Multi-platform e-commerce operation data joint statistical method and system based on privacy calculation

Abstract

The application provides a multi-platform e-commerce operation data joint statistical method and system based on privacy calculation, and relates to the technical field of privacy calculation. The inside sensor and the outside sensor of the box output the inside trigger time and the outside trigger time, and the radio frequency reader outputs the tag data; the processing device establishes the corresponding relationship of the product tag code, the box tag code and the outside trigger time according to the continuous same direction condition of the relative phase increment between tags, corrects the subsequent tag selection by using the phase change of the relationship outside product tag, and makes the product tag code unique by using the subsequent selection result; a plurality of platform servers screen platform inventory records and form secret shares, at least two computing nodes maintained by independent subjects perform security comparison and secret deduplication counting, and the result receiving device only reconstructs the number of boxes and the number of metal canned goods that coincide in the cross-platform inventory occupation period. The application can reduce the identification deviation of piece-by-piece unloading and limit the exposure of platform inventory records.

Description

Technical Field

[0001] This application relates to the field of privacy computing technology, and in particular to a method and system for joint statistical analysis of e-commerce business data across multiple platforms based on privacy computing. Background Technology

[0002] Multi-platform e-commerce inventory statistics typically require linking product identification, packaging information, and inventory holding periods. During the process of receiving entire cases of canned metal goods into the warehouse and then shipping them out piece by piece, box labels identify the packaging boxes, and product labels identify each canned metal item. RFID readers read the tag codes and tag phase values ​​according to the inventory cycle. Accurately matching the shipped products with their respective packaging boxes provides a data foundation for subsequent statistics on boxes and canned metal items with overlapping inventory holding periods across different platforms.

[0003] Regarding the determination of RFID tag movement, patent publication number CN107886018A discloses a method and apparatus for determining the tag movement direction, which obtains in-phase and quadrature signals of the tag relative to two antennas, and determines the tag movement direction based on the changing pattern of phase difference information.

[0004] However, in the scenario of shipping metal canned goods one by one as addressed in this application, the removal of one can will remove the metal obstruction formed by adjacent cans and change the support relationship between cans in the box. Adjacent product labels may be read for the first time in the same shipping period. The short-distance positional change of the cans in the box will also form a phase increment in the same direction as the shipping label, causing a deviation between the product label code and the trigger time on the outside. This deviation will be included in the cross-platform inventory occupancy period statistics along with the platform inventory records. Summary of the Invention

[0005] This application provides a method and system for joint statistical analysis of e-commerce business data across multiple platforms based on privacy computing, in order to solve the problem that the removal of the obscuring of adjacent product labels and changes in the position inside the box caused by the removal of metal canned goods from the box leads to inaccurate correspondence between product label codes and the box removal event, and affects the statistical analysis of overlapping inventory occupancy periods across platforms.

[0006] To address the aforementioned technical problems, firstly, this application provides a method for joint statistical analysis of multi-platform e-commerce operational data based on privacy-preserving computation, comprising: The system obtains the opening time of the box lid, acquires the inner and outer trigger times through sensors inside and outside the box opening, and reads the tag codes and tag phase values ​​of the box label, product label, and reference label through an RFID reader. From the product labels first read after the opening of the box lid or the previous outer trigger time, select product labels whose relative phase increments with the reference label are continuously in the same direction between the inner trigger time and the outer trigger time, and establish the correspondence between the product label code, the box label code and the outer trigger time. The relative phase increment between subsequent labels is corrected by using the relative phase increment between the corresponding product label and the reference label to obtain the product label selection result; When the correspondence contains multiple product label codes, delete the product label codes that are still being read but not selected based on the subsequent product label selection results, until the product label codes are unique; if the correspondence is still not unique after no more product labels are read inside the packaging box, delete the correspondence. Each platform filters platform inventory records according to the product tag codes in the corresponding relationship, and then secretly shares the platform inventory records after associating them with the corresponding relationship to obtain a secret share; Secure multi-party computation is performed on the secret share. Based on the secure comparison results indicating that the inventory occupancy periods of different platforms overlap, secret deduplication is performed, and only the number of boxes and metal canned goods with overlapping inventory occupancy periods across platforms are reconstructed.

[0007] Optionally, from the product labels first read after the opening time of the box lid or the previous outer trigger time, product labels whose relative phase increments with the reference label are continuously in the same direction between the inner trigger time and the outer trigger time are selected, and a correspondence between the product label code, the box label code, and the outer trigger time is established, including: The radio frequency reader is controlled to read the tag phase value according to the storage cycle using the same antenna and frequency point, and the tag phase value is unwrapped to obtain the unwrapped phase value. The inventory period in which both the product label and the reference label are read is determined as the effective inventory period. The difference between the untangled phase values ​​of the product label and the reference label in each effective inventory period is calculated to obtain the relative phase value between the labels. The difference between the relative phase values ​​of the tags in consecutive valid inventory cycles with consecutive inventory serial numbers is calculated to obtain the relative phase increment between the tags; When the relative phase increment between at least two consecutively obtained tags is positive or negative, the corresponding product tag is selected, and a correspondence is established using the selected product tag code, the box tag code read simultaneously, and the outer trigger time; when a product tag or the reference tag is missed, the relative phase increment between the tags is zero, or the direction of change changes, the number of consecutive items in the same direction is re-accumulated.

