Feature standardization method and device in cross-feature federation, electronic equipment and medium

Abstract

This disclosure relates to a feature standardization method, apparatus, electronic device, and computer-readable medium in cross-feature federation, and pertains to the field of computer technology. The method is applied to at least two participants, each holding feature fragments generated based on sample features during the stealth sample alignment process. Each participant can determine a feature difference between the first mean of its held feature fragments and the feature fragments themselves, and then determine a target standard deviation among the participants based on this feature difference. Furthermore, each participant can standardize its held feature fragments based on the feature difference and the target standard deviation. This method breaks the limitation of local feature standardization by determining the feature difference locally within each participant and determining the target standard deviation among the participants using their respective feature differences. It achieves feature fragment standardization for each participant based on stealth sample alignment, supporting subsequent federated training tasks.

Description

Technical Field

[0001] This disclosure relates to the field of privacy computing technology, and more specifically, to a method, apparatus, electronic device, and computer-readable medium for feature standardization in a cross-feature federation. Background Technology

[0002] Cross-feature federated modeling typically involves communication between different participants. For example, sample alignment is required among the participants first. After obtaining the intersection of the participants through a safe intersection method, the features corresponding to the intersection in each participant's local area are determined, and feature standardization is performed in each participant's local area to support subsequent federated training.

[0003] However, during the secure intersection process, the data within the intersection is shared with all participants, posing a certain data privacy security risk. Using anonymous sample alignment for intersection can ensure that the intersection data is not known to the participants, which improves data privacy security to a certain extent. However, it will lead to the fragmentation of the features corresponding to the intersection and their distribution among different participants, making it impossible to complete feature standardization locally on each participant's site. Summary of the Invention

[0004] The purpose of this disclosure is to provide a feature standardization method, apparatus, electronic device, and computer-readable medium in cross-feature federation. This method can complete feature standardization to support subsequent federated model training tasks while ensuring the security of privacy data during sample alignment in cross-feature federation modeling scenarios.

[0005] According to a first aspect of this disclosure, a feature standardization method is provided in a cross-feature federation. This method can be applied to at least two participants, each of whom holds feature fragments generated based on sample features during the cloaking sample alignment process. The method may include: each participant determining a feature difference between the first mean and the feature fragments held by it; determining a target standard deviation among the participants based on the feature difference; and standardizing the feature fragments held by each participant based on the feature difference and the target standard deviation.

[0006] Optionally, determining the target standard deviation among the participants based on the feature differences includes: each participant determining a second mean of the squared differences based on the feature differences; each participant performing a secure multi-party operation based on the feature differences to obtain a third mean of the product of the feature differences between each pair of participants; and determining the target standard deviation among the participants based on the second mean and the third mean.

[0007] Optionally, the participants include a first participant and a second participant. The target standard deviation is determined among the participants based on the second mean and the third mean, including: the first participant and the second participant each exchange the second mean; the first participant and the second participant each calculate the target standard deviation based on the second mean and the third mean.

[0008] Optionally, if the number of participants is greater than or equal to 3, the target standard deviation is determined among the participants based on the second mean and the third mean, including: each participant synchronizing the second mean and the third mean to the target participant, where the target participant is any one of the participants; and the target participant determining the target standard deviation based on the second mean and the third mean of each participant.

[0009] Optionally, based on the feature difference and the target standard deviation, the feature fragments held by each participant are standardized, including: the target participant synchronizing the target standard deviation with each participant; and each participant comparing the feature difference with the target standard deviation to standardize the feature fragments held by each participant.

[0010] Optionally, the target participant is the first participant among all participants.

[0011] Optionally, each feature fragment satisfies additive secret sharing.

[0012] According to a second aspect of this disclosure, a feature standardization apparatus is provided in a cross-feature federation. This apparatus is applied to at least two participants, each holding feature fragments generated based on sample features during the cloaking sample alignment process. The apparatus may include: a difference determination module, used by each participant to determine a feature difference between the first mean and the feature fragments held, based on the first mean of the feature fragments held; a standard deviation determination module, used to determine a target standard deviation among the participants based on the feature difference; and a feature standardization module, used to standardize the feature fragments held by each participant based on the feature difference and the target standard deviation.

