System, method, and computer program product for masking model input data

Abstract

Systems, methods, and computer program products are provided for masking model input data. The system includes at least one processor configured to receive a data set comprising a plurality of records, each record comprising a plurality of parameters, transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record, transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record, compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records, and input the modified data set into a model.

Description

Attorney Docket No. 08223-2407813 (8711 WO01 )SYSTEM, METHOD, AND COMPUTER PROGRAM PRODUCT FOR MASKING MODEL INPUT DATACROSS REFERENCE TO RELATED APPLICATION

[0001] This application claims priority to Greek Patent Application No. 20240100621 , filed on September 10, 2024, the disclosure of which is hereby incorporated by reference in its entirety.BACKGROUND1 . Technical Field

[0002] This disclosure relates generally to models and, in non-limiting embodiments or aspects, to systems, methods, and computer program products for masking model input data.2. Technical Considerations

[0003] Using machine-learning models, including generative artificial intelligence (Al) models, presents several challenges with respect to sensitive data. Many data sets include sensitive data such as personal identifiable information (PH), monetary values, account numbers, and / or other data that raises regulatory and / or privacy concerns. This can limit the amount of data and / or information used to prompt a generative Al model, leading to less desirable outputs. Applications such as fraud detection are prone to being negatively affected by a lack of data that can be used to detect fraudulent patterns, as an example.

[0004] Moreover, inputting numerous records into a model, such as a generative Al model, is computationally resource-intensive and results in latency and increased processing costs. Further, a limited amount of data (e.g., a limited number of tokens to be processed) can be input into such models at once.SUMMARY

[0005] According to non-limiting embodiments or aspects, provided is a system for comprising at least one processor configured to: receive a data set comprising a plurality of records, each record comprising a plurality of parameters; transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a5ZN5938.DOCX Page 1 of 27Attorney Docket No. 08223-2407813 (8711WO01) modified data set comprising a plurality of compressed records; and input the modified data set into a model.

[0006] In non-limiting embodiments or aspects, the first type comprises a monetary numerical value, and the first transformation algorithm comprises a cryptographic commitment function. In non-limiting embodiments or aspects, the second type comprises a non-monetary value, and the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof. In non-limiting embodiments or aspects, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values. In non-limiting embodiments or aspects, the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity. In non-limiting embodiments or aspects, the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

[0007] According to non-limiting embodiments or aspects, provided is a computer- implemented method, comprising: receiving a data set comprising a plurality of records, each record comprising a plurality of parameters; transforming at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transforming at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compressing each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and inputting the modified data set into a model.

[0008] In non-limiting embodiments or aspects, the first type comprises a monetary numerical value, and the first transformation algorithm comprises a cryptographic commitment function. In non-limiting embodiments or aspects, the second type comprises a non-monetary value, and the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof. In non-limiting embodiments or aspects, wherein truncating is based on other5ZN5938.DOCX Page 2 of 27Attorney Docket No. 08223-2407813 (8711WO01) values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values. In non-limiting embodiments or aspects, the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity. In non-limiting embodiments or aspects, the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

[0009] According to non-limiting embodiments or aspects, provided is a computer program product comprising at least one non-transitory computer-readable medium including program instructions that, when executed by at least one processor, cause the at least one processor to: receive a data set comprising a plurality of records, each record comprising a plurality of parameters; transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and input the modified data set into a model.

[0010] In non-limiting embodiments or aspects, the first type comprises a monetary numerical value, and the first transformation algorithm comprises a cryptographic commitment function. In non-limiting embodiments or aspects, the second type comprises a non-monetary value, and the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof. In non-limiting embodiments or aspects, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values. In non-limiting embodiments or aspects, the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity. In non-limiting embodiments or aspects, the model comprises a large language model, and wherein5ZN5938.DOCX Page 3 of 27Attorney Docket No. 08223-2407813 (8711WO01) inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

[0011] Further non-limiting embodiments or aspects are set forth in the following numbered clauses:

[0012] Clause 1 : A system for comprising at least one processor configured to: receive a data set comprising a plurality of records, each record comprising a plurality of parameters; transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and input the modified data set into a model.

[0013] Clause 2: The system of clause 1 , wherein the first type comprises a monetary numerical value, and wherein the first transformation algorithm comprises a cryptographic commitment function.

[0014] Clause 3: The system of clause 1 or 2, wherein the second type comprises a non-monetary value, and wherein the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof.

[0015] Clause 4: The system of any of clauses 1 -3, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values.

[0016] Clause 5: The system of any of clauses 1 -4, wherein the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity.

[0017] Clause 6: The system of any of clauses 1 -5, wherein the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.5ZN5938.DOCX Page 4 of 27Attorney Docket No. 08223-2407813 (8711WO01)

[0018] Clause 7: A computer-implemented method, comprising: receiving a data set comprising a plurality of records, each record comprising a plurality of parameters; transforming at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transforming at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compressing each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and inputting the modified data set into a model.

[0019] Clause 8: The method of clause 7, wherein the first type comprises a monetary numerical value, and wherein the first transformation algorithm comprises a cryptographic commitment function.

