Detailed Action
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Acknowledgments
The Amendment filed on 05/04/26 is acknowledged.
Status of Claims
Claims 22-41 are pending.
In the Amendment filed on 05/04/26, claims 22, 23, 26, 28, 32, 33, 36, 38 and 41 were amended, and no claims were cancelled or added (claims 1-21 were cancelled in a previous paper).
Claims 22-41 are rejected.
Response to Arguments
Regarding priority
In view of the amendments, the indication in the previous Office Action regarding lack of support in parent application no. 18/189,952 for claims 23, 26, 28, 33, 36 and 38 is withdrawn.
Regarding the objections to the drawings/specification
For clarification of the record:
The replacement drawing as filed does not comply with 37 CFR 1.121(d) ("All changes to the drawings shall be explained, in detail, in either the drawing amendment or remarks section of the amendment paper.") No explanation of the changes to the drawings was found in the entirety of the Response, including the replacement drawing, as filed.
As best understood, the instant changes to the drawings (namely, in replacement figure 17A) are as follows:
- reference numeral 1700 has been deleted
If the above-indicated changes do not constitute the entirety of the changes made to the drawings in Applicant's instant Response, then Applicant should so indicate and clarify in the next Response, in order to clarify the record.
As a courtesy to Applicant, the replacement drawing is entered.
In view of the replacement drawing and the changes to the specification, the objections to the drawings/specification are withdrawn.
Regarding the rejection under 35 U.S.C. 101
Applicant's arguments have been fully considered but they are not persuasive.
The Office responds to Applicant's arguments below. Headings in the discussion below refer to Applicant's Response.
A. Step 1: The Claims Fall Within One of the Four Statutory Categories
The Office agrees that the claims fall within one of the four statutory categories of 35 U.S.C. 101.
B. Step 2A: Two-Prong Inquiry
1. Prong One: The Claims Do Not Recite an "Abstract Idea"
Applicant argues:
As amended, independent claim 22 is directed to "a computer-implemented method for identifying unauthorized activity in a computing system including at least one processor." In particular, the claim recites "predicting, using the machine learning model, an outcome based on the three-digit risk indicator, wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action, flagging the processed action for review, or allowing the processed action." These are machine-implemented control actions that govern execution of a processed action within a computing system, not human judgment, economic planning, or organizational behavior. (Response, pp. 14-15; emphasis added)
In response:
The bolded content shown above in Applicant's argument represents abstract idea. The underlined content represents additional elements. As seen from the bolded and non-bolded portions, the additional elements are generic elements, recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, in other words, the additional elements merely 'apply' the abstract idea, or alternatively, the additional elements merely generally link the use of a judicial exception to a particular technological environment or field of use. Likewise, the last sentence in the quote block above reconfirms this point, where "control actions that govern execution of a processed action" constitutes abstract idea, implemented by a machine/computing system, i.e., generic elements, recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, thus merely 'applying' the abstract idea, or alternatively, merely generally linking to a particular technological environment or field of use.
Applicant further argues:
As described in the specification, the machine learning model executes within the financial institution's computing infrastructure and generates risk indicators that directly control system behavior, such as automatically allowing a transaction, placing the transaction on hold, or flagging the transaction for analyst review. See, e.g., Specification at ¶¶ [0055], [0058] - [0059], and [0071]. The specification explains that, based on the generated risk indicator, the processor automatically enforces outcomes including allowance, analyst review, or auto-holding of the transaction. Id. These actions are performed by the system itself, rather than merely presenting information for human evaluation. Accordingly, the claims recite a machine-implemented control mechanism, not a human decision-making process or a fundamental economic practice. (Response, p. 15)
In response:
The above account of the claimed subject matter merely describes a system where a generic computer element processes a transaction by performing actions according to a set of rules or conditions. Again, the additional elements/computer elements are recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, in other words, the additional elements merely 'apply' the abstract idea, or alternatively, the additional elements merely generally link the use of a judicial exception to a particular technological environment or field of use.
Applicant further argues:
The amended claims further recite, "retaining, by the machine learning model, information associated with one or more previously generated risk indicators" including transactional data, customer characteristics, or historical data, and "tuning the numerical risk indicator based on the information, wherein the tuning identifies optimal values of one or more hyperparameters to maximize performance of the machine learning model." The specification describes training the machine learning model to retain information from previously generated indicators and using that retained information in an experimental hyperparameter optimization process to improve model performance. Specification at ¶ [0072]. As disclosed, hyperparameter tuning refers to selecting values that control the learning process of the model to optimize detection of unauthorized transactions based on transactional data, customer characteristics, and historical data. Id. Retaining prior model outputs and tuning hyperparameters are core technical challenges in machine learning system design, not abstract business rules or mental processes. (Response, pp. 15-16; emphasis added)1
In response:
The bolded content shown above in Applicant's argument represents abstract idea. The underlined content represents additional elements. As claimed and as described here, the additional elements represent merely generic computer elements, recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, in other words, the additional elements merely 'apply' the abstract idea, or alternatively, the additional elements merely generally link the use of a judicial exception to a particular technological environment or field of use. In particular, the machine learning limitations do not represent any improvement in machine learning, but merely constitute what machine learning is (in machine learning, the machine learns from its previous operations/results and refines its behavior based on those previous operations/results (feedback), in the service of improving its performance).
Applicant further argues:
As the USPTO's Memorandum on Reminders on Evaluating Subject Matter Eligibility cautions, examiners should not classify claim limitations as mental processes or methods of organizing human activity when those limitations cannot practically be performed in the human mind. See Memorandum on Reminders on Evaluating Subject Matter Eligibility, U.S. Patent and Trademark Office (Aug. 4, 2025) ("Memorandum"). Here, the amended claims could not practically be performed in the human mind for at least the above-recited reasons. (Response, p. 16)
In response:
The claims were not rejected based on being a mental process or being able to practically be performed in the human mind. Accordingly, the argument is not on point to the rejection.
Applicant further argues:
When considered as a whole, the amended claims are directed to operation and control of a machine learning system that generates risk indicators, adapts model behavior over time, and automatically controls execution of processed actions. Accordingly, the claims are directed to a technological process rather than an abstract idea and therefore do not recite a judicial exception under Step 2A, Prong One. (Response, p. 16; emphasis added)
In response:
The above paragraph merely summarizes the previous arguments. The bolded content shown above in Applicant's argument corresponds to the abstract idea. The underlined content corresponds to additional elements. The above description conforms to the previous arguments, wherein the claimed subject matter is seen to be abstract idea carried out using additional elements/generic computer elements, recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, in other words, the additional elements merely 'apply' the abstract idea, or alternatively, the additional elements merely generally link the use of a judicial exception to a particular technological environment or field of use. As such, the claimed process is technological only insofar as it uses a generic computer to apply the abstract idea, or generally links the use of a judicial exception to a particular technological environment or field of use, namely to generic computer elements.