[0013] Optionally, the standard deviation determination module is specifically used for each participant to determine the second mean of the squared difference based on the feature difference; for each participant to perform a secure multi-party operation based on the feature difference to obtain the third mean of the product of the feature differences between each pair of participants; and for each participant to determine the target standard deviation based on the second mean and the third mean.

[0014] Optionally, the participants include a first participant and a second participant. The standard deviation determination module is specifically used for the first participant and the second participant to exchange their second means; the first participant and the second participant each calculate the target standard deviation based on the second means and the third means.

[0015] Optionally, if the number of participants is greater than or equal to 3, the standard deviation determination module is specifically used for each participant to synchronize the second mean and the third mean to the target participant, where the target participant is any one of the participants; the target participant determines the target standard deviation based on the second mean and the third mean of each participant.

[0016] Optionally, the feature standardization module is specifically used to synchronize the target standard deviation among all participants in the target direction; each participant compares the feature difference with the target standard deviation to standardize the feature fragments held by each participant.

[0017] Optionally, the target participant is the first participant among all participants.

[0018] Optionally, each feature fragment satisfies additive secret sharing.

[0019] According to a third aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the feature standardization method across feature federations described in the first aspect.

[0020] According to a fourth aspect of this disclosure, an electronic device is provided, comprising:

[0021] Processor; and

[0022] Memory, used to store computer programs for the processor;

[0023] The processor is configured to implement the feature standardization method in the cross-feature federation described above by executing a computer program.

[0024] According to a fifth aspect of this disclosure, a computer program product is provided that, when run on an electronic device, causes the electronic device to perform steps such as implementing the feature standardization method across feature federations as described in the first aspect.

[0025] This disclosure provides a feature standardization method for cross-feature federation. This method is applied to at least two participants, each holding feature fragments generated after aligning stealth samples. Each participant can determine the feature difference between the first mean of its held feature fragments and the feature fragment itself, and then determine a target standard deviation among the participants based on the feature difference. Furthermore, each participant can standardize its held feature fragments based on the feature difference and the target standard deviation. This method breaks the limitation of local feature standardization by determining the feature difference locally at each participant and determining the target standard deviation among the participants using their respective feature differences. It achieves feature fragment standardization for each participant based on stealth sample alignment, enabling standardization even when sample features are dispersed across participants, thus supporting subsequent federated training tasks.

[0026] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0027] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0028] Figure 1 This is one of the flowcharts of a feature standardization method in a cross-feature federation provided in an embodiment of this disclosure.

[0029] Figure 2 The second flowchart illustrates the steps of the feature standardization method in a cross-feature federation provided in this embodiment of the disclosure.

[0030] Figure 3 This is one of the interactive flowcharts of the feature standardization method in cross-feature federation provided in this embodiment of the disclosure.

[0031] Figure 4 This is the second schematic diagram of the interaction flow of the feature standardization method in cross-feature federation provided in this embodiment of the disclosure.

[0032] Figure 5 A schematic diagram of the structure of a feature standardization device in a cross-feature federation provided in an embodiment of this disclosure.

[0033] Figure 6 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0034] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0035] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0036] It should be noted that the data obtained in this public disclosure, including the participating parties, samples held by the participating parties, and characteristics, are all accessed, collected, stored, and used for subsequent analysis and processing after the user or relevant data owner has been clearly informed of the content of the data collection, the purpose of the data, and the processing method, and with the consent and authorization of the user or relevant data owner. Furthermore, the public may send the user or relevant data owner access, correction, or deletion of the data, as well as the method to revoke consent or authorization.

[0037] The method provided in this disclosure is applied to federated training. To further enhance data security, anonymization can also be used during the sample alignment stage to obtain fully anonymous samples that do not expose any participant. After anonymization, each participant holds fragments of the fully anonymous sample, including fragmented sample labels and fragmented sample features. Since each participant only holds fragments of the fully anonymous sample, it is difficult to standardize the overall features locally on each participant's end.