[0008] Optionally, the step of correcting the relative phase increment between subsequent labels using the relative phase increment between the product label outside the correspondence and the reference label to obtain the product label selection result includes: Identify product tags that are still being read, not selected, and not uniquely associated with the previous external trigger time, outside of the aforementioned correspondence; The difference between the untangling phase values ​​of each product label and the reference label in each effective inventory cycle is calculated to obtain the relative phase value between each product label. Calculate the difference in relative phase values ​​between the labels of each product label in adjacent effective inventory cycles to obtain the relative phase increment between the labels of each product label; determine the median of the relative phase increment between the labels, and subtract the median from the subsequent relative phase increment between the labels to obtain the corrected relative phase increment between the labels; When the relative phase increment between at least two consecutively obtained corrected labels is either positive or negative, the corresponding product label is determined as the selected product label, and the product label selection result is obtained.

[0009] Optionally, when the correspondence contains multiple product label codes, deleting product label codes that are still being read but not selected based on subsequent product label selection results, until the product label codes are unique, and deleting the correspondence when the product label codes are still not unique after no more product labels are read inside the packaging box, includes: When the correspondence includes multiple product label codes, store the multiple product label codes, the corresponding box label codes, and the external trigger time; At subsequent external triggering times, based on the product label selection results, delete product label codes that are still being read but not selected, move the selected product label codes into the corresponding relationship for the subsequent external triggering times, and retain product label codes that are no longer being read. The product tag selection result at subsequent external trigger times is reused to process the stored product tag code until the stored product tag code is unique, or the number of inventory cycles in which product tags are not read consecutively is greater than the maximum number of consecutive missed reads of product tags in the static box. When the stored product label code is unique, the unique product label code, the stored box label code, and the external trigger time are used to form the correspondence containing only one product label code; when the number of inventory cycles in which product labels are not read consecutively is greater than the maximum number of consecutive missed reads and the stored product label code is still not unique, the correspondence is deleted.

[0010] Optionally, each platform filters platform inventory records according to the product tag codes in the correspondence, associates the platform inventory records with the correspondence, and secretly shares them to obtain a secret share, including: Extract the product label code and box label code from the correspondence containing only one product label code; Each platform filters out platform inventory records with the same product label code based on the extracted product label code; Associate the platform identifier and inventory holding period in the filtered platform inventory records with the extracted product label code and box label code; The platform identifier, product label code, and box label code are converted into non-zero integers, and the start and end times of the inventory occupancy period are converted into integers according to a unified time base. The additive secret sharing method is used to split each integer obtained by the transformation into at least two secret shares, such that the sum of the secret shares of the same integer under a common modulus is equal to the corresponding integer; Different secret shares of the same integer are sent to different computing nodes maintained by independent entities to form secret shares for the computing nodes to perform secure multi-party computation; neither the platform server nor the result receiver obtains the entire secret share of the same integer.

[0011] Optionally, unwrapping the tag phase value to obtain the unwrapped phase value includes: When the actual phase change of adjacent effective inventory cycles is less than half a phase cycle, when the tag phase value crosses the phase cycle boundary, one phase cycle is added or subtracted along the continuous direction of phase change, and the adjusted tag phase value is used as the unwrapping phase value; when the product tag or the reference tag is missed, the corresponding inventory cycle is excluded and the continuous same-direction quantity is re-accumulated.

[0012] Optionally, each platform filters platform inventory records with the same product tag code based on the extracted product tag code, including: Each platform records the product label code, platform identifier, and start time when the inventory begins to be used, and writes the end time when the inventory stops being used, forming a platform inventory record. The platform inventory record is then selected from the platform inventory record to find the platform inventory record whose product label code is the same as the extracted product label code.

[0013] Optionally, the step of comparing the secret share using secure multi-party computation includes: A security equivalence comparison is performed on the secret shares of the box label code and the product label code respectively to obtain platform inventory record pairs with the same box label code and platform inventory record pairs with the same product label code; A security comparison is performed on the platform identifier and inventory occupancy time endpoint of the two types of platform inventory record pairs respectively. A secret AND operation is performed on the obtained comparison results to obtain a security comparison result that represents the overlap of inventory occupancy time of different platforms and corresponds to the box label code and the product label code respectively.

[0014] Optionally, the step of performing secret deduplication counting based on the secure comparison results indicating overlapping inventory holding periods across different platforms, and reconstructing only the number of boxes and metal canned goods with overlapping inventory holding periods across platforms, includes: The box label code secret share is selected based on the security comparison result corresponding to the box label code, and the product label code secret share is selected based on the security comparison result corresponding to the product label code. The selected box label code secret share and product label code secret share are securely sorted, adjacent codes are compared and the number of different codes is accumulated, the deduplicated number of box label codes is reconstructed as the number of boxes with overlapping inventory occupation time periods across platforms, and the deduplicated number of product label codes is reconstructed as the number of metal canned products with overlapping inventory occupation time periods across platforms.