2. Prong Two: The Claims Integrate Any Alleged Abstract Idea Into a Practical Application
In this section, Applicant presents the same substance/arguments as in the previous section (re Prong One). Accordingly, Applicant's arguments have been addressed above.
Note the "automated control of transaction execution within the computing system itself based on outputs of the machine learning model" (p. 17) merely represents the abstract idea (transaction execution) carried out using additional elements/generic computer elements (automated control … within the computing system itself based on outputs of the machine learning model), recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, in other words, the additional elements merely 'apply' the abstract idea, or alternatively, the additional elements merely generally link the use of a judicial exception to a particular technological environment or field of use.
Note the alleged fact that the "abstract concept is applied to control real-world transaction processing through a specific machine-implemented workflow" (p. 17) does not constitute a practical application. The real-world transaction processing controlled through a workflow is merely abstract idea. The machine implementation/control is merely implementing the abstract idea using generic computer elements, recited at a high level of generality and not described, used off-the shelf as a tool in their ordinary capacities, in other words, the additional elements merely 'apply' the abstract idea, or alternatively, the additional elements merely generally link the use of a judicial exception to a particular technological environment or field of use.
The new content in this section (Prong Two) relative to the previous section (Prong One) is as follows:
The Patent Trial and Appeal Board has emphasized that eligibility analysis must account for claim language reflecting improvements in artificial intelligence and machine learning technology. See Ex parte Desjardins, Appeal 2024-000567 (P.T.A.B. Sept. 26, 2025). Desjardins highlights that claims directed to improvements in model operation (such as reduced complexity or improved learning behavior) can satisfy Step 2A, Prong Two. (Response, p. 18)
In response:
The subject matter of the instant claims is not related to that of Desjardins, nor is it analogous in respect of 35 U.S.C. 101. Desjardins was found to reflect the following improvements in computer functionality:
xiii. An improved way of training a machine learning model that protected the model’s knowledge about previous tasks while allowing it to effectively learn new tasks; Ex Parte Desjardins, Appeal No. 2024-000567 (PTAB September 26, 2025, Appeals Review Panel Decision) (precedential); and
xiv. Improvements to computer component or system performance based upon adjustments to parameters of a machine learning model associated with tasks or workstreams; Ex Parte Desjardins, Appeal No. 2024-000567 (PTAB September 26, 2025, Appeals Review Panel Decision) (precedential).2
As seen from the above, Desjardins contained specific improvements to machine learning. In contrast, as explained above, the instant claims merely reflect what generic machine learning is, not an improvement to machine learning.
C. Step 2B: The Claims Recite Significantly More Than Any Alleged Judicial Exception
In this section, Applicant presents the same substance/arguments as in the previous sections (re Step 2A, Prongs Two and One). Accordingly, Applicant's arguments have been addressed above.
The new content in this section (Step 2B) relative to the previous section (Step 2A, Prongs One and Two) is as follows:
Viewed individually and as an ordered combination, the amended claim
limitations recite specific technical mechanisms for adaptive machine-learning operation and automated transaction control that are not well-understood, routine, or conventional. The claims therefore recite significantly more than any alleged judicial exception, and Applicant respectfully requests withdrawal of the rejection under 35 U.S.C. § 101. (Response, p. 20; emphasis added)
In response:
The claims were not rejected based on being well-understood, routine, or conventional. Accordingly, this argument is not on point to the rejection.
Rather, the claims were rejected under step 2B based on the fact that they constitute merely an abstract idea applied using a generic computer element, or alternatively an abstract idea generally linked to a particular technological environment or field of use. As such, the claims are not enough to qualify as "significantly more" under step 2B. See MPEP 2106.05 I.A.:
Limitations that the courts have found not to be enough to qualify as "significantly more" when recited in a claim with a judicial exception include:
i. Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, e.g., a limitation indicating that a particular function such as creating and maintaining electronic records is performed by a computer, as discussed in Alice Corp., 573 U.S. at 225-26, 110 USPQ2d at 1984 (see MPEP § 2106.05(f));
…
iv. Generally linking the use of the judicial exception to a particular technological environment or field of use, e.g., a claim describing how the abstract idea of hedging could be used in the commodities and energy markets, as discussed in Bilski v. Kappos, 561 U.S. 593, 595, 95 USPQ2d 1001, 1010 (2010) or a claim limiting the use of a mathematical formula to the petrochemical and oil-refining fields, as discussed in Parker v. Flook, 437 U.S. 584, 588-90, 198 USPQ 193, 197-98 (1978) (MPEP § 2106.05(h)). (emphasis added)
Regarding the rejections under 35 U.S.C. 103
Applicant's arguments have been fully considered but they are not persuasive (as explained below) and/or are moot in view of the new combinations of references being used in the current rejections.
The Office responds to Applicant's arguments below. Page numbers in the discussion below refer to Applicant's Response unless otherwise indicated.
Applicant argues:
As acknowledged in the Office Action, Beckman does not explicitly disclose determining a probability that a processed action belongs to a class, as recited in independent claim 22. See Office Action at 15. Rather, Beckman describes categorizing activity into qualitative risk levels (e.g., low/medium/high) (see Beckman at 7:7-9), without "determining, by a machine learning model, a probability," as recited by the claims. (Response, p. 21; emphasis added)
In response:
First, the underlined portion shown above is misleading by omitting relevant related content. Specifically, the Office Action stated:
since it is not explicitly stated that the determination of belonging to a class is a determination of a probability of belonging to a class, it would be obvious to combine embodiments and incorporate a probabilistic determination such as … (Office Action, p. 15)
In other words, the Office Action stated that, although Beckman does not explicitly state that the determination in question is a determination of a probability, it would be obvious to combine Beckman's embodiments so as to incorporate a probabilistic determination into the determination in question, such that the combination of Beckman's embodiments makes it obvious that the determination in question is a determination of a probability, in other words, the combination of Beckman's embodiments renders obvious "determining a probability that the processed action belongs to a class." The rejection then explained in detail and at length how/which particular teachings of Beckman taught/rendered obvious the limitation in question.
Second, it is noted for clarification of the record that Beckman's teaching cited here by Applicant ("categorizing activity into qualitative risk levels (e.g., low/medium/ high)") is not the particular teaching of Beckman cited by the Office Action as teaching the limitation in question. Thus, Applicant's argument is again misleading.