[0020] Clause 9: The method of clause 7 or 8, wherein the second type comprises a non-monetary value, and wherein the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof.

[0021] Clause 10: The method of any of clauses 7-9, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values.

[0022] Clause 1 1 : The method of any of clauses 7-10, wherein the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity.

[0023] Clause 12: The method of any of clauses 7-11 , wherein the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

[0024] Clause 13: A computer program product comprising at least one non- transitory computer-readable medium including program instructions that, when executed by at least one processor, cause the at least one processor to: receive a data set comprising a plurality of records, each record comprising a plurality of parameters; transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record;5ZN5938.DOCX Page 5 of 27Attorney Docket No. 08223-2407813 (8711WO01) transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and input the modified data set into a model.

[0025] Clause 14: The computer program product of clause 13, wherein the first type comprises a monetary numerical value, and wherein the first transformation algorithm comprises a cryptographic commitment function.

[0026] Clause 15: The computer program product of clause 13 or 14, wherein the second type comprises a non-monetary value, and wherein the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof.

[0027] Clause 16: The computer program product of any of clauses 13-15, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values.

[0028] Clause 17: The computer program product of any of clauses 13-16, wherein the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity.

[0029] Clause 18: The computer program product of any of clauses 13-17, wherein the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

[0030] These and other features and characteristics of the present disclosure, as well as the methods of operation and functions of the related elements of structures and the combination of parts and economies of manufacture, will become more apparent upon consideration of the following description and the appended claims with reference to the accompanying drawings, all of which form a part of this specification, wherein like reference numerals designate corresponding parts in the various figures. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the disclosed subject matter.5ZN5938.DOCX Page 6 of 27Attorney Docket No. 08223-2407813 (8711WO01)BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Additional advantages and details are explained in greater detail below with reference to the non-limiting, exemplary embodiments that are illustrated in the accompanying schematic figures, in which:

[0032] FIG. 1 is a schematic diagram of a system for masking model input data according to some non-limiting embodiments or aspects;

[0033] FIG. 2 is a flow diagram of a method for masking model input data according to some non-limiting embodiments or aspects;

[0034] FIG. 3 is a schematic diagram of an electronic payment processing network used in connection with some non-limiting embodiments or aspects; and

[0035] FIG. 4 is a schematic diagram of example components of one or more devices of FIG. 1 according to some non-limiting embodiments or aspects.DETAILED DESCRIPTION

[0036] For purposes of the description hereinafter, the terms “end,” “upper,” “lower,” “right,” “left,” “vertical,” “horizontal,” “top,” “bottom,” “lateral,” “longitudinal,” and derivatives thereof shall relate to the embodiments as they are oriented in the drawing figures. However, it is to be understood that the present disclosure may assume various alternative variations and step sequences, except where expressly specified to the contrary. It is also to be understood that the specific devices and processes illustrated in the attached drawings, and described in the following specification, are simply exemplary and non-limiting embodiments or aspects of the disclosed subject matter. Hence, specific dimensions and other physical characteristics related to the embodiments or aspects disclosed herein are not to be considered as limiting.

[0037] Some non-limiting embodiments or aspects are described herein in connection with thresholds. As used herein, satisfying a threshold may refer to a value being greater than the threshold, more than the threshold, higher than the threshold, greater than or equal to the threshold, less than the threshold, fewer than the threshold, lower than the threshold, less than or equal to the threshold, equal to the threshold, etc.

[0038] No aspect, component, element, structure, act, step, function, instruction, and / or the like used herein should be construed as critical or essential unless explicitly described as such. Also, as used herein, the articles “a” and “an” are intended to include one or more items and may be used interchangeably with “one or more” and “at least one.” Furthermore, as used herein, the term “set” is intended to include one5ZN5938.DOCX Page 7 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) or more items (e.g., related items, unrelated items, a combination of related and unrelated items, and / or the like) and may be used interchangeably with “one or more” or “at least one.” Where only one item is intended, the term “one” or similar language is used. Also, as used herein, the terms “has,” “have,” “having,” or the like are intended to be open-ended terms. Further, the phrase “based on” is intended to mean “based at least partially on” unless explicitly stated otherwise. In addition, reference to an action being “based on” a condition may refer to the action being “in response to” the condition. For example, the phrases “based on” and “in response to” may, in some non-limiting embodiments or aspects, refer to a condition for automatically triggering an action (e.g., a specific operation of an electronic device, such as a computing device, a processor, and / or the like).

[0039] As used herein, the term “communication” may refer to the reception, receipt, transmission, transfer, provision, and / or the like of data (e.g., information, signals, messages, instructions, commands, and / or the like). For one unit (e.g., a device, a system, a component of a device or system, combinations thereof, and / or the like) to be in communication with another unit means that the one unit is able to directly or indirectly receive information from and / or transmit information to the other unit. This may refer to a direct or indirect connection (e.g., a direct communication connection, an indirect communication connection, and / or the like) that is wired and / or wireless in nature. Additionally, two units may be in communication with each other even though the information transmitted may be modified, processed, relayed, and / or routed between the first and second unit. For example, a first unit may be in communication with a second unit even though the first unit passively receives information and does not actively transmit information to the second unit. As another example, a first unit may be in communication with a second unit if at least one intermediary unit processes information received from the first unit and communicates the processed information to the second unit. In some non-limiting embodiments or aspects, a message may refer to a network packet (e.g., a data packet and / or the like) that includes data. It will be appreciated that numerous other arrangements are possible.