[0015] Secondly, this application provides a multi-platform e-commerce business data joint statistical system based on privacy computing, including: The inner and outer sensors of the can opening are used to output the inner and outer trigger times to the processing equipment when the metal canned goods pass through the inner and outer sides of the can opening, respectively. Radio frequency (RFID) readers, connected to processing equipment, are used to output tag codes and tag phase values ​​for box tags, product tags, and reference tags to the processing equipment. The processing equipment is used to obtain the opening time of the box lid, and determine the label code of the outgoing goods based on the opening time of the box lid, the inner trigger time, the outer trigger time, the label code and the label phase value, forming a correspondence between the product label code, the box label code and the outer trigger time, which contains only one product label code. Several platform servers are connected to the processing device and store the platform inventory records of the corresponding platform. They are used to filter the platform inventory records according to the product tag codes in the corresponding relationship, associate the filtered platform inventory records with the corresponding relationship, and generate secret shares that are sent to at least two computing nodes respectively using additive secret sharing. Each computing node is connected to the respective platform server and maintained by an independent entity. Different secret shares are used to receive the same data. Secure multi-party computation is used to determine the box label code and product label code that overlap in the inventory occupancy period of different platforms, and deduplication count shares are generated respectively. The result receiving device, connected to each computing node, is used to reconstruct the number of boxes and metal canned goods that overlap in the cross-platform inventory occupancy period based on the deduplication count share.

[0016] The technical solution provided in this application has the following beneficial effects: This application establishes a correspondence between tags based on the relative phase increment between the trigger times on the inside and outside of the box opening. It uses the phase change of the product tags outside the correspondence to correct the selection of subsequent product tags, and uses subsequent box-out events to make the product tag code uniquely correspond to the box-out event. The sensors inside and outside the box opening, the RFID reader, the processing equipment, the platform server, the computing node, and the result receiving equipment sequentially complete the trigger time acquisition, tag data reading, box-out product determination, local inventory screening, dense state statistics, and quantity reconstruction. This allows the unique correspondence to be directly entered into the cross-platform inventory occupancy period overlap statistics, which can reduce the statistical deviation caused by the removal of obstruction and changes in the position inside the box, and limit the exposure of complete platform inventory records between different hardware. 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 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 example diagram of the hardware structure for item-by-item out-of-box statistics provided in an embodiment of this application.

[0019] Figure 2 A flowchart illustrating the multi-platform e-commerce business data joint statistical method based on privacy computing provided in this application embodiment.

[0020] Figure 3 A flowchart illustrating the multi-product label encoding process provided in this application embodiment.

[0021] Figure 4 Hardware block diagram of a multi-platform e-commerce business data joint statistics system based on privacy computing provided in this application embodiment. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In multi-platform e-commerce operations, the inventory occupancy records of metal canned goods within the same packaging box can be generated separately by different platforms. During the unpacking process, the joint statistics first determine the product label code and box label code corresponding to each shipment. Then, based on the determined label codes, the inventory occupancy records of each platform are filtered, and boxes and metal canned goods with overlapping inventory occupancy periods across different platforms are statistically analyzed. Label correspondence discrepancies generated during the shipment identification stage will be included in subsequent joint statistics along with the filtering results. Therefore, this embodiment focuses on the continuous process of the same packaging box from opening the lid, unpacking items one by one, to emptying the box.

[0024] Taking a packaging box containing twelve metal cans of goods as an example, the cans inside the box are in contact with each other, and the product labels located between adjacent cans are obscured by the metal. When the operator opens the box lid and removes the goods one by one, the RFID reader continuously reads the box label, the product label, and the reference label. After one product is removed, the adjacent product label may change from being unread to being read, and adjacent cans may also roll, tilt, slide, or fall back due to changes in their support positions. The aforementioned changes in the position inside the box may coincide with the time when the goods pass through the box opening, causing the product labels that are actually removed and those that remain inside the packaging box to show the same phase change in direction. This embodiment needs to distinguish between the product labels that actually pass through the box opening and the product labels that are read due to the removal of obstruction or changes in the position inside the box.

[0025] To collect the timing and tag data during the unpacking process, such as Figure 1 As shown, the sensors inside the box opening, the sensors outside the box opening, and the RFID reader are connected to the processing device, the processing device is connected to each platform server, each platform server is connected to at least two computing nodes, and each computing node is connected to the result receiving device. Figure 1 The data flow of a single platform is illustrated using one platform server and two compute nodes. Multiple platform servers are configured as follows: Figure 1 Participate in joint statistics as shown; Figure 1 The result recipient receives the statistical results through the result receiving device. The time of lid opening serves as an existing input for the processing device. Figure 1 Not shown repeatedly. The box label is set on the packaging box, the product label is set on the metal canned product, and the reference label is fixed at the unpacking station; the inner trigger time and the outer trigger time are generated when the metal canned product passes through the inner side and the outer side of the box opening, respectively. The process of the metal canned product triggering the two sensors in sequence constitutes a box unpacking event.

[0026] To address the aforementioned phase changes, this embodiment utilizes two sensors to define each out-of-box event. A fixed reference label is used to establish the relative phase change of the product labels. Then, the median phase change of the product labels still inside the packaging box relative to the reference label is used to correct the relative phase increment between labels in subsequent reading periods. When multiple product labels are selected in a single out-of-box event, the processing device retains multiple product label codes and successively excludes product label codes still inside the box in subsequent out-of-box events. After the product label code is unique, it then enters the inventory data filtering and secure multi-party calculation process for each platform, ensuring continuous and consistent statistical correlation between the item-by-item out-of-box results and cross-platform inventory occupancy periods.