In sum, while the rejection explained in detail how Beckman taught/rendered obvious the limitation in question, Applicant neglects to address the substance of this explanation in the rejection, and instead merely posits a conclusory statement ("Beckman does not explicitly disclose determining a probability that a processed action belongs to a class") or at best a very incomplete argument, in that, while Applicant offers a teaching of Beckman ("Beckman describes categorizing activity into qualitative risk levels (e.g., low/medium/high)"), Applicant does not address the particular teachings of Beckman cited as teaching the limitation in question.
Applicant further argues:
Further, Beckman conflates the concepts of a "risk indicator" and an "outcome." Specifically, the Office Action relies on the same Beckman disclosure (e.g., low/medium/high risk) to allegedly meet both the claimed "numerical risk indicator" and the claimed "outcome." See Office Action at 16-17. However, the claims recite that the outcome is distinct from and determined based on the numerical risk indicator. Beckman does not disclose or suggest "an outcome based on the three-digit risk indicator," as recited in the claims. Accordingly, Beckman fails to teach at least these limitations. (Response, p. 21; emphasis added)
In response:
First, the underlined portion shown above is misleading. Beckman was not cited as teaching the entirety of what Applicant alleges Beckman does not teach. Specifically, Beckman as not cited as teaching "three-digit."
Second, the rejection has now been revised to clarify how Beckman teaches the risk indicator and the outcome as separate items. Explanation is provided in the rejection in the body of the Office Action hereinbelow.
In this regard, as best understood, the portions of Applicant's specification that support the recited "predicting … an outcome based on the … risk indicator" are 0005-0006 and 0057-0058 (Fig. 10). According to 0005-0006, "the generated indicator causes the at least one processor to: stop the processed action; flag the processed action for review; or allow the processed action." According to 0057-0058, the "risk results" are the "outcomes based on the risk indicators," and "the processed action …, when assigned [a particular, e.g., low, medium, or high] risk indicator …, would result in [the particular outcome, e.g., allowance, review, hold]." Thus, as per the disclosure, the indicator causes the outcome (namely, stopping, flagging, or allowing) (0005-0006), or the processed action, when assigned a particular risk indicator, results in a particular outcome, based on the particular risk indicator (0057-0058). This mode of causation/ operation does not conform to the ordinary meaning of "predict." Accordingly, Applicant is deemed to be acting as its own lexicographer in defining "predict" to refer to the necessary determination ("cause"; "would result") of an outcome, subsequent to and consequent to (based on) a preceding prediction (in the ordinary meaning of "predict") of a risk indicator.3
Applicant further argues:
Poduval does not remedy the deficiencies of Beckman with respect to the subject matter of the independent claims. Poduval is directed to chargeback prediction for transactions that have already been processed, focusing on post-transaction analysis and risk scoring associated with subsequent chargeback events. See, e.g., Poduval at ¶ [0003]. This is fundamentally different from the claims, which determine "a probability that the processed action belongs to a class" and predict "an outcome based on the three-digit risk indicator." (Response, pp. 21-22; emphasis added)
In response:
First, the underlined portion shown above is misleading. Poduval was not cited as teaching the entirety of the claimed subject matter that Applicant here cites. Specifically, Poduval was relied upon as teaching that the risk indicator is "a three-digit number."4 Thus, the underlined portion shown above is misleading.
Second, while Poduval deals with chargebacks, the rejection sets forth a motivation statement articulating a motivation/reason for combining the particular teachings of Poduval in question with Beckman. Applicant neglects to address this motivation/reason for combining.
Regarding the amended claim language:
wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action; flagging the processed action for review; or allowing the processed action
it is noted that this language is set forth as a disjunction ("at least one of … or"). Therefore, the prior art need teach only one of the three recited alternatives (namely, (1) stopping, (2) flagging, (3) allowing, as recited) in order to teach the recitation.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 22-41 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Claims 22-41 are directed to a computer-implemented method, computing system, or non-transitory computer-readable medium, which are/is one of the statutory categories of invention. (Step 1: YES)
Claims 22, 32 and 41 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims recite a computer-implemented method, computing system, and non-transitory computer-readable medium for determining a likelihood of fraud of a transaction (see specification 0043-0044 for clarification of the term "class").
For claims 22, 32 and 41 (claim 32 being deemed representative), the limitations (indicated below in bold) of:
at least one processor configured to:
receive, by the at least one processor, a processed action of a user;
determine, by a machine learning model, a probability that the processed action belongs to a class;
based on the probability, generate, using the machine learning model, a numerical risk indicator, wherein the numerical risk indicator is a three-digit risk indicator associated with unauthorized activity; and
predict, using the machine learning model, an outcome based on the three-digit risk indicator, wherein the outcome includes the at least one processor performing at least one of:
automatically stopping the processed action;
flagging the processed action for review; or
allowing the processed action.
as drafted, constitute a process that, under the broadest reasonable interpretation, covers "certain methods of organizing human activity," specifically, "fundamental economic practices or principles" and/or "commercial or legal interactions," but for recitation of generic computer components and generally linking the use of a judicial exception to a particular technological environment or field of use. The Examiner notes that "fundamental economic practices" or "fundamental economic principles" describe concepts relating to the economy and commerce, including hedging, insurance, and mitigating risks, and "commercial interactions" or "legal interactions" include agreements in the form of contracts, legal obligations, advertising, marketing or sales activities or behaviors, and business relations. MPEP 2106.04(a)(2)II.A.,B. If a claim limitation, under its broadest reasonable interpretation, covers "fundamental economic practices or principles" and/or "commercial or legal interactions," but for recitation of generic computer components and generally linking the use of a judicial exception to a particular technological environment or field of use, then it falls within the "certain methods of organizing human activity" grouping of abstract ideas. Accordingly, claims 22, 32 and 41 recite an abstract idea. (Step 2A - Prong 1: YES. The claims recite an abstract idea.)
This judicial exception is not integrated into a practical application. Claims 22, 32 and 41 recite the additional elements of at least one processor, a machine learning model (the foregoing recited in claims 22, 32 and 41), and a non-transitory computer-readable medium storing a set of instructions for identifying unauthorized activity in a computing system including at least one processor, the set of instructions comprising one or more instructions that, when executed by one or more processors of the computing system, cause the computing system to [perform operations] (the foregoing recited in claim 41), that implement the abstract idea. These additional elements are not described by the applicant and they are recited at a high level of generality (i.e., one or more generic computer elements performing generic computer functions, or generally linking the use of a judicial exception to a particular technological environment or field of use), such that they amount to no more than mere instructions to apply the exception using generic computer elements (namely, at least one processor, a machine learning model, and a non-transitory computer-readable medium storing a set of instructions for identifying unauthorized activity in a computing system including at least one processor, the set of instructions comprising one or more instructions that, when executed by one or more processors of the computing system, cause the computing system to [perform operations]), or such that they amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (namely, a machine learning model). Accordingly, even in combination these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. (Step 2A - prong 2: NO. The additional elements do not integrate the abstract idea into a practical application.)