[0040] As used herein, the term “computing device” may refer to one or more electronic devices configured to process data. A computing device may, in some examples, include the necessary components to receive, process, and output data, such as a processor, a display, a memory, an input device, a network interface, and / or5ZN5938.DOCX Page 8 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) the like. A computing device may be a mobile device. As an example, a mobile device may include a cellular phone (e.g., a smartphone or standard cellular phone), a portable computer, a wearable device (e.g., watches, glasses, lenses, clothing, and / or the like), a personal digital assistant (PDA), and / or other like devices. A computing device may also be a desktop computer or other form of non-mobile computer.

[0041] As used herein, the term “server” may refer to or include one or more computing devices that are operated by or facilitate communication and processing for multiple parties in a network environment, such as the Internet, although it will be appreciated that communication may be facilitated over one or more public or private network environments and that various other arrangements are possible. Further, multiple computing devices (e.g., servers, point-of-sale (POS) devices, mobile devices, etc.) directly or indirectly communicating in the network environment may constitute a “system.”

[0042] As used herein, the term “system” may refer to one or more computing devices or combinations of computing devices (e.g., processors, servers, client devices, software applications, components of such, and / or the like). Reference to “a device,” “a server,” “a processor,” and / or the like, as used herein, may refer to a previously-recited device, server, or processor that is recited as performing a previous step or function, a different device, server, or processor, and / or a combination of devices, servers, and / or processors. For example, as used in the specification and the claims, a first device, a first server, or a first processor that is recited as performing a first step or a first function may refer to the same or different device, server, or processor recited as performing a second step or a second function.

[0043] As used herein, the term “issuer institution” may refer to one or more entities, such as a bank, that provide accounts to customers for conducting transactions (e.g., payment transactions), such as initiating credit and / or debit payments. For example, an issuer institution may provide an account identifier, such as a personal account number (PAN), to a customer that uniquely identifies one or more accounts associated with that customer. The account identifier may be embodied on a portable financial device, such as a physical financial instrument, e.g., a payment card, and / or may be electronic and used for electronic payments. The term “issuer system” refers to one or more computer systems operated by or on behalf of an issuer institution, such as a server computer executing one or more software applications. For example, an issuer system may include one or more authorization servers for authorizing a transaction.5ZN5938.DOCX Page 9 of 27Attorney Docket No. 08223-2407813 (8711 WO01 )

[0044] As used herein, the term “acquirer institution” may refer to an entity licensed and / or approved by the transaction service provider to originate transactions (e.g., payment transactions) using a portable financial device associated with the transaction service provider. The transactions the acquirer institution may originate may include payment transactions (e.g., purchases, original credit transactions (OCTs), account funding transactions (AFTs), and / or the like). In some non-limiting embodiments, an acquirer institution may be a financial institution, such as a bank. As used herein, the term “acquirer system” may refer to one or more computer systems, computer devices, software applications, and / or the like operated by or on behalf of an acquirer institution.

[0045] As used herein, the term “merchant” may refer to an individual or entity that provides goods and / or services, or access to goods and / or services, to customers based on a transaction, such as a payment transaction. The term “merchant” or “merchant system” may also refer to one or more computer systems operated by or on behalf of a merchant, such as a server computer executing one or more software applications. A “point-of-sale (POS) system,” as used herein, may refer to one or more computers and / or peripheral devices used by a merchant to engage in payment transactions with customers, including one or more card readers, near-field communication (NFC) receivers, RFID receivers, and / or other contactless transceivers or receivers, contact-based receivers, payment terminals, computers, servers, input devices, and / or other like devices that can be used to initiate a payment transaction.

[0046] As used herein, the term “transaction service provider” may refer to an entity that receives transaction authorization requests from merchants or other entities and provides guarantees of payment, in some cases through an agreement between the transaction service provider and an issuer institution. For example, a transaction service provider may include a payment network such as Visa® or any other entity that processes transactions. The term “transaction processing system” may refer to one or more computer systems operated by or on behalf of a transaction service provider, such as a transaction processing server executing one or more software applications. A transaction processing server may include one or more processors and, in some non-limiting embodiments, may be operated by or on behalf of a transaction service provider.

[0047] Non-limiting embodiments described herein are directed to improvements in data processing and generative Al models. Through combining masking and5ZN5938.DOCX Page 10 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) compression with sensitive data elements, non-limiting embodiments allow for large sets of sensitive data to be used with generative Al models. Non-limiting embodiments allow for dynamic masking to be performed based on a type of parameter, such that different aspects of the data that may be usable for identifying one or more patterns and / or analyzing may be preserved. The use of dynamic masking and the application of multiple transformation algorithms allows the amount of data to be compressed in an efficient manner while still allowing for analysis. Non-limiting embodiments may allow for the use of generative Al models with reduced computational resources and latency based on the masked and compressed nature of the data used. It will be appreciated that various other improvements and advantages are provided by nonlimiting embodiments.