[0038] In this embodiment, each participant can determine the feature difference between the first mean and the feature fragments they hold, and determine the target standard deviation of the fully anonymous sample based on the feature difference among different participants. Then, standardization is performed locally by each participant based on the feature difference and the target standard deviation. This achieves feature fragment standardization for each participant based on the alignment of the anonymous sample, supporting subsequent federated training tasks. The specific details are explained below with reference to the accompanying drawings:

[0039] Figure 1 This is one of the flowcharts illustrating the steps of a feature normalization method in a cross-feature federation provided in this disclosure. The method is applied to at least two parties, each holding feature fragments generated based on sample features during the cloaked sample alignment process.

[0040] In this embodiment, each participant is a different participant in the cross-feature federated modeling process. Each participant holds sample data, which may include sample labels such as user identity information, user identifiers, product names, and merchant information. The sample data may also include different types of sample features, such as the user's age and the amount spent, or the product's price, transaction volume, and transaction counterparty. Among the participants in the cross-feature federated modeling, the types of sample features in their sample data are different, and there is at least some overlap in the sample labels.

[0041] Furthermore, in this embodiment of the disclosure, during cross-feature federated modeling, stealth sample alignment can be used to determine the intersection of sample labels among the participants. This process involves dense-state obfuscation, preventing each participant from knowing the sample labels in each row of the other's intersection, and also preventing them from knowing the sample labels in each row of their own intersection. This ensures that the intersection is not leaked while maintaining sample alignment. After stealth sample alignment, each participant holds fragments of the intersection, including fragments of sample labels, fragments of sample features, etc. These feature fragments can then be processed for subsequent feature standardization.

[0042] like Figure 1 As shown, the method may specifically include steps 101 to 103 as shown below.

[0043] Step 101: Each participant determines the feature difference between the first mean and the feature fragment based on the first mean of the feature fragments they hold.

[0044] In this embodiment of the disclosure, each participant can calculate the first mean value by performing an average calculation on the feature fragments it holds locally, and determine the feature difference between the feature fragments it holds and the corresponding first mean value based on the first mean value.

[0045] Step 102: Determine the target standard deviation among the participants based on the feature differences.

[0046] In this embodiment, the standard deviation refers to the arithmetic square root of the variance. Since the variance can be obtained by averaging the squares of the feature differences, the variance of all feature fragments at intersection can be determined among the participating parties based on all feature differences, thereby determining the corresponding target standard deviation. In this process, each participating party no longer directly standardizes its own sample features locally. Instead, it first determines the feature differences locally, and then determines the target standard deviation of all feature fragments at intersection based on all feature differences through communication between the participating parties.

[0047] Step 103: Standardize the feature fragments held by each participant based on feature difference and target standard deviation.

[0048] In this embodiment, Z-Score standardization (standard deviation standardization) can be used to standardize the feature fragments held by each participant. The standardized representation of the feature fragments is obtained by the feature difference and the target standard deviation. By standardizing the feature fragments held by each participant, the overall sample features are standardized. In subsequent federated modeling, the convergence speed of the weight parameters can be improved and the model training efficiency can be improved based on the standardization process.

[0049] Figure 2 This is the second flowchart illustrating the steps of a cross-feature federation feature normalization method provided in this embodiment of the disclosure. The method is applied to at least two participants, each holding feature fragments generated based on sample features during the cloaked sample alignment process.

[0050] Step 201: Each participant determines the feature difference between the first mean and the feature fragment based on the first mean of the feature fragments they hold.

[0051] In this embodiment of the disclosure, step 201 can be referred to the relevant description of step 101 above. To avoid repetition, it will not be repeated here.

[0052] In this embodiment of the disclosure, the Z-Score normalization method can be used, which is implemented in conventional data processing based on the following formula (1):

[0053]

[0054] Where mean is the first mean of the features in the intersection set, and std is the target standard deviation of the features in the intersection set. <x> 1, <x>2 represents fragments with different characteristics that intersect.

[0055] It can be seen that the standardization of the overall sample features in the intersection can be achieved by adding the different feature fragments generated based on the sample features after standardization.

[0056] In this embodiment of the disclosure, it can be seen from the derivation of the above formula (1) that the standardization of different feature fragments can be achieved based on the feature difference between the feature fragment and its first mean, and the target standard deviation of the sample features in the intersection. On this basis, each participant can determine the first mean of its own feature fragment locally, and further determine the feature difference between the feature fragment held by the participant and its corresponding first mean.

