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 .
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claim 1 is a process type claim. Claim 12 is a machine type claim. Claim 18 is a manufacture type claim. Therefore, claims 1-20 are directed to either a process, machine, manufacture or composition of matter.
As per claim 1,
2A Prong 1:
“detecting attributes and values in rules contained in a rules set, wherein the attributes comprise condition attributes and result attributes” A user mentally or with pencil and paper looks at rules from a rules set.
“determining definitions of the attributes detected in the rules from a data model that includes data objects with attributes that map to the attributes in the rules of the rules set” The user mentally or with pencil and paper identifies attributes in the rules using a data model associated with the rules and matches up the rules to the appropriate attributes.
“generating an unlabeled dataset comprising multiple different unlabeled data entries having fields associated with the condition attributes detected in the rules and containing data values associated with the condition attributes, the generating comprising populating the fields associated with the condition attributes with the data values according to the values detected in the rules and the definitions of the condition attributes determined from the data model” The user mentally or with pencil and paper creates and fills in the appropriate fields with values and attributes taken from the rule set.
“forming a labeled dataset using the unlabeled data entries and logic contained in the rules set such that the labeled dataset comprises multiple different labeled data entries comprising fields and data values of the unlabeled data entries and labels, wherein the forming comprises determining the labels using rules of the rules set that match the unlabeled data entries such that the labels are derived from the result attributes of the rule” The user mentally or with pencil and paper creates and labels a dataset using the rule set.
“training a … model using at least a portion of the labeled dataset” The user mentally or with pencil and paper works out the mathematics to make their model operate.
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
“computer” (mere instructions to apply the exception using a generic computer component);
“a machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: The claimed machine learning model is a generic machine learning model with no details or limitations that make it beyond a generic, off the shelf machine learning model.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
“computer” (mere instructions to apply the exception using a generic computer component)
“a machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: The claimed machine learning model is a generic machine learning model with no details or limitations that make it beyond a generic, off the shelf machine learning model.
As per claims 2-7 contain additional mental steps similar to claim 1, and are rejected similarly to claim 1.
As per claims 8-11, these claims contain additional mental steps and generic machine learning steps similar to claim 1, and are rejected for similar reasons to claim 1.
As per claim 12,
2A Prong 1:
“detecting attributes and values in rules contained in a rules set wherein the attributes comprise condition attributes and result attributes” A user mentally or with pencil and paper looks at rules from a rules set.
“determining definitions of the attributes detected in the rules contained in the rules set from a data model that includes data objects with attributes that map to the attributes in the rules of the rules set” The user mentally or with pencil and paper identifies attributes in the rules using a data model associated with the rules and mapping the values that match together.
“generating an unlabeled dataset comprising multiple different unlabeled data entries having fields associated with the condition attributes detected in the rules and containing data values associated with the condition attributes, the generating comprising populating the fields associated with the condition attributes with the data values according to the values detected in the rules and the definitions of the condition attributes determined from the data model” The user mentally or with pencil and paper creates and fills in the appropriate fields with values and attributes taken from the rule set.
“forming a labeled dataset using the unlabeled data entries and logic contained in the rules set wherein the labeled dataset comprises multiple different labeled data entries comprising fields and data values of the unlabeled data entries and labels, wherein the forming comprises selecting a unlabeled data entry from the unlabeled data entries, executing the rules set on the unlabeled data entry to obtain a result, and using the result as a label for the unlabeled data entry such that the label is derived from the result attributes of the rule” The user mentally or with pencil and paper creates and labels a dataset by mentally or with pencil and paper executing rules and using the output as a label.
“forming a training dataset from the labeled dataset” The user mentally or with pencil and paper identifies the labeled dataset as a training dataset.
“applying the training dataset to a … model during training of the … model” The user mentally or with pencil and paper works out the mathematics to make their model operate.
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
“computing system”, “one or more processing units coupled to memory”, “one or more computer readable storage media” (mere instructions to apply the exception using a generic computer component);
“a machine learning model” and “the machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: The claimed machine learning model is a generic machine learning model with no details or limitations that make it beyond a generic, off the shelf machine learning model.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
“computing system”, “one or more processing units coupled to memory”, “one or more computer readable storage media” (mere instructions to apply the exception using a generic computer component)
“a machine learning model” and “the machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: The claimed machine learning model is a generic machine learning model with no details or limitations that make it beyond a generic, off the shelf machine learning model.