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception itself. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of at least one processor, a machine learning model (the foregoing recited in claims 22, 32 and 41), and a non-transitory computer-readable medium storing a set of instructions for identifying unauthorized activity in a computing system including at least one processor, the set of instructions comprising one or more instructions that, when executed by one or more processors of the computing system, cause the computing system to [perform operations] (the foregoing recited in claim 41), to perform the noted steps amount to no more than mere instructions to apply the exception using generic computer elements or generally linking the use of a judicial exception to a particular technological environment or field of use. Mere instructions to apply an exception using generic computer elements or generally linking the use of a judicial exception to a particular technological environment or field of use cannot provide an inventive concept ("significantly more"). Accordingly, even in combination, these additional elements do not provide significantly more. As such, claims 22, 32 and 41 are not patent eligible. (Step 2B: NO. The claims do not provide significantly more.)
Dependent claims 23-31 and 33-40 are similarly rejected because they further define/narrow the abstract idea of independent claims 22, 32 and 41 as discussed above, and/or do not integrate the abstract idea into a practical application or provide an inventive concept such as would render the claims eligible, whether each is considered individually or as an ordered combination.
As for further defining/narrowing the abstract idea:
Dependent claims 23 and 33 merely further describe predicts the outcome based on a comparison of the three-digit risk indicator with at least one threshold.
Dependent claims 24 and 34 merely further describe wherein the at least one threshold is determined … based on at least one of a deposit type, a deposit amount, a deposit location, or a fraud history.
Dependent claims 25 and 35 merely further describe wherein the probability that the processed action belongs to the class is based on at least one of transactional data, a customer characteristic, or historical data.
Dependent claims 26 and 36 merely further describe retaining … information associated with one or more previously generated risk indicators, wherein the information includes at least one of the transactional data, the customer characteristic, or the historical data, and tuning the numerical risk indicator based on the information, wherein the tuning ….
Dependent claims 27 and 37 merely further describe wherein the three-digit risk indicator is derived from a … probability.
Dependent claims 28 and 38 merely further describe wherein … based on information associated with at least one of the processed action of the user, the probability, the three-digit risk indication, or the outcome.
Dependent claims 29 and 39 merely further describes wherein … based on a relative precision.
Dependent claims 30 and 40 merely further describe wherein the class is generated.
Dependent claim 31 merely further describes wherein the processed action of the user is enriched in real time.
As for additional elements:
Dependent claims 23, 24, 30, 33, 34 and 40 recite "the machine learning model." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element or generally linking the use of a judicial exception to a particular technological environment or field of use. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Dependent claims 26 and 36 recite "the machine learning model" and "identifies optimal values of one or more hyperparameters to maximize performance of the machine learning model." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element or generally linking the use of a judicial exception to a particular technological environment or field of use. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Dependent claims 27 and 37 recite "model" (referring to the machine learning model). This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element or generally linking the use of a judicial exception to a particular technological environment or field of use. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Dependent claims 28 and 38 recite "wherein the machine learning model is tuned." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element or generally linking the use of a judicial exception to a particular technological environment or field of use. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Dependent claims 29 and 39 recite "wherein the machine learning model is benchmarked." This recitation is at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer element or generally linking the use of a judicial exception to a particular technological environment or field of use. Even in combination these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself.
Dependent claims 25, 31, and 35 do not recite any additional elements, and accordingly, for the reasons provided above with respect to the independent claims, are not patent eligible.
Therefore, dependent claims 23-31 and 33-40 are not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries set forth in Beckman v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 22, 23, 25, 27, 30, 32, 33, 35, 37, 40 and 41 are rejected under 35 U.S.C. 103 as being unpatentable over Beckman et al. (U.S. Patent No. 10,872,341 B1), hereafter Beckman, in view of Poduval et al. (U.S. Patent No. 2022/0358507 A1), hereafter Poduval.
Regarding Claims 22, 32 and 41
Beckman teaches:
(claim 22) the method being performed by the at least one processor and comprising: (4:20-48 payment system 130 may include processor(s), software, etc. to perform operations; 5:66-6:20 risk assessment engine 150 may include processor(s), software, etc. to perform operations)
(claim 32) at least one processor configured to: (4:20-48 payment system 130 may include processor(s), software, etc. to perform operations; 5:66-6:20 risk assessment engine 150 may include processor(s), software, etc. to perform operations)
(claim 41) A non-transitory computer-readable medium storing a set of instructions for identifying unauthorized activity in a computing system including at least one processor, the set of instructions comprising: one or more instructions that, when executed by one or more processors of the computing system, cause the computing system to: (4:20-48 payment system 130 may include processor(s), software, etc. to perform operations, including as per 4:49-5:7 performing authorization and authentication; 5:66-6:20 risk assessment engine 150 may include processor(s), software, etc. to perform operations, including as per 6:21-58 determining whether communications/transactions are fraudulent)
receive, by the at least one processor, a processed action of a user; (8:13-17, Fig. 2, 204 payment system 130 receives transaction authorization request)
determine, by a machine learning model, a probability that the processed action belongs to a class; (8:13-43, Fig. 2, 208 payment server 130 "[determines] whether the transaction may be fraudulent," "using any suitable technique and fraud detection process," based on risk factors as described at 8:20-27. The result of this process is a determination that the transaction is not fraudulent (8:28-37) or a determination that the transaction is fraudulent (8:38-64), i.e., a determination that the processed action belongs to a class, either the class of fraudulent transactions or the class of not fraudulent transactions. As the determination is made by any suitable fraud detection process, e.g., based on the risk factors at 8:20-27, it is understood that such a determination is a determination of a probability, not an absolute determination (e.g., the fact that "the account … has been recently flagged for fraud" or not is a risk factor that makes it more likely that the transaction is fraudulent or not; such factors do not yield absolute determinations). Nonetheless, since it is not explicitly stated that the determination of belonging to a class is a determination of a probability of belonging to a class, it would be obvious to combine embodiments and incorporate a probabilistic determination such as (1) that described as capable of being performed by transaction