[0057] That is, participant 1 holds feature fragments. <x>If 1, then its first mean is Based on this, the characteristic difference of participant 1 is determined.

[0058] Participant 2 holds feature fragments <x>2, then its first mean is Based on this, the characteristic difference of participant 2 is determined.

[0059] Step 202: Each participant determines the second mean of the squared difference based on the feature difference.

[0060] Step 203: Perform secure multi-party computation based on feature differences among the participants to obtain the third mean of the product of feature differences between each pair of participants.

[0061] Step 204: Determine the target standard deviation among all participants based on the second and third means.

[0062] In this embodiment of the disclosure, the standard deviation is the arithmetic square root of the variance, and therefore the square of the standard deviation can be expressed as the following formula (2):

[0063]

[0064] Where std is the target standard deviation, x j The features of the intersecting samples. <x j >1. <x j >2 represents the feature fragments of the intersection; furthermore, with <x j >′1 means and with <x j >′2 means

[0065] As can be seen from the derivation of the aforementioned formula (2), the standard deviation of the features of the intersection samples can be obtained based on the feature differences of each feature fragment, specifically including the square mean of the feature differences corresponding to different feature fragments, and the mean of the product of each feature difference.

[0066] Based on this, according to steps 202 to 204, each participant can determine the second mean of the square of the difference and the third mean of the product of the difference based on the feature difference, and obtain the target standard deviation std based on the second mean and the third mean through formula (2).

[0067] In this embodiment of the disclosure, secure multi-party computation is used to compute the product of feature fragments among the participants. Multiplication can be performed without disclosing the privacy inputs and outputs of the participants. Secure multi-party computation can be built on basic technologies such as hash functions, homomorphic encryption, and pseudo-random functions. It also includes basic tools such as secret sharing, unintentional transmission protocols, and unintentional pseudo-random functions. Those skilled in the art can choose the method to implement secure multi-party computation according to actual needs. This embodiment of the disclosure does not impose specific limitations on this.

[0068] In an optional embodiment of this disclosure, if the participants include a first participant and a second participant, then step 204 may include the following steps S11 to S12.

[0069] Step S11: The first participant and the second participant each exchange the second mean;

[0070] Step S12: The first participant and the second participant each calculate the target standard deviation based on the second mean and the third mean.

[0071] In this embodiment of the disclosure, during the joint modeling process, when the participants include a first participant and a second participant, the first participant and the second participant can exchange the second mean after determining the second mean, so that each participant can obtain the second mean corresponding to all feature fragments. Then, based on the third mean obtained by secure multi-party computation, the numerical terms required to calculate the target standard deviation based on the second mean, the third mean, etc., can be calculated locally by each participant.

[0072] Specifically, Figure 3 This illustration shows one of the interactive flow diagrams of the feature normalization method in a cross-feature federation according to an embodiment of this disclosure, such as... Figure 3 As shown, if a first participant and a second participant are included, the interaction process may include the following steps 301 to 307.

[0073] Step 301: The first participant inputs feature fragments. <x>1. Calculate the first mean of the feature fragments locally. and feature difference

[0074] Step 302: The second participant inputs feature fragments. <x>2. Calculate the first mean of the feature fragments locally. and feature difference

[0075] Step 303: The first and second participants perform a secure multi-party computation and each obtains a third mean.

[0076] Step 304: The first participant calculates the squared feature difference locally. <x>′1 2 Second mean And send it to the second participant;

[0077] Step 305: The second participant calculates the squared feature difference locally. <x>'2 2 Second mean And send it to the first participant;

[0078] Step 306, the first participant based on Determine the target standard deviation (std) and perform feature fragmentation based on std. <x>1. Standardize.

[0079] Step 307, the second participant based on Determine the target standard deviation (std) and perform feature fragmentation based on std. <x>2. Standardize.

[0080] In an optional embodiment of this disclosure, if the number of participants is greater than or equal to 3, then step 204 may include steps S21 to S22 as follows.

[0081] Step S21: Each participant synchronizes the second and third average values ​​to the target participant, which is any one of the participants.

[0082] Step S22: The target participant determines the target standard deviation based on the second and third means of each participant.