[0027] Based on the above scenarios, such as Figure 2 As shown, the method is implemented in the order of tag data acquisition, outbound product identification, multiple product tag coding processing, inventory data filtering and security statistics, specifically including steps S101 to S106.

[0028] In step S101, the RFID reader reads three types of tags using the same antenna and the same frequency, ensuring that the tag phase values ​​within the same inventory cycle originate from the same read / write link. Two sensors are sequentially positioned along the direction of product removal, with the distance between them such that at least three complete inventory cycles are included between the inner and outer trigger times to obtain the relative phase increment between the two tags. The inventory cycle is the time interval between the start of one tag read and the start of the next tag read by the RFID reader.

[0029] The inventory cycle is determined based on the maximum time required for the RFID reader to complete one inventory of tags within the box, the minimum time required for goods to move from the inner sensor to the outer sensor at the fastest permissible box exit speed, and the phase unwrapping continuity setting. During setup, the aforementioned maximum and minimum times are measured, and the inventory cycle is shortened or the distance between the two sensors is increased to ensure that at least three complete inventory cycles exist between two trigger moments, and that the actual phase change between adjacent inventory cycles at the fastest permissible box exit speed is less than half a phase cycle. These conditions are directly determined by the read / write timing and station dimensions, without setting a separate motion judgment threshold.

[0030] After completing the cycle settings, upon the first box shipment, the processing device uses the box lid opening moment as the starting point for reading the product label; for subsequent box shipments, the trigger moment from the previous outer side is used as the starting point. When the product label code appears in the inventory results for the first time after the reading starting point, the processing device identifies the corresponding product label as the first product label read. When both the product label and the reference label obtain label phase values ​​in the same inventory cycle, the corresponding inventory cycle is identified as a valid inventory cycle; only two consecutive valid inventory cycles constitute adjacent valid inventory cycles. If any label is missed, the missing label phase value is not interpolated, the corresponding inventory cycle is excluded, and the consecutive same-direction quantities are re-accumulated.

[0031] After obtaining the tag phase values ​​for each valid inventory cycle, step S102 can employ the existing phase unwrapping method to handle jumps in tag phase values ​​that cross phase cycle boundaries. When the tag phase value is expressed in radians, one phase cycle is twice pi. Under the condition that the actual phase change between adjacent valid inventory cycles is less than half a phase cycle, when the tag phase value crosses the phase cycle boundary, it is increased or decreased by one phase cycle along the continuous direction of phase change to obtain continuous unwrapped phase values. If adjacent valid inventory cycles are discontinuous, phase unwrapping and continuous counting in the same direction restart.

[0032] Based on the unwrapping phase value, the processing device subtracts the unwrapping phase value of the reference tag from the unwrapping phase value of the product tag within the same inventory cycle to obtain the relative phase value between tags. This relative phase value represents the phase difference between the product tag and the reference tag within the same inventory cycle. Then, the relative phase value between tags in the current inventory cycle is subtracted from the relative phase value in the previous inventory cycle to obtain the relative phase increment between tags. This relative phase increment represents the change in the phase difference between the product tag and the reference tag between adjacent inventory cycles. Since the reference tag is fixed at the unpacking station, the phase change of the reference tag can be used to mitigate common changes generated within the same read / write link.

[0033] Continuous same-direction determination requires at least two relative phase increments between tags, meaning at least three consecutive valid inventory cycles with the same inventory number are needed. When the relative phase increment between tags is zero, it is not classified as positive or negative. If a product tag or reference tag is missed, the relative phase increment between tags is zero, or the direction of change changes, the processing device re-accumulates the number of consecutive same-direction items. Single-direction changes caused by occasional phase noise do not result in a continuous same-direction outcome.

[0034] According to the above quantity settings, in the current outgoing event, when the relative phase increments between at least two consecutively obtained tags are both positive or both are negative, it is called that the relative phase increments between tags are continuously in the same direction, and the processing device selects the corresponding product tag. When a product tag is selected, the product tag code, the simultaneously read box tag code, and the external trigger time are written into the same data record, and the data record is called the correspondence relationship; when at least two product tags are selected, each product tag code is associated with the same box tag code and the external trigger time and stored in the correspondence relationship. Thus, the correspondence relationship obtained in step S102 becomes the basis for distinguishing product tags inside and outside the relationship in step S103.

[0035] Corresponding to the first unpacking event of the packaging box, product A is removed from the box, triggering two sensors in sequence. The label of product B is unobstructed due to the removal of product A, and product B simultaneously undergoes a short-distance roll. Both product A and product B generate positive relative phase increments between their labels within two consecutive inventory cycles. The processing device establishes a correspondence between the product label codes of product A and product B, the box label code, and the time of the first external trigger. Other product labels that are still being read but not yet selected are located outside this correspondence.