verification service 140 (5:11-26 "Transaction verification service 140 may comprise software-based services, APIs, SDKs, or the like configured to perform various fraud detection operations, as discussed further herein. For example, transaction verification service 140 may comprise a mobile device authentication and fraud prevention software solution, such as the INAUTH SECURITY PLATFORM™ offered by InAuth, Inc. The mobile device authentication and fraud prevention software solution may provide additional fraud detection services, such as, for example, the generation of a fraud score based on captured user device data. The fraud score may be configured to provide an analysis of the likelihood that user device 110, or an email account for user 101, has been compromised by a third party.") or again such as (2) that described as being performed by risk assessment engine 150, based on comparable risk factors (6:21-7:44, 9:58-10:33 e.g., "the captured device data may be input into the statistical model, the machine learning model, or the artificial intelligence model to determine a risk of fraud. … Based on the data consumption [e.g., device data, historical transaction fraud data, non-device related attributes], the model may be leveraged to predict whether the verification is coming from a fraudulent device."), because determinations as to whether a transaction is fraudulent, based on risk data, in the context of a payment system/payment processor, are generally probability determinations, as such determinations do not generally admit of absolute certainty, rather a probability/likelihood is a more reliable/plausible determination, hence affording more accurate, useful, and effective, fraud detection and mitigation/prevention; regarding by a machine learning model: 6:59-76 "risk assessment engine 150 may implement statistical models, machine learning, artificial intelligence, and the like to aid in identifying possible fraud. In that regard, the captured user device data may be input into the statistical model, the machine learning model, or the artificial intelligence model to determine a risk of fraud."; 15:7-16 "any of the operations may be conducted or enhanced by … machine learning")
based on the probability, generate, using the machine learning model, a … risk indicator, wherein the … risk indicator is a … risk indicator associated with unauthorized activity; and (9:58-10:33, Fig. 3, 318-324; the risk indicator is the output/fraud determination classification of "low risk," "medium risk" or "high risk," as taught by 7:7-30, see also 10:34-11:3; regarding based on the probability: note that step 318 (determining a secondary fraud risk (e.g., 9:61-67), based on captured device data, historical data, and non-device attributes (e.g., 10:25-32)) (generating a risk indicator) is based on step 208 (Fig. 2) (determining whether the transaction is fraudulent) (determining a probability that the processed action belongs to a class), via a series of intermediate steps (namely, Fig. 2, 212, 214, Fig. 3, 302-306, 314, 316); regarding using the machine learning model: 6:59-76, 15:7-16 see quotations in previous bullet point (determining step) immediately above)
predict, using the machine learning model, an outcome based on the … risk indicator, wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action; flagging the processed action for review; or allowing the processed action. (7:7-30 "in response to the captured data fraud risk being a “low risk,” (predict an outcome based on the … risk indicator) risk assessment engine 150 may transmit a verification approval to payment system 130. In response to receiving the verification approval, payment system 130 may remove any pending fraud statuses, flags, or the like from the transaction or associated transaction account. (wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action; flagging the processed action for review; or allowing the processed action) … As a further example, and in accordance with various embodiments, in response to the captured data fraud risk being a “high risk,” (predict an outcome based on the … risk indicator) risk assessment engine 150 may transmit a fraudulent verification notification to payment system 130. In response to receiving the fraudulent verification notification, payment system 130 may flag the associated transaction account for a fraud resolution follow-up. The flag may block all self-service channels (e.g., online account access) from verifying the transaction activity" (wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action; flagging the processed action for review; or allowing the processed action); 10:30-11:3 "For example, in response to the captured user device data comprising a “low risk,” (predict an outcome based on the … risk indicator) risk assessment engine 150 may transmit the verification approval to payment system 130. In response to receiving the verification approval, payment system 130 may remove any pending fraud statuses from the transaction account. (wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action; flagging the processed action for review; or allowing the processed action) … ¶ … For example, in response to the captured user device data comprising a “high risk,” (predict an outcome based on the … risk indicator) risk assessment engine 150 may transmit the fraudulent verification notification to payment system 130. In response to receiving the fraudulent verification notification, payment system 130 may flag the associated transaction account for a manual fraud resolution follow-up. Payment system 130 may also block all digital access to the transaction account (e.g., via an online portal or the like)." (wherein the outcome includes the at least one processor performing at least one of: automatically stopping the processed action; flagging the processed action for review; or allowing the processed action); regarding using the machine learning model: 6:59-76, 15:7-16 see quotations at determining step above)
Beckman does not explicitly disclose that the risk indicator is a three-digit number, but Poduval teaches:
… a numerical risk indicator, … wherein the numerical risk indicator (chargeback (fraud) risk probability score) is a three-digit indicator associated with unauthorized activity; and (0097; note although Poduval refers to "chargeback risk probability score" and the like terminology, Poduval's disclosure (and terminology) are deemed to deal with (and refer to) -- and in any event are applicable to -- fraud risk determination, in view of Poduval's teachings, such as: 0034 "One of the most common reasons for the chargeback is fraud."; 0037 "The set of transaction indicators includes, … fraud risk features, …. " (note the transaction indicators are used to generate the transaction features that are inputted into the machine learning models to predict chargeback/fraud, see 0037, 0067-0075, 0079-0080, 00083, 0087, 0108, 0111-0112); 0038 "Furthermore, the server system is configured to implement or run a chargeback risk prediction model to compute a set of chargeback risk probability scores corresponding to one or more time intervals associated with the account holder based, at least in part, on the set of transaction features. … Moreover, the server system is configured to transmit a notification to an issuer server associated with the account holder based, at least in part, on the set of chargeback risk probability scores. In an example, the issuer server may analyze the set of chargeback risk probability scores to perform one or more downstream tasks (e.g., prediction of fraudulent payment transactions, etc.)."; 0070 "The set of transaction features may be determined from or engineered from the payment transaction data of the past payment transactions. … The set of transaction features includes at least one of: spend transaction features, merchant features of a plurality of merchants involved in the payment transactions, and fraud risk features. … In an example, the fraud risk features are generated based on the payment transactions performed due to fraud."; 0073 "In one example, the fraud risk features may include total fraud amount for card-not-present cross-border payment transactions performed in 1 month, 3 months, and so on, chargeback amount for fraudulent payment transactions performed at a merchant in 1 month, 3 months, and so on, and the like.")