[0083] In this embodiment, if the number of participants is greater than or equal to 3, it can be confirmed that the number of participants is relatively large. At this time, any one of the participants can be selected as the target participant, and the other participants other than the target participant will synchronize the second mean, third mean, etc. to the target participant. The target participant calculates its local second mean, obtains its local third mean through pairwise multi-party secure computation, and obtains all the second and third means of other participants through synchronization. Thus, the target participant obtains the second and third means of all feature fragments of the intersection, and can calculate the numerical terms required for the target standard deviation based on the second and third means, etc., and determine the target standard deviation of the intersection features locally on the target participant.

[0084] In one embodiment of the method disclosed herein, the target participant is the first participant among all participants.

[0085] In this embodiment, the participants can be ranked to determine their position in data processing and modeling training. This ranking can be used to distinguish different participants or to adjust the execution of process steps based on the ranking. The ranking can be based on the order in which participants engage in cross-feature federated modeling, or participants can be randomly assigned to corresponding positions, or priorities can be pre-set for each participant and ranked accordingly. The target participant can be the first participant in the sequence. The first participant, after calculating the second and third means, receives synchronized second and third means from other participants at the second, third, ..., nth positions, and calculates the target standard deviation locally based on all the second and third means. At this time, the first participant can hold the position information of other participants to confirm successful reception of the second and third means from the corresponding participants. If the first participant determines that synchronization has failed, it can also communicate with the corresponding participants to determine the reason for the synchronization failure.

[0086] In multi-party data processing, Z-Score standardization can be achieved based on the following formula (3):

[0087]

[0088] Where i is determined by the number of participants. In this embodiment, formula (3) can be referred to the relevant description of the aforementioned formula (1), and will not be repeated here to avoid repetition.

[0089] Based on this, the standard deviation can be derived from the following formula (4):

[0090]

[0091] Where i is determined by the number of participants. In this embodiment, formula (4) can be referred to the relevant description of the aforementioned formula (2), and will not be repeated here to avoid repetition.

[0092] Specifically, Figure 4 This illustrates the second interactive flowchart of the feature normalization method in a cross-feature federation according to an embodiment of this disclosure, as shown below. Figure 3 As shown, it includes a first participant, a second participant, and a third participant. Where the first participant is set as the target participant, the interaction process can include steps 401 to 410 as follows.

[0093] Step 401: The first participant inputs feature fragments. <x>1. Calculate the first mean of the feature fragments locally. and feature difference

[0094] Step 402: The second participant inputs feature fragments. <x>2. Calculate the first mean of the feature fragments locally. and feature difference

[0095] Step 403: The third party inputs feature fragments. <x>3. Calculate the first mean of the feature fragments locally. and feature difference

[0096] Step 404: The first and second participants perform a secure multi-party computation and each obtains a third mean.

[0097] Step 405: The second and third participants perform a secure multi-party computation, and each obtains the third mean. And send it to the first participant;

[0098] Step 406: The first participant and the third participant perform a secure multi-party computation and each obtains the third mean.

[0099] Step 407: The first participant calculates the squared feature difference (x>'1) locally. 2 Second mean

[0100] Step 408: The second participant calculates the squared feature difference locally. <x>'2 2 Second mean And send it to the first participant;

[0101] Step 409: The third participant calculates the squared feature difference locally. <x>'3 2 Second mean And send it to the first participant;

[0102] Step 410, the first participant based on Determine the target standard deviation (std) and send it to the second and third participants.

[0103] Step 205: Standardize the feature fragments held by each participant based on feature difference and target standard deviation.

[0104] In this embodiment of the disclosure, step 205 can be referred to the relevant description of step 103 above. To avoid repetition, it will not be repeated here.

[0105] In an optional embodiment of this disclosure, based on the aforementioned step S22, step 205 may include the following steps S31 to S32.

[0106] Step S31: Target participation direction: All participants synchronize the target standard deviation.

[0107] Step S32: Each participant compares the feature difference with the target standard deviation to standardize the feature fragments held by each participant.

[0108] In this embodiment, after determining the target standard deviation, the target participant can synchronize the target standard deviation with each participant in the joint modeling process. This allows each participant to compare the feature differences with the target standard deviation locally, thereby standardizing the feature fragments held by each participant to support subsequent federated modeling tasks. For example, the first participant can send the target standard deviation (std) to the second and third participants, and the first participant can calculate... Second Participant Calculation And calculations by third-party participants Each entity completes the standardization of its respective local feature fragments.