[0036] Since multiple product tag codes are obtained from the first unpacking event, step S103 identifies product tags that are still being read, not selected, and have not yet established a unique correspondence with the previous external trigger time, from outside the correspondence. The unpacking station allows the product to leave the effective reading area of ​​the RFID reader after passing through the external sensor; therefore, the aforementioned product tags correspond to metal canned goods still located inside the packaging box. The processing equipment calculates the difference in unwrapped phase values ​​between each product tag and the reference tag in each effective inventory cycle to obtain the relative phase value between tags. Then, it calculates the difference in relative phase values ​​between tags in adjacent effective inventory cycles to obtain the relative phase increment between each product tag.

[0037] The processing equipment sorts the relative phase increments between tags obtained from the same pair of adjacent valid inventory cycles. If the number of tags is odd, the median is taken; if the number is even, the average of the two median values ​​is taken to obtain the median. The median comes from product tags that are still inside the box but not corresponding to the original tag. Both the median and the value to be corrected use the reference tag as the phase reference, representing the change in the box's position caused by the remaining cans after this shipment. The processing equipment subtracts the median from the subsequent relative phase increments between tags to obtain the corrected relative phase increments between tags. Then, it obtains the product tag selection result according to the condition of continuous same direction. If there is only one product tag that meets the condition outside the original tag, the relative phase increment between the corresponding product tag is directly used; if there are no product tags that meet the condition, the correction value is zero.

[0038] After the above corrections, when product B is subsequently removed, the processing device obtains the relative phase increment between product B and the reference label in each adjacent valid inventory cycle, and subtracts the median of the same pair of adjacent valid inventory cycles. When the relative phase increment between the labels after two consecutive corrections is positive, product B is included in the current product label selection result. This result is used in step S104 to process the two product label codes saved in the first out-of-box event.

[0039] Based on the selection results of product tags, such as Figure 3As shown, if a stored product label code is read again in a subsequent inventory cycle but not selected by the product label selection result, the processing device deletes the product label code from the original correspondence. If a stored product label code is selected by a subsequent product label selection result, the product label code is moved into the correspondence corresponding to the subsequent external trigger time and associated with the same box label code. If a stored product label code is no longer read, the product label code is retained in the original correspondence. The processing device repeats the above process at each subsequent external trigger time until the product label code in the original correspondence is unique.

[0040] Therefore, in this embodiment, after product B is selected in the second out-of-box event, it is moved into the correspondence corresponding to the second external trigger time. The correspondence corresponding to the first out-of-box event only retains the product label code of product A. The product label code of product A, the box label code, and the first external trigger time thus form a correspondence containing only one product label code. The result of the second product label selection is directly used to determine the product label of the first out-of-box event.

[0041] The confirmation cycle for no longer reading product labels inside the packaging box is determined based on the measured maximum number of consecutive missed reads of product labels within the stationary box. During setup, with the products stationary and the packaging box at the unpacking station, continuous inventory is conducted, recording the number of consecutive missed reads for each product label, and the maximum number of consecutive missed reads is recorded. During actual unpacking, if the number of inventory cycles with consecutively unread product labels exceeds the maximum number of consecutive missed reads, it is determined that product labels inside the packaging box will no longer be read. If product labels inside the packaging box are no longer being read, but the original correspondence still contains multiple product label codes, the processing equipment deletes the correspondence, preventing the corresponding product label codes from entering subsequent inventory data filtering.

[0042] After obtaining the mapping relationship containing only one product label code, in step S105, each platform records the product label code, platform identifier, and start time when the inventory begins to be occupied, and writes the end time when the inventory stops being occupied, thus associating the platform identifier, start time, and end time of the same product label code to form a platform inventory record. The processing device extracts the product label code and box label code from the mapping relationship containing only one product label code. Each platform filters platform inventory records with the same product label code locally based on the extracted product label code, and associates the platform identifier and inventory occupation period in the filtered records with the product label code and box label code.

[0043] To ensure data from different platforms enters the same calculation process, each platform uses a unified encoding rule to convert platform identifiers, product label codes, and box label codes into non-zero integers, and a unified time base to convert the start and end times of inventory holding periods into integers. When the label code is a character, it can be converted into an integer according to the byte sequence of the unified character encoding; the same label code is converted into the same integer across platforms. The unified time base can be 00:00 on the statistical date, with the time unit being seconds, ensuring that the endpoints of inventory holding periods across different platforms have the same unit.

[0044] After integer conversion, the existing additive secret sharing method is used to split each integer into at least two values, each called a secret share. The sum of the secret shares of the same integer under a common modulus equals the corresponding integer. The common modulus is a prime number greater than all integer codes and determined according to the largest integer code. Different computing nodes are maintained by independent entities, and different secret shares of the same integer are sent to different computing nodes. No platform server, individual computing node, or result receiver obtains all secret shares of the same integer, and at least one computing node does not exchange its secret shares with other computing nodes. For example, under a common modulus of 257, the integer 73 can be split into secret shares 120 and 210. The remainder of their sum 330 divided by 257 is 73. The first computing node holds only 120, the second computing node holds only 210, and neither computing node can obtain the integer 73 from a single secret share.

[0045] After receiving the secret share, each computing node, in step S106, first performs a security equivalence comparison on the secret shares of the box label codes to obtain platform inventory record pairs with the same box label codes; then, it performs a security equivalence comparison on the secret shares of the product label codes to obtain platform inventory record pairs with the same product label codes. For each type of platform inventory record pair, the computing node compares whether the platform identifiers are different, and compares whether the start time of one platform inventory record is earlier than the end time of another platform inventory record, and whether the start time of another platform inventory record is earlier than the end time of the previous platform inventory record, thereby determining whether there is any overlap in the inventory occupancy periods of different platforms for the same box label code or product label code.