… the three-digit risk indicator. (0097, 0109-0111; regarding using the machine learning model: 0033, 0038, 0040, 0075, 0084, 0108, 0112, 0124, 0132 machine learning model, GBDT model)
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified Beckman's systems and methods for determining fraud risk, by incorporating therein these teachings of Poduval regarding use of a three-digit number as a risk score, because it would provide for more comprehensive, fine-tuned risk scoring and consequent responsive actions compared to Beckman (Beckman merely teaches a tripartite risk classification of "high risk," "medium risk," and "low risk" and mentions in another context fraud scores without further specifying the nature of the scores). This more comprehensive, fine-tuned process would allow for more precise risk scores/classifications and thus would be more likely to treat transactions appropriately according to their actual risk level, thus leading to more effective and satisfactory outcomes, and thus amounts to an improvement upon Beckman.
Regarding Claims 23 and 33
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 as set forth above. Beckman further teaches:
wherein the machine learning model predicts the outcome …. (7:7-30, 10:30-11:3 see quotations at predicting step of claims 22, 32 and 41, above; regarding using the machine learning model: 6:59-76, 15:7-16 see quotations at determining step of claims 22, 32 and 41, above)
Poduval further teaches:
… based on a comparison of the three-digit risk indicator with at least one threshold. (0090, 0109-0110, 0145 risk score is compared to threshold; based on this comparison, notification (alert) is transmitted (outcome))
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score, by incorporating therein these further teachings of Poduval regarding comparing a risk score to a threshold, because it provides for increased precision (finer resolution) in making risk classifications in support of further action and thus should improve outcomes (should yield more proper actions taken).
Regarding Claims 25 and 35
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 as set forth above. Beckman further teaches:
wherein the probability that the processed action belongs to the class is based on at least one of transactional data, a customer characteristic, or historical data. (8:13-27 determination as to whether transaction is fraudulent (determination of probability that the processed action belongs to the class) may be based on "data regarding the transaction account associated with user 101 to check if the account is active, has been recently flagged for fraud, and/or the like. Payment system 130 may retrieve recent purchases and determine whether the geographical codes align with the geographical code of the transaction authorization request (e.g., user 101 purchases goods in Spain and Brazil on the same day)" -- this data teaches transactional data, a customer characteristic, or historical data.)
Regarding Claims 27 and 37
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 as set forth above. Poduval further teaches:
wherein the three-digit risk indicator is derived from a model probability. (0097 "In one implementation, each chargeback risk probability score is a three-digit numeric value ranging from 001 to 999, indicative of the probability of chargeback to be experienced for the future payment transactions to be performed by the account holders. In an example, the chargeback risk probability score of 234 for the time interval 0-6 months indicates that there is a probability of 23.4% that the account holder 104 will raise the chargeback request in the next 6 months. In another example, the chargeback risk probability score of 876 for the time interval 0-12 months indicates that there is a probability of 87.6% that the account holder 104 will raise the chargeback request in the next 12 months." -- these scores (risk indicators) 234 and 876 are merely representations of the probabilities of 23.4% and 87.6%, respectively; regarding a model probability: as per 0033, 0038, 0040, 0075, 0084, 0108, 0124, 0132, the probability is determined by a machine learning/statistical/AI/ GBDT model, hence the probability is a model probability)
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score, by incorporating therein these further teachings of Poduval regarding the three-digit number being derived from a model probability, because it is appropriate for a fraud detection/prevention system, and will yield proper results, if the risk score / risk of fraud represents, and hence is derived from, the probability of fraud, and because where the fraud detection/prevention system uses a model to determine probability / risk of fraud, the probability will be generated by the model.
Regarding Claims 30 and 40
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 as set forth above. Beckman further teaches:
wherein the class is generated by the machine learning model.(8:13-64, Fig. 2, 208 payment server 130 "[determines] whether the transaction may be fraudulent," "using any suitable technique and fraud detection process," based on risk factors as described at 8:20-27. The output of this process is a determination that the transaction is not fraudulent (8:28-37) or a determination that the transaction is fraudulent (8:38-64), i.e., an assignment of the transaction to a class (either the class of fraudulent transactions or the class of not fraudulent transactions), in other words, the process outputs or generates the class (or generates the output, which is the class) to which the transaction is assigned; regarding by the machine learning model: 6:59-76 "risk assessment engine 150 may implement statistical models, machine learning, artificial intelligence, and the like to aid in identifying possible fraud. In that regard, the captured user device data may be input into the statistical model, the machine learning model, or the artificial intelligence model to determine a risk of fraud."; 15:7-16 "any of the operations may be conducted or enhanced by … machine learning")
Claims 24, 28, 34 and 38 are rejected under 35 U.S.C. 103 as being unpatentable over Beckman et al. (U.S. Patent No. 10,872,341 B1), hereafter Beckman, in view of Poduval et al. (U.S. Patent No. 2022/0358507 A1), hereafter Poduval, and further in view of Melul et al. (U.S. Patent Application Publication No. 2022/0198470 A1), hereafter Melul.
Regarding Claims 24 and 34
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 and intervening claims 23 and 33 as set forth above. Beckman in view of Poduval does not explicitly disclose but Melul teaches:
wherein the at least one threshold is determined by the machine learning model based on at least one of a deposit type, a deposit amount, a deposit location, or a fraud history. (0022 Melul incorporates by reference Amitai (U.S. Patent Application Publication No. 2022/0044248 A1; U.S. Application No. 16/985,773); Amitai, 0026-0034, 0040-0049: Amitai, 0026, teaches that a rule set has thresholds. As such, different rule sets have different thresholds, and changing a rule set changes the thresholds. Changing a threshold amounts to setting or determining a new threshold. Amitai, 0028, 0031, teaches that a draft rule set is refined, by tuning the model, and so transformed into a model rule set. (Alternatively, Amitai, 0048, teaches that a new rule set may be adopted in the model tuning process.) Therefore, Amitai teaches that thresholds are changed, i.e., new thresholds are determined, by tuning the model. Further, per Amitai, 0040-0049 (describing Fig. 2), tuning the model includes calculating an "accuracy metric … by calculating the percentage of False Positives and False Negatives" (0043, 0048) -- as such, since the false positives and negatives constitute an aspect of the fraud history, the tuning of the model is based on fraud history. Thus, the transformation of the rules, e.g., of the draft rules into model rules, and the concomitant changing of thresholds or determination of new thresholds, is based on a fraud history. Regarding by the machine learning model: Melul 0050 "the constant threshold [i.e., the threshold to which confidence (fraud risk) scores are compared] is adjusted through machine learning techniques"; note, under broadest reasonable interpretation, adjustment of a threshold teaches setting or determining a new threshold)
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score and comparing the risk score to a threshold, by incorporating therein these further teachings of Melul/Amitai regarding a model tuning process, including determining new thresholds based on a fraud history, that serves to increase model accuracy and quality by minimizing false negatives and positives, because incorporating such a model tuning process, including determining new thresholds based on a fraud history, would further improve model accuracy/quality and hence model results. See Melul, 0023 (Amitai, 0032, 0043, 0048; note Amitai, 0043, 0048 are continuations/elaborations of the content of Amitai, 0026-0034, cited in the rejection, as explained at Amitai, 0034 ("The tuning software 103 is described in more detail in the discussion of FIG. 2 below [namely, 0040-0049].")).