As per claims 13-15 contain additional mental steps similar to claim 1, and are rejected similarly to claim 12.
As per claims 16-17, these claims contain additional mental steps and generic machine learning steps similar to claim 1, and are rejected for similar reasons to claim 12.
As per claim 18,
2A Prong 1:
“detecting attributes and values in rules contained in a rules set, wherein the attributes comprise condition attributes and result attributes” A user mentally or with pencil and paper looks at rules from a rules set.
“determining definitions of the attributes detected in the rules contained in the rules set from a data model the rules set, that includes data objects with attributes that map to the attributes in the rules of the rules set” The user mentally or with pencil and paper identifies attributes in the rules using a data model associated with the rules.
“generating an initial dataset comprising multiple different initial data entries having fields associated with the condition attributes detected in the rules and containing data values associated with the condition attributes, the generating comprising populating the fields with the data values according to the values detected in the rules and the definitions of the condition attributes determined from the data model” The user mentally or with pencil and paper creates and fills in the appropriate fields with values and attributes taken from the rule set.
“forming a labeled dataset, using the initial data entries and logic contained in the rules set wherein the labeled dataset comprises multiple different labeled data entries comprising fields and data values of the initial data entries and labels, wherein the forming comprises selecting an initial data entry from the initial data entries, executing the rules set on the initial data entry to obtain a result, and using the result as a label for the initial data entry wherein the label is derived from the result attributes of the rule” The user mentally or with pencil and paper creates and labels a dataset by mentally or with pencil and paper executing rules and using the output as a label.
“forming a training dataset from the labeled dataset” The user mentally or with pencil and paper identifies the labeled dataset as a training dataset.
“applying the training dataset to a … model during training of the … model” The user mentally or with pencil and paper works out the mathematics to make their model operate.
“Wherein populating the fields with data values comprises selecting values randomly” The user mentally or with pencil and paper chooses random values to fill the fields.
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
“non-transitory computer readable storage media” (mere instructions to apply the exception using a generic computer component);
“a machine learning model” and “the machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: The claimed machine learning model is a generic machine learning model with no details or limitations that make it beyond a generic, off the shelf machine learning model.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
“non-transitory computer readable storage media” (mere instructions to apply the exception using a generic computer component)
“a machine learning model” and “the machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: The claimed machine learning model is a generic machine learning model with no details or limitations that make it beyond a generic, off the shelf machine learning model.
As per claims 19-20 contain additional mental steps similar to claim 1, and are rejected similarly to claim 1.
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 (i.e., changing from AIA to pre-AIA ) 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.
Claims 1-4, 8 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Taylor et al (“ClassBench: A Packet Classification Benchmark”) in view of Bhatia et al (US 20200272741 A1).
As per claim 1, Taylor discloses, “a computer-implemented method comprising” (Pg.2078, particularly section VI; EN: this denotes monitoring load, power consumption, and other aspects of a computer system running programs, which inherently includes some form of computer/processor/memory to run the system).
“Detecting attributes” (Pg.2070, particularly C2, section B; EN; This denotes looking at filter sets (i.e. rules) and what attributes they had, such as protocols, port ranges, port pair class, etc). “and values” (Pg.2070, particularly C2, section B; EN: This denotes the values found in these things such as TCP IP, types of port ranges, etc). “in rules contained in a rule set” (Pg.2070, C1, Section III; EN: this denotes looking at premade filter sets to analyze them). “wherein the attributes comprise condition attributes” (Pg.2070, particularly section B; EN: This denotes protocols, port ranges, and port pair class, all of which are conditions of incoming data). “and result attributes” (Pg.2068, particularly C1, introduction section; EN: this denotes using the filter sets to apply security policies, application processing, and QoS guarantees, all of which are examples of results).
“determining definitions of the attributes detected in the rules contained in the rules set from a data model that includes data objects with attributes that map to the attributes in the rules of the rules sets” (Pg.2070, particularly C2, Section B; EN: this denotes the system identifying the different aspects of the rules).
“generating an unlabeled dataset comprising multiple different unlabeled data entries having fields…” (Pg.2074, particularly section V; EN: this denotes creating new filter sets based on statistical values and distributions from a parameter file. This is “unlabeled” because the rules are created via statistics and are not made for a specific purpose). “associated with the condition attributes detected in the rules” (Pg.2073, particularly C1, section Iv; EN: this denotes making the parameter file based off of real filter sets). “and containing data values associated with the condition attributes” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters). “the generating comprising populating the fields associated with the condition attributes with the data values according to the values detected in the rules and definitions of the condition attributes determined from the data model” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters).