[0109] In an optional method embodiment of this disclosure, each feature fragment satisfies additive secret sharing.

[0110] In this embodiment, feature fragments belonging to different participants can be reconstructed into sample features through addition. Based on this, after separately processing the feature fragments in a split state, the result of the sample feature calculation can be obtained through addition. In this embodiment, each participant only needs to provide the feature difference and the second mean, where the feature difference and the second mean are indirectly obtained values, without directly disclosing the participant's local mean information, thus effectively protecting the security of the feature fragment information held locally by the participant.

[0111] This disclosure provides a feature standardization method for cross-feature federation. This method is applied to at least two participants, each holding feature fragments generated after aligning stealth samples. Each participant can determine the feature difference between the first mean of its held feature fragments and the feature fragment itself, and then determine a target standard deviation among the participants based on the feature difference. Furthermore, each participant can standardize its held feature fragments based on the feature difference and the target standard deviation. This method breaks the limitation of local feature standardization by determining the feature difference locally at each participant and determining the target standard deviation among the participants using their respective feature differences. It achieves feature fragment standardization for each participant based on stealth sample alignment, enabling standardization even when sample features are dispersed across participants, thus supporting subsequent federated training tasks.

[0112] Figure 5 A feature standardization apparatus 500 in a cross-feature federation provided for embodiments of this disclosure, such as Figure 5 As shown, the device is applied to at least two participants, each of whom holds feature fragments generated based on sample features during the cloaking sample alignment process. The device may include: a difference determination module 501, used by each participant to determine the feature difference between the first mean and the feature fragments held by them; a standard deviation determination module 502, used to determine a target standard deviation among the participants based on the feature difference; and a feature standardization module 503, used to standardize the feature fragments held by each participant based on the feature difference and the target standard deviation.

[0113] In an optional device embodiment of this disclosure, the standard deviation determination module 502 is specifically used for each participant to determine a second mean of the squared difference based on the feature difference; each participant performs a secure multi-party operation based on the feature difference to obtain a third mean of the product of the feature differences between each pair of participants; and determines a target standard deviation among each participant based on the second mean and the third mean.

[0114] In an optional device embodiment of this disclosure, the participants include a first participant and a second participant. The standard deviation determination module 502 is specifically used for the first participant and the second participant to exchange their respective second means; the first participant and the second participant each calculate the target standard deviation based on the second means and the third means.

[0115] In an optional device embodiment of this disclosure, the number of participants is greater than or equal to 3. The standard deviation determination module 502 is specifically used for each participant to synchronize the second mean and the third mean to the target participant, where the target participant is any one of the participants; the target participant determines the target standard deviation based on the second mean and the third mean of each participant.

[0116] In an optional device embodiment of this disclosure, the feature standardization module 503 is specifically used to synchronize the target standard deviation of each participant in the target participation direction; each participant compares the feature difference with the target standard deviation to standardize the feature fragments held by each participant.

[0117] In an optional device embodiment of this disclosure, the target participant is the first participant among all participants.

[0118] In an optional device embodiment of this disclosure, each feature fragment satisfies additive secret sharing.

[0119] This disclosure provides a feature standardization apparatus for cross-feature federation. The apparatus is applied to at least two participants, each holding feature fragments generated based on sample features during the stealth sample alignment process. Each participant can determine the feature difference between the first mean of its held feature fragments and the feature fragment itself, and then determine a target standard deviation among the participants based on the feature difference. Furthermore, each participant can standardize its held feature fragments based on the feature difference and the target standard deviation. This method breaks the limitation of local feature standardization by determining the feature difference locally at each participant and determining the target standard deviation among the participants using their respective feature differences. It achieves feature fragment standardization for each participant based on stealth sample alignment, enabling standardization even when sample features are distributed across participants, thus supporting subsequent federated training tasks.

[0120] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0121] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0122] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0123] In an exemplary embodiment of this disclosure, an electronic device capable of implementing the above-described method is also provided.