Regarding Claims 28 and 38
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 as set forth above. Beckman in view of Poduval does not explicitly disclose but Melul teaches:
wherein the machine learning model is tuned based on information associated with at least one of the processed action of the user, the probability, the three-digit risk indicator, or the outcome. (0022 "… once the artificial neural network model is generated, transactions seen on the rail 106 [the processed action of the user] are used to tune the production model, …. In some cases, … the production model 104 is re-tuned 103 periodically.", 0023 "The model tuning software 103 outputs a production model 104 that is tuned by the latest transaction received from the rail 106 [the processed action of the user].")
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score, by incorporating therein these teachings of Melul regarding periodically tuning a machine learning model used to detect fraudulent transactions, because periodically tuning a machine learning model is necessary to keep the model up to date with the latest transaction data and therefore periodic tuning maintains the good performance of the model and the accuracy of the model results and, as such, would improve upon a system/method that does not perform tuning. Melul, 0023.
Claims 26 and 36 are rejected under 35 U.S.C. 103 as being unpatentable over Beckman et al. (U.S. Patent No. 10,872,341 B1), hereafter Beckman, in view of Poduval et al. (U.S. Patent No. 2022/0358507 A1), hereafter Poduval, and further in view of Shevyrev et al. (U.S. Patent Application Publication No. 2023/0281629 A1), hereafter Shevyrev.
Regarding Claims 26 and 36
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 and intervening claims 25 and 35 as set forth above. Beckman in view of Poduval does not explicitly disclose but Shevyrev teaches:
retaining, by the machine learning model (Fig. 4, 308; 0028, 0078), information associated with one or more previously generated risk indicators, wherein the information includes at least one of the transactional data, the customer characteristic, or the historical data; and tuning the numerical risk indicator based on the information, wherein the tuning identifies optimal values of one or more hyperparameters to maximize performance of the machine learning model. (0080-0082: 0081 "Further, the loss function 408 can return quantifiable data regarding the difference between a given training … prediction from the training … predictions 405 and a corresponding ground truth from the ground truth check data 406 [retaining, by the machine learning model, information associated with one or more previously generated risk indicators]. In particular, the loss function 408 can return losses 410 to the … machine-learning model 308 [retaining, by the machine learning model, information associated with one or more previously generated risk indicators] based upon which the … system 102 adjusts various parameters/hyperparameters [wherein the tuning identifies optimal values of one or more hyperparameters]. In so doing, the … system 102 can improve the quality/accuracy of training … predictions in subsequent training iterations—by narrowing the difference between training … predictions and ground truth … data in subsequent training iterations [tuning the numerical risk indicator based on the information, wherein the tuning identifies optimal values of one or more hyperparameters to maximize performance of the machine learning model]."; further regarding wherein the information includes at least one of the transactional data, the customer characteristic, or the historical data; and tuning the numerical risk indicator based on the information: as per Fig. 4, 402, Table 1 (0077), Fig. 3, 304 (e.g., 0053-0054), Fig. 2, 204 (e.g., 0043), the features (model inputs) that are used to generate the output probability/prediction scores (Fig. 4, 308, 405, (e.g., 0078), Fig. 3, 306 (e.g., 0064), Fig. 2, 206 (e.g., 0044)) include at least one of the transactional data, the customer characteristic, or the historical data; one of ordinary skill in the art understands that the adjustment of parameters (e.g., adjusting of weights assigned to inputs)/hyperparameters is based on consideration of the outputs generated in light of the inputs used, and therefore the features/inputs are included in or considered along with the "quantifiable data regarding the differences …" (0081))
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score, by incorporating therein these teachings of Shevyrev regarding adjusting parameters/hyperparameters of a machine learning model based on a loss function (data regarding difference between training prediction and ground truth), because such tuning of a machine learning model is necessary to keep the model up to date with the latest transaction data and therefore such tuning maintains the good performance of the model and the accuracy of the model results and, as such, would improve upon a system/method that does not perform tuning, see Shevyrev, 0080-0082.
Claims 29 and 39 are rejected under 35 U.S.C. 103 as being unpatentable over Beckman et al. (U.S. Patent No. 10,872,341 B1), hereafter Beckman, in view of Poduval et al. (U.S. Patent No. 2022/0358507 A1), hereafter Poduval, and further in view of Lu (U.S. Patent Application Publication No. 2019/0197442 A1).
Regarding Claims 29 and 39
Beckman in view of Poduval teaches the limitations of base claims 22 and 32 as set forth above. Beckman in view of Poduval does not explicitly disclose but Lu teaches:
wherein the machine learning model is benchmarked based on a relative precision. (0054 "data preprocessor 210 includes implementation of … performance benchmarking techniques to compare the performance measures such as accuracy, precision and recall"; note Applicant's specification 0086 defines "relative precision" as "[t]he number of true positives divided by the sum of the true positives and false negatives"; Lu 0054 teaches, inter alia, benchmarking based on "recall"; the definition of "recall" in the art is the same as Applicant's definition of "relative precision"5; therefore, Lu's teaching of "recall" teaches Applicant's recitation of "relative precision")
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score, by incorporating therein these teachings of Lu regarding benchmarking a model based on performance measures such as accuracy, precision and recall, because it would improve model performance/results by having the model meet a standard/criterion/threshold.
Claim 31 is rejected under 35 U.S.C. 103 as being unpatentable over Beckman et al. (U.S. Patent No. 10,872,341 B1), hereafter Beckman, in view of Poduval et al. (U.S. Patent No. 2022/0358507 A1), hereafter Poduval, and further in view of Chisholm (U.S. Patent Application Publication No. 2014/0351137 A1).