“Forming a … dataset using the unlabeled data entries and logic contained in the rules set, wherein the … dataset comprises multiple different … data entries comprising fields and data values of the unlabeled data entries and labels…wherein the forming comprises using rules of the rules set that match the unlabeled data entries…” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters based on the parameter set generated from the real filter set).
However, Taylor fails to explicitly disclose, “forming a labeled dataset using the unlabeled data entries … wherein the labeled dataset comprises multiple different labeled entries comprising fields and data values of the unlabeled data entries and labels, wherein the forming comprises determining the labels” and “training a machine learning model using at least a portion of the labeled dataset”
Bhatia discloses “forming a labeled dataset using the unlabeled data entries … wherein the labeled dataset comprises multiple different labeled entries comprising fields and data values of the unlabeled data entries and labels, wherein the forming comprises determining the labels” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc).
“training a machine learning model using at least a portion of the labeled dataset” (Pg.4, particularly paragraph 0043; EN: this denotes using the rules to train a recurrent neural network using supervised learning, which makes use of labels in the training set).
Taylor and Bhatia are analogous art because both involve network rule creation.
Before the effective filing date it would have been obvious to one skilled in the art of network rule creation to combine the work of Taylor and Bhatia in order to properly label and import rules in order to use them to train a machine learning algorithm.
The motivation for doing so would be to “use[] natural language processing (NLP techniques … to identify and eliminate duplicate rules, combine similar rules together into ‘super rules’ align STEM rules with frameworks and/or standard rules from standard rules repositories, decompose the rules and their conditions into principal components for use in automatically generating new STEM rules, and train a machine learning model, such as a Recurrent Neural Network (RNN) to generate automated rules based on specific threat intelligence and learning of rule components that correspond to threat characteristics” (Bhatia, Pg.7, paragraph 0059) or in the case of Taylor, allow the system to fully parse and understand the pieces of the generated rules in order to improve the system via machine learning for rule generation.
Therefore before the effective filing date it would have been obvious to one skilled in the art of network rule creation to combine the work of Taylor and Bhatia in order to properly label and import rules in order to use them to train a machine learning algorithm.
As per claim 2, Taylor discloses, “wherein the definitions determined from the data model comprise value domains for the attributes detected in the rules” (Pg.2073, particularly C2, the bullet points; EN: this denote show each of the different types of data are determined and what kind of values they hold). “wherein generating the unlabeled dataset comprises determining permissible values for the fields based on the values detected in the rules and the value domains from the data model” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each).
As per claim 3, Taylor discloses, “wherein generating the unlabeled dataset further comprises randomly assigning values to the fields of the unlabeled data entries from the permissible values” (Pg.2074, particularly C2, last paragraph; EN: this denotes using random variables to create the data).
As per claim 4, Taylor discloses, “Wherein generating the unlabeled dataset further comprises assigning values to the unlabeled data entries within the permissible values” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each). “and from existing data with values for the attributes and permissible values” (Pg.2073, particularly C1, section Iv; EN: this denotes making the parameter file based off of real filter sets).
As per claim 8, Taylor discloses, “receiving a new rules set and a new data model associated with the new rules set” (Pg.2070, particularly C1, section III; EN: this denotes working with multiple different rule sets in different formats (i.e. different data models)).
“forming a new … dataset from the new rules set and the new data model” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters based on the parameter set generated from the real filter set).
Bhatia discloses, “forming a new labeled dataset from the new rules set and the new data model” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc).
“and … training the machine learning model with the new labeled dataset” (Pg.4, particularly paragraph 0043; EN: this denotes using the rules to train a recurrent neural network).
However, Taylor and Bhatia fail to explicitly disclose “retraining.” However, the Examiner takes official notice that it would be obvious to one of ordinary skill at the time of filing to retrain a machine learning model based upon newly available data, as this allows the system to be kept up to date on new developments related to incoming data as needed.
As the Applicant failed to argue or respond to the official notice in the response filed 6/13/2025, the aspects of the official notice are applicant admitted prior art as per MPEP 2144.03(C).
As per claim 11, Bhatia discloses, “further comprising making a prediction using the machine learning model” (pg.12, particularly paragraph 0096; EN: this denotes predicting potential rule components for new rules).