[0048] Referring now to FIG. 1 , shown is a system 1000 for masking model input data according to some non-limiting embodiments or aspects. A computing device 100 may be in communication with a records database 101 , which may include one or more records such as but not limited to transaction records. Each record in the record database 101 may include a plurality of parameters of different types. For example, a record may include numeric parameters, textual parameters, temporal parameters, and / or the like. In non-limiting embodiments, parameter types may include string parameters (e.g., personally identifying information such as personal names, addresses, telephone numbers, as well as merchant names, issuer names, and / or the like), amounts / value numerical parameters (e.g., transaction values, account balances, and / or any other numerical value), score parameters (e.g., risk scores, fraud scores, and / or the like), and miscellaneous parameters (e.g., country codes, zip codes, auxiliary fields, and / or the like). It will be appreciated that various types of records and types of parameters may be used in non-limiting embodiments.

[0049] With continued reference to FIG. 1 , the computing device 100 may be in communication with a compression engine 104 and a masking engine 102. The compression engine 104 and / or masking engine 102 may include one or more computing devices and / or software functions executable by one or more computing devices. In some non-limiting embodiments, the compression engine 104 and / or masking engine 102 may be part of the computing device 100. In some non-limiting embodiments, the compression engine 104 and / or masking engine 102 may be part of a computing system 1 10, such as an internal local area network and / or the like. In some non-limiting embodiments the compression engine 104 and / or masking engine5ZN5938.DOCX Page 1 1 of 27Attorney Docket No. 08223-2407813 (8711 WO01 )102 may be external to the computing device 100 and / or computing system 1 10, such as one or more remote server computers in communication with the computing device 100 through a network.

[0050] In non-limiting embodiments, the masking engine 102 is configured to transform portions of a set of records received from the records database 101. The masking engine 102 may process records individually and / or in batches. The masking engine 102 may apply different transformation functions to different types of parameters. For example, the masking engine 102 may determine the type (e.g., classification, category, etc.) of a parameter in the record and then select a transformation function (e.g., transformation algorithm) that corresponds to that type. There may be a set of rules specifying parameter types associated with transformation functions, as an example.

[0051] In non-limiting embodiments, for string parameters, such as names, addresses, and / or the like, a deterministic masking algorithm may be applied as a transformation function. As an example, string parameters may be hashed with a oneway cryptographic function. In this manner, identical inputs will result in identical masked parameters such that the masked parameters can be matched without knowing the underlying plaintext value. In non-limiting embodiments, such parameters may additionally or alternatively be converted to a hexadecimal representation or another form of data representation. Such conversion may occur before or after the transformation function.

[0052] In non-limiting embodiments, for numerical parameters that represent amounts, such as transaction values and balances, a transformation algorithm may be applied that masks the underlying value but still allows for comparisons and / or mathematical functions. As an example, such numerical parameters may be transformed via a cryptographic commitment such as but not limited to Pedersen commitments using elliptical curves. It will be appreciated that various zero-knowledge processes may be used in non-limiting embodiments to determine if one transformed value is greater or less than another transformed value, without revealing the values themselves. In some non-limiting embodiments, a hash function may be used to mask numerical parameters without preserving the values for comparison.

[0053] In non-limiting embodiments, for score parameters, such as risk scores, a noise-based masking algorithm may be applied in non-limiting embodiments. Risk scores may be provided by a separate risk scoring engine, as an example. In non-5ZN5938.DOCX Page 12 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) limiting embodiments, a parameterized amount of noise may be added to the value to mask the underlying value without having a significant effect on a subsequent analysis by a machine-learning model. As an example, a value of 50 may be transformed to 51 or 49 by adding or subtracting. In non-limiting embodiments, the noise-based masking algorithm may use a random number generator to generate random values to be introduced into the value (e.g., via a mathematical operator such as but not limited to addition, subtraction, multiplication, and / or the like). In some non-limiting embodiments, the amount of noise and / or modification may be less than a percentage of the total value (e.g., 1 %, 5%, 10%, and / or the like) to lessen any impact on accuracy. In some non-limiting embodiments, only some of the score parameters may be transformed with noise and others may be kept as-is. In non-limiting embodiments, for other parameters, such as country codes, zip codes, auxiliary fields, and / or the like, there may be no transformation applied and the values may remain in plaintext.

[0054] In non-limiting embodiments, the records are transformed with these transformation functions by replacing the original value in the record with the transformed value, thus removing the original value and resulting in a modified record. In non-limiting embodiments, the records may be represented by only the parameters, arranged in an array or in any other form of structured data.

[0055] In non-limiting embodiments, all parameters may be classified as a type, even if the parameter is not masked (e.g., a non-masked type). In non-limiting embodiments, such parameters may additionally or alternatively be converted to a hexadecimal representation or another form of data representation.