Regarding Claim 31
Beckman in view of Poduval teaches the limitations of base claim 22 as set forth above. Beckman in view of Poduval does not explicitly disclose but Chisholm teaches:
wherein the processed action of the user is enriched in real time. (0042, 0101, claims 8 and 14; regarding in real time: 0024, 0041, 0042, 0044, 0049, 0054, 0101, the enrichment is performed by the decisioning platform (e.g., 0042, 0101, Fig. 4, preprocessor 204), which performs real-time fraud scoring/fraud prediction of transactions, such that the fraud scoring/fraud prediction can be and is used to decide whether to approve or decline a transaction (the candidate transaction currently being evaluated by the system (0049)) in real time)
It would have been obvious to one of ordinary skill in the art not later than the effective filing date of the claimed invention to have modified the combination of Beckman's systems and methods for determining fraud risk, as modified by Poduval's teachings regarding use of a three-digit number as a risk score, by incorporating therein these teachings of Chisholm regarding enriching a transaction in real time to facilitate a process of determining whether a transaction is fraudulent, based on the following reasoning:
Beckman (e.g., 10:25-31) teaches using other data (risk factors) in assessing risk/generating a risk score, which other data (risk factors) include data comparable to Chisholm's historical and cardholder data with which a transaction is enriched, but Beckman does not teach "enriching" the transaction (the transaction record that is being analyzed for risk of fraud) with this other data (risk factors). Thus, while Beckman uses this other data (risk factors) in performing the risk assessment, Beckman does not provide implementation detail as to how this other data (risk factors) (e.g., "historical data and non-device related attributes" (10:25-31)) is obtained so that it is available for use in performing the fraud risk assessment. However, Chisholm's enrichment process provides implementation detail appropriate to implement this under-specified aspect of Beckman's systems and methods, as Chisholm's enrichment constitutes a known way of having other supplementary data at hand together with the primary data under test, for use in analyzing the primary data, as Beckman requires. In addition, the combination (incorporation of Chisholm's teaching of enrichment into Beckman) would have predictable results, e.g., the enrichment can be incorporated into Beckman in a mechanical-like manner without adversely affecting any other relevant aspects of Beckman. Thus, combining Chisholm's teaching of enrichment with Beckman provides implementation detail (namely, for having the risk factors at hand for use in the risk analysis) that permits Beckman to actually perform its intended function of assessing fraud risk of transactions so as to allow/provide for appropriate remedial actions.
Note although Beckman's operations are not explicitly described as being performed in real-time, as best understood Beckman's fraud detection/risk scoring is performed in real-time such that the fraud score/prediction is used to decide whether to approve or decline a current transaction. On this understanding, the fact that Chisholm's enrichment is performed in real-time aligns with and serves Beckman's requirements. On the other hand, if this understanding is incorrect, Chisholm's real-time enrichment would facilitate improved performance by Beckman, by permitting transactions to be enriched in real time, thus permitting fraud scoring / prediction that is both more accurate (on account of the presence/inclusion of the enriched data) and that can be performed in real-time rather than after the fact.
Conclusion
The prior art made of record and not relied upon, as set forth in the accompanying Notice of References Cited (PTO-892), is considered pertinent to applicant's disclosure.
Comeaux (US-11669844-B1) teaches evaluating a transaction for fraud, including generating an alert probability score (fraud risk score), based on a wide variety of risk factors (e.g., behavioral profile including user personal data, financial data, and user social network data; historical user financial data), generating an alert, and taking action to prevent processing of a fraudulent transaction, including using and training a machine learning model.
Comeaux (US-10567402-B1) and Comeaux (US-11722502-B1) teach fraud detection/prevention similar to Comeaux (US-11669844-B1) but to greater depth in certain aspects.
Phatak (US-2022/0006899-A1) and Anderson (US-12136096-B1) teach a fraud alert queue that prioritizes fraud alerts based on fraud importance.
Vaswani (US-2022/0377090-A1) teaches fraud detection/prevention (including risk scores and alerts) similar to Comeaux (US-11669844-B1).
Karpovsky (US-2022/0191173-A1) teaches determining fraud risk based on VPN and/or proprietary knowledge and periodic monitoring.
Pavlovic ("Log-normal Distribution - A simple explanation”) teaches content about log-normal distribution similar to that of Applicant's disclosure (specification paragraph 0045).
Vimal (US-2023/0186311-A1) (qualifying as prior art based on Indian priority date) teaches, inter alia, benchmarking a machine learning model based on precision, recall, F1, and/or F2 scores, see 0097.
Thomas (US-10997596-B1) teaches appending a fraud accuracy tag to a declined transaction, where the fraud accuracy tag is indicative of whether the decline of the transaction is a true positive decline or a false positive decline, whereby the fraud accuracy tag is suitable to provide insight into accuracy of a fraud strategy implemented in connection with the declined transaction.
Beckman (US-10872341-B1) teaches secondary fraud detection during transaction verification, where the transaction verification process with the user itself is evaluated and scored for likelihood of fraud, using machine learning, and including assigning designations of high risk, medium risk, and low risk.
Selway (US-2013/0013491-A1) teaches evaluating a transaction for fraud, including generating a risk score, based on risk factors, generating an alert, and taking remedial action to prevent processing of a fraudulent transaction, including using neural models, and where the transaction may be received from a specified one of several specified transaction channels (e.g., bank, merchant, ATM, remote, etc.) and the alert/notification may be sent to the same transaction channel or to any of the transaction channels.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DOUGLAS W PINSKY whose telephone number is (571)272-4131. The examiner can normally be reached on 8:30 am - 5:30 pm ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jessica Lemieux can be reached on 571-270-3445. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/DOUGLAS W PINSKY/
Examiner, Art Unit 3626
/EMMETT K. WALSH/Primary Examiner, Art Unit 3626
1 Note the subject matter argued here is claimed in dependent claims 26 and 36, not in the independent claims.
2 MEMORANDUM to Patent Examining Corps, "Advance notice of change to the MPEP in light of Ex Parte Desjardins," December 5, 2025, p. 4.
3 Note Applicant's instant remarks confirm this unordinary meaning Applicant ascribes to the term "predict":
As described in the specification, the machine learning model executes within the financial institution's computing infrastructure and generates risk indicators that directly control system behavior, such as automatically allowing a transaction, placing the transaction on hold, or flagging the transaction for analyst review. See, e.g., Specification at ¶¶[0055], [0058] - [0059], and [0071]. The specification explains that, based on the generated risk indicator, the processor automatically enforces outcomes including allowance, analyst review, or auto-holding of the transaction. Id. These actions are performed by the system itself, rather than merely presenting information for human evaluation. Accordingly, the claims recite a machine-implemented control mechanism, not a human decision-making process or a fundamental economic practice. (Response, p. 15; emphasis added)
4 Note in the rejection in the previous Office Action Poduval was cited as teaching additional claimed subject matter that constitutes context surrounding the teaching that the risk indicator is "a three-digit number." However, that context is taught by Beckman.
5 See, e.g., Wilber ("Precision and Recall"), p. 9.
Prosecution Insights