[0046] When determining whether inventory holding periods overlap, the inventory holding period is treated as an interval including the start time but excluding the end time; inventory holding periods with two endpoints connected are not considered overlapping. The computing node performs a secret AND operation on the comparison results of different platform identifiers and the comparison results of two times, respectively obtaining secure comparison results corresponding to the box label code and the product label code. The secure comparison result remains a secret share, without reconstructing the platform identifier, inventory holding period, or label code. For example, the following values ​​only represent the comparison relationship corresponding to the secret share; the actual calculation does not reconstruct the platform identifier and time. When platform identifier 1 and platform identifier 2 are different, and the two inventory holding periods are [100, 160) and [140, 220) respectively, 100 is earlier than 220 and 140 is earlier than 160, the secure comparison result indicates that the two inventory holding periods overlap; when the start time of the second inventory holding period is 160, 160 is not earlier than the end time 160 of the first inventory holding period, the secure comparison result indicates that the two inventory holding periods do not overlap.

[0047] After obtaining the two types of security comparison results, the computing node selects the box label code secret share based on the security comparison result indicating overlapping inventory holding periods across different platforms and corresponding to the box label code, and selects the product label code secret share based on the security comparison result corresponding to the product label code. Unselected code secret shares are set to zero. Therefore, subsequent counting objects are limited to box label codes and product label codes that participate in at least one set of records with overlapping inventory holding periods across different platforms. Valid label codes are all converted to non-zero integers, thus distinguishing them from unselected zero values. When performing security sorting on the two types of code secret shares, an existing bitonic sorting network can be used to arrange identical codes adjacently. The computing node skips zero values, compares adjacent codes, and increments the secret share by one when codes are different, obtaining the deduplicated quantity secret shares of box label codes and product label codes for overlapping inventory holding periods across platforms. For example, if the integers corresponding to the box label codes are 7, 7, 7, and another unselected record is set to zero, the safe sorting order is 0, 7, 7, 7. After skipping the zero value, the deduplicated quantity is 1. If the integers corresponding to the product label codes are 23, 23, 41, and another unselected record is set to zero, the safe sorting order is 0, 23, 23, 41. After skipping the zero value, the deduplicated quantity is 2. The receiver can then reconstruct the quantity of boxes (1) and the quantity of metal canned goods (2) that overlap in inventory usage time across platforms.

[0048] After completing the secret deduplication count, each computing node sends the secret share of the two types of deduplication quantities to the result recipient. The result recipient only reconstructs the deduplication quantities of box label codes and product label codes that overlap in cross-platform inventory occupancy periods, which are respectively used as the quantity of boxes and metal canned goods that overlap in cross-platform inventory occupancy periods. During the processing of packaging boxes, the platform inventory record corresponding to product A participates in the above calculation after product A forms a unique correspondence with the first external trigger time. Therefore, the phase selection result, the processing results of multiple product label codes, the platform inventory record, and the cross-platform overlapping occupancy statistics are generated sequentially along the same processing chain.

[0049] This application also provides a multi-platform e-commerce business data joint statistical system based on privacy computing. For example... Figure 4 As shown, the system is in Figure 1 Based on the hardware connection relationship, several platform servers and at least two computing nodes are deployed, including sensors inside the box opening, sensors outside the box opening, radio frequency readers, processing equipment, several platform servers, at least two computing nodes, and result receiving equipment. Figure 1 The processing device shown is the same as the processing device in this embodiment. Figure 1 The platform server shown represents any platform server. Figure 1 The first and second computing nodes shown represent two of at least two computing nodes. Figure 1 The recipient of the results receives the statistical results through the result receiving device.

[0050] The inner and outer sensors of the can opening are respectively located on the inner and outer sides of the can opening. When the metal canned goods pass through the two sensors in sequence, the inner sensor outputs the inner trigger time to the processing device, and the outer sensor outputs the outer trigger time to the processing device. The RFID reader is connected to the processing device and outputs the tag codes and tag phase values ​​of the can label, product label, and reference label to the processing device.

[0051] The processing device includes a processor and a memory. It acquires the time when the lid opens and unwraps the received tag phase value to form the relative phase increment between the product tag and the reference tag. The processing device combines the inner trigger time, the outer trigger time and the subsequent box-out event to determine the product tag code. After the product tag code is unique, it outputs the correspondence between the product tag code, the box tag code and the outer trigger time, which contains only one product tag code, to each platform server.

[0052] Several platform servers correspond to different e-commerce platforms. Each platform server stores the platform inventory records for its respective platform and is connected to the processing device. Each platform server receives the correspondence from the processing device, filters the platform inventory records locally according to the product tag codes in the correspondence, and retains the platform inventory records that are not filtered in the corresponding platform server.

[0053] Each platform server associates the selected platform inventory records with their corresponding relationships, and converts the platform identifier, inventory occupancy period endpoint, product label code, and container label code into integers. Then, using additive secret sharing, each integer is split into at least two secret shares, and the different secret shares of the same integer are sent to different computing nodes. The platform servers do not exchange plaintext platform inventory records with each other.

