithm yields the largest amount of effective information and has the highest sensitivity.

[0035] S2: Based on user needs and personalized requirements, perform deep learning and prediction of its own privacy protection level;

[0036] S3: Then, based on different scenarios, select and distinguish between sensitive and non-sensitive attributes for categorized protection;

[0037] The published data satisfies γ-privacy if and only if the privacy level pl at that point and the differential privacy protection budget ε allocated to that point satisfy ε × pl = γ. Given γ, the higher the privacy level pl, the smaller the allocated privacy protection budget ε, and the stronger the privacy protection. When pl = 1, ε = γ. Conversely, the lower the privacy level pl, the larger ε, and the weaker the privacy protection. When pl is close to 0, ε is close to ∞.

[0038] S4: Introduce 5 parameters: sensitivity, specificity, data precision, false positives, and false negatives;

[0039] Sensitivity is defined as sensitivity, specificity as specificity, precision as precision, false positives as false positives, and false negatives as false negatives.

[0040] S5: Calculate the sensitivity of the data = t_pos / pos;

[0041] S6: Calculate data specificity = t_neg / neg;

[0042] S7: Calculate the error negatives of the data = 1 - t_pos / pos;

[0043] S8: Error positives in technical data = 1 - t_neg / neg;

[0044] S9: Calculate data precision = t_pos(t_pos+f_pos);

[0045] Where t_pos is the true number of samples, pos is the number of positive samples, t_neg is the number of true negative samples, neg is the number of negative samples, and f_pos is the number of false positive samples;

[0046] S10: Determine the final accuracy rate;

[0047] The final accuracy rate was:

[0048] A=(t_pos / pos)*pos / (pos+neg)+(t_neg / neg)*neg / (pos+neg).

[0049] In practical applications, those skilled in the art, in order to obtain sensitive privacy attributes, use entropy algorithms to calculate the amount of information for each attribute that needs to be protected. If an attacker obtains information about a certain attribute, the entropy algorithm yields the largest amount of effective information, indicating the highest sensitivity. This demonstrates the strength of the sensitive privacy attributes, facilitating timely privacy protection. By using deep learning and prediction based on user needs and personalized requirements, the privacy attribute protection level can be determined. Then, selections can be made according to different scenarios, distinguishing between sensitive and insensitive attributes for categorized protection, thereby increasing the level of privacy protection. When γ is given, the higher the privacy level pl, the smaller the allocated privacy protection budget ε, and the greater the privacy protection strength. When pl = 1, ε = γ. Conversely, the lower the privacy level (pl), the larger the ε, and the weaker the privacy protection. When pl approaches 0, ε approaches ∞, thus classifying and grading sensitive privacy attributes to prevent privacy leaks. Since the purpose of data publishers releasing personal privacy data is for researchers to conduct data mining to meet the needs of multiple researchers completing different data mining tasks, it is necessary to consider how to improve data utility while ensuring user privacy. This requires assessing data usability. However, in many data mining applications, using the number of global error rates to describe model performance is inappropriate. To describe model quality, more complex and global metrics are needed. Therefore, five parameters need to be introduced: sensitivity, specificity, and... Data accuracy, positive errors, and negative errors are defined as follows: data sensitivity = t_pos / pos, data specificity = t_neg / ne, negative errors = 1 - t_pos / pos, positive errors = 1 - t_neg / neg, and data accuracy = t_pos(t_pos + f_pos). This allows for the calculation of data loss rate, enabling timely processing and remediation of privacy leaks. Finally, the accuracy rate of privacy information is calculated using the final accuracy rate A = (t_pos / pos)*pos / (pos + neg) + (t_neg / neg)*neg / (pos + neg). This facilitates timely processing and remediation of leaked privacy information.

[0050] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for protecting online lending data privacy with low data loss rate based on suppression strategies, characterized in that: The following steps are included: S1: Calculate the amount of information for each attribute that needs to be protected using the entropy algorithm; S2: Based on user needs and personalized requirements, perform deep learning and prediction on the privacy protection level of the model to improve the dynamism of the privacy protection model; S3: Then, based on different scenarios, select and distinguish between sensitive and non-sensitive attributes for categorized protection; S4: Introduce 5 parameters: sensitivity, specificity, data precision, false positives, and false negatives; S5: Calculate the sensitivity of the data = t_pos / pos; S6: Calculate data specificity = t_neg / neg; S7: Calculate the error negatives of the data = 1 - t_pos / pos; S8: Error positives in technical data = 1 - t_neg / neg; S9: Calculate data precision = t_pos(t_pos+f_pos); S10: Determine the final accuracy.

2. The method for protecting online lending data privacy with low data loss rate based on suppression strategy according to claim 1, characterized in that, The S1 algorithm yields the largest amount of effective information and has the highest sensitivity.

3. The method for protecting online lending data privacy with low data loss rate based on suppression strategy according to claim 1, characterized in that, In S3, when pl = 1, ε = γ. Conversely, the lower the privacy level pl, the larger ε is, and the lower the privacy protection strength is. When pl is close to 0, ε is close to ∞.

4. The method for protecting online lending data privacy with low data loss rate based on suppression strategy according to claim 1, characterized in that, In S5, t_pos is the actual number of samples, and pos is the number of positive samples.

5. The method for protecting online lending data privacy with low data loss rate based on suppression strategy according to claim 1, characterized in that, In S6, t_neg represents the number of truly negative samples, and neg represents the number of negative samples.

6. The method for protecting online lending data privacy with low data loss rate based on suppression strategy according to claim 1, characterized in that, In S9, f_pos represents the number of false positive samples.

7. The method for protecting online lending data privacy with low data loss rate based on suppression strategy according to claim 1, characterized in that, The final accuracy rate in S10 is: A=(t_pos / pos)*pos / (pos+neg)+(t_neg / neg)*neg / (pos+neg).