[0056] In non-limiting embodiments, the compression engine 104 is configured to compress the transformed records (e.g., masked records) to reduce the input size made to a model 106. Since patterns are identified in the records that may not need full data values, the masked parameters (e.g., fields of the record) may be truncated in their respective bitwise representations. Multiple records in a data set may be processed and compressed together, such as but not limited to all transaction data for a time period, transaction data for a specific account identifier, transaction data for a specific account holder, transaction data for a specific merchant, transaction data for a specific issuer, and / or the like. It will also be appreciated that other types of records, such as but not limited to medical records, may be compressed.

[0057] In non-limiting embodiments, the compression may be performed in a manner that avoids collisions from parameters with different values being truncated to5ZN5938.DOCX Page 13 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) the same compressed value. The collisions may be avoided across a set of records (e.g., across multiple records in a data set). An example algorithm for compressing the data elements while avoiding collisions is as follows:Step 1 : For field / and each data element in row i, set k = 1 ; dataj = dataij[O: ], where dataij[O :k] denotes the truncated version of dataj;Step 2: If there are collisions within field / (i.e., exist x,y where dataxj [0: ] = datayj[O: ], where dataxj + datayj), then k = +1 and go to Step 1 for next data element;Step 3: else, record truncated data elements and repeat for next field / .

[0058] The example compression algorithm described above operates by truncating values (e.g., transformed values or non-transformed values) of parameters in iterations until a collision is detected. For example, one (1 ) byte or character (or any other predetermined portion) may be truncated from the end of a parameter at a time, after which other parameters in the data set are compared to determine if there are any collisions (e.g., different original values that are truncated to the same value). This process may be continued until the parameters are as truncated as possible without incurring any collisions. Once a collision is detected (e.g., two different parameters match), the last truncation step may be reversed to revert the parameters to a compressed form without any collisions. In non-limiting embodiments, a hexadecimal or other representation of the parameter may be truncated.

[0059] It will be appreciated that various other algorithms may be used to truncate values of parameters. The compression engine 104 may also be configured to remove redundant values from the records. In a sample dataset, a record may be reduced a fraction of the size of the original record (e.g., half the size, a quarter of the size, an eighth of the size, a tenth of the size, and / or the like).

[0060] With continued reference to FIG. 1 , in non-limiting embodiments the computing device 100 may communicate the compressed records to the model 106. The model 106 may be, as an example, a large language model (LLM) configured to provide an output to a prompt. It will be appreciated that the model 106 may also be any type of model, such as but not limited to a machine-learning model, that is configured to generate an output based on patterns identified in one or more inputted compressed records. The model 106 may be trained with any technique based on, for example, historical transaction data and fraud determinations. Additionally or5ZN5938.DOCX Page 14 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) alternatively, the model 106 may be a general model prompted with contextual information about the patterns being identified, examples, and / or the like. In FIG. 1 the model is pictured as outside the computing system 1 10 (such as a local area network), but it will be appreciated that the model 106 may be local to the computing device 100 and / or part of a computing system 1 10 in non-limiting embodiments. The model 106 may return an output, such as but not limited to a risk score or fraud score, that is stored in the output database 108. The output database 108 may be part of the records database 101 such that the output of the model 106 is stored in relation to the record(s) being input.

[0061] The number and arrangement of systems and devices shown in FIG. 1 are provided as an example. There may be additional systems and / or devices, fewer systems and / or devices, different systems and / or devices, and / or differently arranged systems and / or devices than those shown in FIG. 1 . Furthermore, two or more systems or devices shown in FIG. 1 may be implemented within a single system or device, or a single system or device shown in FIG. 1 may be implemented as multiple, distributed systems or devices. Additionally or alternatively, a set of systems (e.g., one or more systems) or a set of devices (e.g., one or more devices) of the system 1000 may perform one or more functions described as being performed by another set of systems or another set of devices of the system 1000.

[0062] Referring now to FIG. 2, shown is a flow diagram for a method for masking model input data according to some non-limiting embodiments or aspects. The steps shown in FIG. 2 are for example purposes only. It will be appreciated that additional, fewer, different, and / or a different order of steps may be used in some non-limiting embodiments or aspects. In some non-limiting embodiments or aspects, a step may be automatically performed in response to performance and / or completion of a prior step. At step 200, a data set of at least one record is received. As an example, records from a records database may be obtained by querying a records database. A data set may include any number of records, and in non-limiting examples may include records pertaining to a particular account, merchant, issuer, and / or individual. For example, the records may be transaction records of payment transactions engaged in by different account identifiers. In non-limiting embodiments, short- or long-term risk profile tables may be generated to establish a baseline of historical transaction activity across certain data points (e.g., merchant name, device identifier, IP address, and / or the like). The table baseline may be dynamic, such as a rolling one-month trend line5ZN5938.DOCX Page 15 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) at a daily refresh rate. It will be appreciated that different sets of records may be used at various time intervals.