[0054] At least two computing nodes are connected to each platform server and maintained by independent entities. Each computing node performs tag code security equivalence comparison, platform identifier security comparison, inventory occupancy period overlap comparison, and secret deduplication counting based on the received secret share to obtain the box tag code deduplication count share and the product tag code deduplication count share; no single computing node obtains the entire secret share of the same data.

[0055] The result receiving device is connected to each computing node and is used to receive the deduplication count share of container label codes and product label codes sent by each computing node. It then reconstructs the quantity of containers and metal canned goods with overlapping inventory occupancy periods across platforms. The result receiving device does not reconstruct platform identifiers, inventory occupancy periods, container label codes, or product label codes.

[0056] Therefore, the sensors inside and outside the container opening only output the trigger time, the RFID reader only outputs the tag data, the processing device only outputs the corresponding relationship to each platform server, each platform server only outputs the secret share to each computing node, and each computing node only outputs the deduplication count share to the result receiving device. Figure 4 according to Figure 1 The data stream shown expands each independent hardware component to perform real-time data acquisition, tag reading, outbound product identification, local inventory filtering, dense state statistics, and quantity reconstruction.

[0057] The foregoing has provided a detailed description of a multi-platform e-commerce operational data joint statistical method and system based on privacy computing provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. Several improvements and modifications can be made to this application without departing from its principles, and all improvements and modifications fall within the protection scope of this application.

Claims

1. A multi-platform e-commerce operational data joint statistical method based on privacy computing, used for item-by-item outbound statistical analysis of metal canned goods, characterized in that... include: The system obtains the opening time of the box lid, acquires the inner and outer trigger times through sensors inside and outside the box opening, and reads the tag codes and tag phase values ​​of the box label, product label, and reference label through an RFID reader. From the product labels first read after the opening time of the box lid or the previous outer trigger time, select product labels whose relative phase increment between the inner trigger time and the outer trigger time is continuously in the same direction as the reference label, and establish the correspondence between the product label code, the box label code and the outer trigger time. The relative phase increment between subsequent labels is corrected by using the relative phase increment between the product label outside the correspondence and the reference label to obtain the product label selection result; When the correspondence contains multiple product label codes, delete the product label codes that are still being read but not selected based on the subsequent product label selection results, until the product label codes are unique; if the correspondence is still not unique after no more product labels are read in the packaging box, delete the correspondence. Each platform filters platform inventory records according to the product tag codes in the corresponding relationship, and then secretly shares the platform inventory records after associating them with the corresponding relationship to obtain a secret share; Secure multi-party computation is performed on the secret share. Based on the secure comparison results indicating that the inventory occupancy periods of different platforms overlap, secret deduplication is performed, and only the number of boxes and metal canned goods with overlapping inventory occupancy periods across platforms are reconstructed.

2. The method according to claim 1, characterized in that, From the product labels first read after the opening of the box lid or the previous outer triggering time, select product labels whose relative phase increments with the reference label are continuously in the same direction between the inner triggering time and the outer triggering time, and establish a correspondence between the product label code, the box label code, and the outer triggering time, including: The radio frequency reader is controlled to read the tag phase value according to the storage cycle using the same antenna and frequency point, and the tag phase value is unwrapped to obtain the unwrapped phase value. The inventory period in which both the product label and the reference label are read is determined as the effective inventory period. The difference between the untangled phase values ​​of the product label and the reference label in each effective inventory period is calculated to obtain the relative phase value between the labels. The difference between the relative phase values ​​of the tags in consecutive valid inventory cycles with consecutive inventory serial numbers is calculated to obtain the relative phase increment between the tags; When the relative phase increment between at least two consecutively obtained tags is positive or negative, the corresponding product tag is selected, and a correspondence is established using the selected product tag code, the box tag code read simultaneously, and the outer trigger time; when a product tag or the reference tag is missed, the relative phase increment between the tags is zero, or the direction of change changes, the number of consecutive items in the same direction is re-accumulated.

3. The method according to claim 2, characterized in that, The step of correcting the relative phase increment between subsequent labels using the relative phase increment between the corresponding product label and the reference label to obtain the product label selection result includes: Identify product tags that are still being read, not selected, and not uniquely associated with the previous external trigger time, outside of the aforementioned correspondence; The difference between the untangling phase values ​​of each product label and the reference label in each effective inventory cycle is calculated to obtain the relative phase value between each product label. Calculate the difference in relative phase values ​​between the labels of each product label in adjacent effective inventory cycles to obtain the relative phase increment between the labels of each product label; determine the median of the relative phase increment between the labels, and subtract the median from the subsequent relative phase increment between the labels to obtain the corrected relative phase increment between the labels; When the relative phase increments between at least two consecutively obtained corrected labels are both positive or both are negative, the corresponding product label is determined as the selected product label, and the product label selection result is obtained.