[0124] Those skilled in the art will understand that various aspects of this disclosure can be implemented as a system, method, or program product. Therefore, various aspects of this disclosure can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "system."

[0125] The following reference Figure 6 To describe an electronic device 600 according to such an embodiment of the present disclosure. Figure 6 The electronic device 600 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0126] like Figure 6 As shown, the electronic device 600 is manifested in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, and a bus 630 connecting different system components (including storage unit 620 and processing unit 610).

[0127] The storage unit stores program code, which can be executed by the processing unit 610, causing the processing unit 610 to perform the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of this disclosure.

[0128] Storage unit 620 may include a readable medium in the form of a volatile storage unit, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include a read-only memory (ROM) 6203.

[0129] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0130] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0131] Electronic device 600 can also communicate with one or more external devices 600 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed through display unit 640 and input / output (I / O) interface 650 connected to display unit 640. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 660. As shown, network adapter 660 communicates with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0132] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0133] In exemplary embodiments of this disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the methods described above is stored. In some possible implementations, various aspects of this disclosure may also be implemented as a program product including program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure.

[0134] In embodiments of this disclosure, a program product for implementing the above-described methods is also provided. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of this disclosure is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0135] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0136] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0137] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0138] Program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0139] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of this disclosure and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0140] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.< / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x> < / x>

Claims

1. A method for feature standardization in a cross-feature federation, the method comprising: The method is applied to at least two participants, each of which holds a feature shard generated based on a sample feature in a de-identification sample alignment process, and the method comprises: Each participant determines a feature difference value between a first mean value of the feature shard held by the participant and the feature shard according to the first mean value; A target standard deviation is determined between the participants based on the feature difference value; The feature shards held by the participants are standardized based on the feature difference value and the target standard deviation; The target standard deviation is determined between the participants based on the feature difference value, comprising: Each participant determines a second mean value of a square of the difference value based on the feature difference value; A third mean value of a product of the feature difference value between each pair of participants is obtained through secure multi-party computation between the participants based on the feature difference value; The target standard deviation is determined between the participants based on the second mean value and the third mean value.

2. The method of claim 1, wherein, The participants include a first participant and a second participant, and the target standard deviation is determined between the participants based on the second mean value and the third mean value, comprising: The first participant and the second participant each exchange the second mean value; The first participant and the second participant each calculate a target standard deviation based on the second mean value and the third mean value.

3. The method of claim 1, wherein, The number of participants is greater than or equal to 3, and the target standard deviation is determined between the participants based on the second mean value and the third mean value, comprising: Each participant synchronizes the second mean value and the third mean value to a target participant, which is any participant among the participants; The target participant determines the target standard deviation based on the second mean value and the third mean value of each participant.

4. The method of claim 3, wherein, The feature shards held by the participants are standardized based on the feature difference value and the target standard deviation, comprising: The target participant synchronizes the target standard deviation to each participant; Each participant standardizes the feature shard held by the participant by comparing the feature difference value with the target standard deviation.

5. The method of claim 3, wherein, The target participant is a first participant among the participants.

6. The method of claim 1, wherein, Each feature shard satisfies an additive secret sharing.

7. A feature standardization apparatus in a cross-feature federation, the apparatus comprising: The device is applied to at least two participants, each of which holds a feature shard generated based on a sample feature in a de-identification sample alignment process, and the device comprises: A difference determination module configured to determine, for each participant, a feature difference value between a first mean value of the feature shard held by the participant and the feature shard according to the first mean value; A standard deviation determination module configured to determine a target standard deviation between the participants based on the feature difference value; A feature standardization module configured to standardize the feature shards held by the participants based on the feature difference value and the target standard deviation. The standard deviation determination module is specifically configured to: determine, by each of the participants, a second mean value of the square of the feature difference value; perform secure multi-party operation among the participants based on the feature difference value to obtain a third mean value of the product of the feature difference value between each pair of the participants; and determine, among the participants, the target standard deviation based on the second mean value and the third mean value.

8. An electronic device, comprising: Comprise: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to implement the feature standardization method in cross-feature federation according to any one of claims 1 to 6 by executing the executable instructions.

9. A computer readable medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, implements the feature standardization method in cross-feature federation according to any one of claims 1 to 6.

Images