[0063] At step 202, the record(s) are analyzed for different types of parameters and a first type of parameter is transformed. In some examples, a classification model and / or classification rules may be applied to determine what portions of the record(s) correspond to classified parameters. For example, string parameters, numerical amount parameters, score parameters, and miscellaneous parameters may be identified. A different transformation algorithm may be applied at step 302 based on the type of parameter as described herein.

[0064] At step 204, it is determined whether there are more types (e.g., classifications) of parameters to mask. If there are additional parameter types not yet processed, the method proceeds back to step 202 to transform the next type of parameter. Once all the parameter types are processed and potentially masked, the method may proceed to step 206. At step 206, the record(s) may be compressed with one or more compression algorithms to truncate the parameters. The record(s) may be compressed individually or as a set of multiple records.

[0065] At step 208, the compressed record(s) are input into a model. For example, in some non-limiting embodiments, the compressed record(s) may be input into an LLM with a prompt that instructs the LLM to identify one or more patterns. Various other types of models may be used in non-limiting embodiments. In some non-limiting embodiments, multiple different models may be used to process the record(s). The model may output a determination or data that may be used to make a determination. In non-limiting embodiments, the prompt for an LLM or other model may be based on the masking and / or compression, including instructions for the model to consider the encoded, masked, and / or compressed nature of the input data. For example, a prompt may identify the data and / or portions of the data (e.g., such as certain parameters) as binary, hexadecimal, hashed, and / or the like.

[0066] At step 210, it is determined if fraud is detected in the record(s) (e.g., is likely or predicted) based on the output of the model. Fraud is one example of various types of patterns that may be analyzed and identified, and is discussed here and illustrated in FIG. 2 only as an example. The output of the model may be a binary indication of fraud. In other examples, the output of the model may be a score or some other output that is compared to a threshold or otherwise processed at step 210 to determine if fraud is detected. If fraud is detected, the method may perform an action at step 2125ZN5938.DOCX Page 16 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) such as, but not limited to, automatically notifying one or more entities, automatically placing a hold on an account or accounts, automatically declining a transaction, and / or the like.

[0067] In non-limiting embodiments, one or more foundational models may be used to process the records before or after the records are masked and / or compressed. As an example, an internal model may be used to process records that are in plaintext to identify risk scores and / or fraud, to determine what generative Al model to use, to select features from the records, and / or the like. After being processed by a foundational model, records with an uncertain risk score (e.g., with a confidence score below a threshold or the like) and / or all records may be processed by a generative Al model.

[0068] In some non-limiting embodiments, the system for masking model input data may be used in a situation where another fraud model (e.g., such as an internal fraud model and / or authentication platform) outputs a potential indication of fraud that is not high enough to trigger fraud protocols (e.g., a score is above 50 but less than 90 on a scale of 1 -99).

[0069] In non-limiting embodiments, the generative Al model may be prompted and / or configured to output different metrics for comparison and pattern detection. For example, an anomaly process may be provided to enable tracking of normal baseline patterns versus anomalies (e.g., a merchant’s transaction count activity, amount activity, international ratio activity, and / or the like). For example, four (4) standard deviations from the baseline may indicate the minimum anomaly threshold. Large scale attacks often see anomalies in the scale of over 100 deviations, as an example. The higher the deviation, the higher the output rating of the anomaly assigned to the potential fraud event.

[0070] In non-limiting embodiments, because a transaction may contain multiple transaction risk data features, a grouping process may be provided to aggregate anomalies across the transaction population. Such aggregation may enhance the reliability of fraud detection when multiple transactions across a short timeframe experience similar anomaly patterns. In non-limiting embodiments, this multi-factor triggering approach may also potentially boost the confidence level in fraud detection when multiple transactions across a short timeframe experience similar anomaly patterns.5ZN5938.DOCX Page 17 of 27Attorney Docket No. 08223-2407813 (8711 WO01 )

[0071] In non-limiting embodiments a foundational and / or internal fraud or risk model may process these filtered and consolidated data sets and, for any transaction meeting an established minimum criteria threshold, a confidence score may be output.

[0072] In non-limiting embodiments, the generative Al model may be trained and / or fine-tuned (e.g., through iterative fine-tuning or the like). The generative Al model may be prompted and / or configured to automatically generate computer program code (e.g., such as Python code or the like) that can be used to scan for and / or classify fraudulent transactions through pattern recognition or the like. The program code generated may then be back tested with data from a general transaction population. This automated feedback loop may provide insights into the parameters used to select training data, such as a platform risk score and anomaly thresholds. Such an arrangement ensures continuous learning and improvement of the automated fraud detection system. In non-limiting embodiments, the masking and / or compression techniques discussed herein may be used to train a model, such as a generative Al model, that can be shared and used by other entities without exposing sensitive information.

[0073] Non-limiting embodiments described herein may be used in various ways. Additionally or alternatively to fraud detection and / or risk scoring, non-limiting embodiments may be used to test models, to perform security testing on new environments (e.g., moving to a server-based platform), and / or the like. For example, it may be beneficial to mask and / or compress transaction records to store the records on a server computer (e.g., cloud storage) for security testing. Non-limiting embodiments may be used for model evaluation where sensitive data is used to train and / or execute the model and it is desired to share the model and / or results with a third-party. Non-limiting embodiments may be used to share data and / or models trained on data among entities, such as issuers, merchants, transaction processing systems, and / or the like. It will be appreciated that non-limiting embodiments may be used to process any type of sensitive data with a generative Al model, including but not limited to financial records, medical records, and / or the like.