4. The method according to claim 3, characterized in that, When the correspondence contains multiple product label codes, the product label codes that are still being read but not selected are deleted based on the subsequent product label selection results, until the product label codes are unique, and the correspondence is deleted when the product label codes are still not unique after no more product labels are read inside the packaging box. This includes: When the correspondence includes multiple product label codes, store the multiple product label codes, the corresponding box label codes, and the external trigger time; At subsequent external triggering times, based on the product label selection results, delete product label codes that are still being read but not selected, move the selected product label codes into the corresponding relationship for the subsequent external triggering times, and retain product label codes that are no longer being read. The product tag selection result at subsequent external trigger times is reused to process the stored product tag code until the stored product tag code is unique, or the number of inventory cycles in which product tags are not read consecutively is greater than the maximum number of consecutive missed reads of product tags in the static box. When the stored product label code is unique, the unique product label code, the stored box label code, and the external trigger time are used to form the correspondence containing only one product label code; when the number of inventory cycles in which product labels are not read consecutively is greater than the maximum number of consecutive missed reads and the stored product label code is still not unique, the correspondence is deleted.

5. The method according to claim 4, characterized in that, Each platform filters its inventory records according to the product tag codes in the corresponding relationship, associates the platform inventory records with the corresponding relationship, and secretly shares them to obtain a secret share, including: Extract the product label code and box label code from the correspondence containing only one product label code; Each platform filters out platform inventory records with the same product label code based on the extracted product label code; Associate the platform identifier and inventory holding period in the filtered platform inventory records with the extracted product label code and box label code; The platform identifier, product label code, and box label code are converted into non-zero integers, and the start and end times of the inventory occupancy period are converted into integers according to a unified time base. The additive secret sharing method is used to split each integer obtained by the transformation into at least two secret shares, such that the sum of the secret shares of the same integer under a common modulus is equal to the corresponding integer; Different secret shares of the same integer are sent to different computing nodes maintained by independent entities to form secret shares for the computing nodes to perform secure multi-party computation; neither the platform server nor the result receiver obtains the entire secret share of the same integer.

6. The method according to claim 2, characterized in that, The step of unwrapping the tag phase value to obtain the unwrapped phase value includes: When the actual phase change of adjacent effective inventory cycles is less than half a phase cycle, when the tag phase value crosses the phase cycle boundary, one phase cycle is added or subtracted along the continuous direction of phase change, and the adjusted tag phase value is used as the unwrapping phase value; when the product tag or the reference tag is missed, the corresponding inventory cycle is excluded and the continuous same-direction quantity is re-accumulated.

7. The method according to claim 5, characterized in that, Each platform filters platform inventory records with the same product tag code based on the extracted product tag code, including: Each platform records the product label code, platform identifier, and start time when the inventory begins to be used, and writes the end time when the inventory stops being used, forming a platform inventory record. The platform inventory record is then selected from the platform inventory record to find the platform inventory record whose product label code is the same as the extracted product label code.

8. The method according to claim 7, characterized in that, The method of comparing the secret share using secure multi-party computation includes: A security equivalence comparison is performed on the secret shares of the box label code and the product label code respectively to obtain platform inventory record pairs with the same box label code and platform inventory record pairs with the same product label code; A security comparison is performed on the platform identifier and inventory occupancy time endpoint of the two types of platform inventory record pairs respectively. A secret AND operation is performed on the obtained comparison results to obtain a security comparison result that represents the overlap of inventory occupancy time of different platforms and corresponds to the box label code and the product label code respectively.

9. The method according to claim 1, characterized in that, The method of performing secret deduplication counting based on the secure comparison results indicating overlapping inventory holding periods across different platforms, reconstructing only the number of boxes and metal canned goods with overlapping inventory holding periods across platforms, includes: The box label code secret share is selected based on the security comparison result corresponding to the box label code, and the product label code secret share is selected based on the security comparison result corresponding to the product label code. The selected box label code secret share and product label code secret share are securely sorted, adjacent codes are compared and the number of different codes is accumulated, the deduplicated number of box label codes is reconstructed as the number of boxes with overlapping inventory occupation time periods across platforms, and the deduplicated number of product label codes is reconstructed as the number of metal canned products with overlapping inventory occupation time periods across platforms.

10. A multi-platform e-commerce operational data joint statistical system based on privacy computing, used for item-by-item outbound statistical analysis of metal canned goods, characterized in that... include: The inner and outer sensors of the can opening are used to output the inner and outer trigger times to the processing equipment when the metal canned goods pass through the inner and outer sides of the can opening, respectively. An RFID reader, connected to the processing device, is used to output the tag codes and tag phase values ​​of the box tag, product tag, and reference tag to the processing device; The processing device is used to obtain the opening time of the box lid, and determine the label code of the outgoing goods based on the opening time of the box lid, the inner trigger time, the outer trigger time, the label code and the label phase value, forming a correspondence between the product label code, the box label code and the outer trigger time, which contains only one product label code. Several platform servers are connected to the processing device and store platform inventory records for their respective platforms. These servers are used to filter platform inventory records according to the product tag codes in the correspondence relationship, associate the filtered platform inventory records with the correspondence relationship, and generate secret shares that are sent to at least two computing nodes using additive secret sharing. Each computing node is connected to the respective platform server and maintained by an independent entity. It is used to receive different secret shares of the same data. Secure multi-party computation is used to determine the box label code and product label code that overlap in the inventory occupancy period of different platforms, and generate deduplication count shares respectively. The result receiving device, connected to each of the computing nodes, is used to reconstruct the number of boxes and metal canned goods that overlap in the cross-platform inventory occupancy period based on the deduplication count share.

Images