[0074] FIG. 3 shows an electronic payment processing network 1 100 according to non-limiting embodiments or aspects. The payment processing network may be used in conjunction with the systems and methods described herein. It will be appreciated that the particular arrangement of the electronic payment processing network 1 100 shown is for example purposes only, and that various arrangements are possible. A5ZN5938.DOCX Page 18 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) transaction processing system 1101 (e.g., a transaction handler) is shown to be in communication with one or more issuer systems (e.g., such as an issuer system 1 106) and one or more acquirer systems (e.g., such as an acquirer system 1 108). Although only a single issuer system 1 106 and single acquirer system 1108 are shown, it will be appreciated that the transaction processing system 1101 may be in communication with a plurality of issuer systems and / or acquirer systems. In some embodiments, the transaction processing system 1 101 may also operate as an issuer system such that both the transaction processing system 1 101 and issuer system 1106 are a single system and / or controlled by a single entity.

[0075] In some non-limiting embodiments or aspects, the transaction processing system 1 101 may communicate with a merchant system 1 104 directly through a public or private network connection. Additionally or alternatively, the transaction processing system 1 101 may communicate with the merchant system 1 104 through a payment gateway 1 102 and / or the acquirer system 1 108. In some non-limiting embodiments or aspects, the acquirer system 1108 associated with the merchant system 1 104 may operate as the payment gateway 1 102 to facilitate the communication of transaction requests from the merchant system 1 104 to the transaction processing system 1 101. The merchant system 1 104 may communicate with the payment gateway 1 102 through a public or private network connection. For example, a merchant system 1 104 that includes a physical POS device may communicate with the payment gateway 1 102 through a public or private network to conduct card-present transactions. As another example, a merchant system 1104 that includes a server (e.g., a web server) may communicate with the payment gateway 1 102 through a public or private network, such as a public Internet connection, to conduct card-not-present transactions.

[0076] In some non-limiting embodiments or aspects, the transaction processing system 1101 , after receiving a transaction request from the merchant system 1 104 that identifies an account identifier of a payor (e.g., such as an account holder) associated with an issued payment device 1 1 10, may generate an authorization request message to be communicated to the issuer system 1 106 that issued the payment device 1 110 and / or account identifier. The issuer system 1 106 may then approve or decline the authorization request and, based on the approval or denial, generate an authorization response message that is communicated to the transaction processing system 1 101. The transaction processing system 1 101 may communicate an approval or denial to the merchant system 1 104. When the issuer system 1 1065ZN5938.DOCX Page 19 of 27Attorney Docket No. 08223-2407813 (8711 WO01 ) approves the authorization request message, it may then clear and settle the payment transaction between the issuer system 1 106 and acquirer system 1108.

[0077] Referring now to FIG. 4, shown is a diagram of example components of a device 400 according to non-limiting embodiments. Device 400 may correspond to the computing device 100, the compression engine 104, and / or the masking engine 102 in FIG. 1 , as an example. In some non-limiting embodiments, such systems or devices may include at least one device 400 and / or at least one component of device 400. The number and arrangement of components shown are provided as an example. In some non-limiting embodiments, device 400 may include additional components, fewer components, different components, or differently arranged components than those shown. Additionally, or alternatively, a set of components (e.g., one or more components) of device 400 may perform one or more functions described as being performed by another set of components of device 400.

[0078] As shown in FIG. 4, device 400 may include a bus 402, a processor 404, memory 406, a storage component 408, an input component 410, an output component 412, and a communication interface 414. Bus 402 may include a component that permits communication among the components of device 400. In some non-limiting embodiments, processor 404 may be implemented in hardware, firmware, or a combination of hardware and software. For example, processor 404 may include a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), etc.), a microprocessor, a digital signal processor (DSP), and / or any processing component (e.g., a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), etc.) that can be programmed to perform a function. Memory 406 may include random access memory (RAM), read only memory (ROM), and / or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, optical memory, etc.) that stores information and / or instructions for use by processor 404.

[0079] With continued reference to FIG. 4, storage component 408 may store information and / or software related to the operation and use of device 400. For example, storage component 408 may include a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, a solid-state disk, etc.) and / or another type of computer-readable medium. Input component 410 may include a component that permits device 400 to receive information, such as via user input (e.g., a touch screen display, a keyboard, a keypad, a mouse, a button, a switch, a microphone, etc.).5ZN5938.DOCX Page 20 of 27Attorney Docket No. 08223-2407813 (8711 WO01 )Additionally, or alternatively, input component 410 may include a sensor for sensing information (e.g., a global positioning system (GPS) component, an accelerometer, a gyroscope, an actuator, etc.). Output component 412 may include a component that provides output information from device 400 (e.g., a display, a speaker, one or more light-emitting diodes (LEDs), etc.). Communication interface 414 may include a transceiver-like component (e.g., a transceiver, a separate receiver and transmitter, etc.) that enables device 400 to communicate with other devices, such as via a wired connection, a wireless connection, or a combination of wired and wireless connections. Communication interface 414 may permit device 400 to receive information from another device and / or provide information to another device. For example, communication interface 414 may include an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a Wi-Fi® interface, a cellular network interface, and / or the like.

[0080] Device 400 may perform one or more processes described herein. Device 400 may perform these processes based on processor 404 executing software instructions stored by a computer-readable medium, such as memory 406 and / or storage component 408. A computer-readable medium may include any non-transitory memory device. A memory device includes memory space located inside of a single physical storage device or memory space spread across multiple physical storage devices. Software instructions may be read into memory 406 and / or storage component 408 from another computer-readable medium or from another device via communication interface 414. When executed, software instructions stored in memory 406 and / or storage component 408 may cause processor 404 to perform one or more processes described herein. Additionally, or alternatively, hardwired circuitry may be used in place of or in combination with software instructions to perform one or more processes described herein. Thus, embodiments described herein are not limited to any specific combination of hardware circuitry and software. The term “configured to,” as used herein, may refer to an arrangement of software, device(s), and / or hardware for performing and / or enabling one or more functions (e.g., actions, processes, steps of a process, and / or the like). For example, “a processor configured to” may refer to a processor that executes software instructions (e.g., program code) that cause the processor to perform one or more functions.5ZN5938.DOCX Page 21 of 27Attorney Docket No. 08223-2407813 (8711 WO01 )

[0081] Although embodiments have been described in detail for the purpose of illustration, it is to be understood that such detail is solely for that purpose and that the disclosure is not limited to the disclosed embodiments or aspects, but, on the contrary, is intended to cover modifications and equivalent arrangements that are within the spirit and scope of the appended claims. For example, it is to be understood that the present disclosure contemplates that, to the extent possible, one or more features of any embodiment or aspect can be combined with one or more features of any other embodiment or aspect.5ZN5938.DOCX Page 22 of 27

Claims

Attorney Docket No. 08223-2407813 (8711WO01)WHAT IS CLAIMED IS:1 . A system for comprising at least one processor configured to: receive a data set comprising a plurality of records, each record comprising a plurality of parameters; transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and input the modified data set into a model.

2. The system of claim 1 , wherein the first type comprises a monetary numerical value, and wherein the first transformation algorithm comprises a cryptographic commitment function.

3. The system of claim 2, wherein the second type comprises a nonmonetary value, and wherein the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof.

4. The system of claim 1 , wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values.

5. The system of claim 1 , wherein the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity.5ZN5938.DOCX Page 23 of 27Attorney Docket No. 08223-2407813 (8711WO01)6. The system of claim 1 , wherein the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

7. A computer-implemented method, comprising: receiving a data set comprising a plurality of records, each record comprising a plurality of parameters; transforming at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transforming at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compressing each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and inputting the modified data set into a model.

8. The method of claim 7, wherein the first type comprises a monetary numerical value, and wherein the first transformation algorithm comprises a cryptographic commitment function.

9. The method of claim 8, wherein the second type comprises a nonmonetary value, and wherein the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof.

10. The method of claim 7, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values.5ZN5938.DOCX Page 24 of 27Attorney Docket No. 08223-2407813 (8711WO01)1 1 . The method of claim 7, wherein the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity.

12. The method of claim 7, wherein the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.

13. A computer program product comprising at least one non- transitory computer-readable medium including program instructions that, when executed by at least one processor, cause the at least one processor to: receive a data set comprising a plurality of records, each record comprising a plurality of parameters; transform at least one parameter of a first type in each record with a first transformation algorithm to generate a first masked parameter in each record; transform at least one parameter of a second type in each record with a second transformation algorithm to generate a second masked parameter in each record; compress each record of the plurality of records by truncating a value of at least one of the following: the first masked parameter, the second masked parameter, or any combination thereof, resulting in a modified data set comprising a plurality of compressed records; and input the modified data set into a model.

14. The computer program product of claim 13, wherein the first type comprises a monetary numerical value, and wherein the first transformation algorithm comprises a cryptographic commitment function.

15. The computer program product of claim 14, wherein the second type comprises a non-monetary value, and wherein the second transformation algorithm comprises at least one of the following: a noise algorithm, a hash algorithm, or any combination thereof.5ZN5938.DOCX Page 25 of 27Attorney Docket No. 08223-2407813 (8711WO01)16. The computer program product of claim 13, wherein truncating is based on other values of the at least one of the first masked parameter and the second masked parameter in other records of the plurality of records, such that truncated values of each value are unique among the other values.

17. The computer program product of claim 13, wherein the model is configured or prompted to determine if one or more records of the plurality of compressed records represents fraudulent activity.

18. The computer program product of claim 13, wherein the model comprises a large language model, and wherein inputting the modified data set comprises uploading the modified data set with a corresponding prompt instructing the model to perform at least one task with the modified data set.5ZN5938.DOCX Page 26 of 27

Images