Claim Rejections - 35 USC § 103
Claims 5-7 are rejected under 35 U.S.C. 103 as being unpatentable over Taylor et al (“ClassBench: A Packet Classification Benchmark”) in view of Bhatia et al (US 20200272741 A1) and further in view of Rogers et al (US 20190147297 A1).
As per claim 5, Taylor discloses, “within the permissible values” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each).
Bhatia discloses, “wherein forming the labeled data set comprise selecting an unlabeled data entry from the unlabeled data entries…” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc).
However, Taylor modified by Bhatia fails to explicitly disclose, “executing the rules set on the unlabeled data entry to obtain a result, and using the result as a label for the unlabeled data entry.”
Rogers discloses, “executing the rules set on the unlabeled data entry to obtain a result, and using the result as a label for the unlabeled data entry” (Pg.2, particularly paragraph 0019; EN: this denotes using rules to label training data).
Rogers and Taylor modified by Bhatia are analogous art because both involve machine learning.
Before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Rogers and Taylor modified by Bhatia in order to use rules to label training data.
The motivation for doing so would be to use “so-called weak supervision techniques … which can be used to concurrently label multiple training data” (Rogers, pg.2, paragraph 0019) in order to provide “accurately labeled data” which can be “daunting, time consuming, and resource intensive task” (Rogers, Pg.2, paragraph 0019) or in the case of Taylor modified by Bhatia, allow the system to have clearly labeled rules from unlabeled rules to perform the supervised learning of the Taylor/Bhatia combination.
Therefore before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Rogers and Taylor modified by Bhatia in order to use rules to label training data.
As per claim 6, Rogers discloses, “Wherein executing the rules set on the unlabeled data entry to obtain a result comprises finding a rule in the rules set having a set of conditions that matches the unlabeled data entry and applying the found rule to the unlabeled data entry” (Pg.2, particularly paragraph 0019; EN: this denotes using rules to label training data).
As per claim 7, Rogers discloses, “wherein using the result as a label for the unlabeled data entry comprises adding the result to the unlabeled data entry to form a labeled data entry” (Pg.2, particularly paragraph 0019; EN: this denotes using rules to label training data).
Claim Rejections - 35 USC § 103
Claims 9-10 are rejected under 35 U.S.C. 103 as being unpatentable over Taylor et al (“ClassBench: A Packet Classification Benchmark”) in view of Bhatia et al (US 20200272741 A1) and further in view of Korjani et al (US 20160179751 A1).
As per claim 9, Taylor modified by Bhatia fails to explicitly disclose, “validating the machine learning model using at least a portion of the labeled dataset.”
Korjani discloses, “validating the machine learning model using at least a portion of the labeled dataset” (Pg.3, particularly paragraph 0025; EN: this denotes breaking up the data used to train the system into training data, validation data, and testing data, and using that to help train the system).
Korjani and Taylor modified by Bhatia are analogous art because both involve machine learning.
Before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Korjani and Taylor modified by Bhatia in order to include validation and test data.
The motivation for doing so would be to use the validating data set to “estimate generalization error” (Korjani, Pg.4, paragraph 0032) or in the case of Taylor modified by Bhatia, allow the training to include a validation set to help estimate generalization errors for the machine learning process.
Therefore before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Korjani and Taylor modified by Bhatia in order to include validation and test data.
As per claim 10, Taylor modified by Bhatia fails to explicitly disclose, “further comprising testing the machine learning model using at least a portion of the labeled dataset”
Korjani discloses, “further comprising testing the machine learning model using at least a portion of the labeled dataset” (Pg.3, particularly paragraph 0025; EN: this denotes breaking up the data used to train the system into training data, validation data, and testing data, and using that to help train the system).
Korjani and Taylor modified by Bhatia are analogous art because both involve machine learning.
Before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Korjani and Taylor modified by Bhatia in order to include validation and test data.
The motivation for doing so would be to use the validating data set to “evaluat[e] the non-linear variable structure regression model with the testing data subset” (Korjani, Pg.4, paragraph 0030) or in the case of Taylor modified by Bhatia, allow the training to include a test set to help test the model being trained.
Therefore before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Korjani and Taylor modified by Bhatia in order to include validation and test data.
Claim Rejections - 35 USC § 103
Claims 12-15, and 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Taylor et al (“ClassBench: A Packet Classification Benchmark”) in view of Bhatia et al (US 20200272741 A1) and Rogers et al (US 20190147297 A1).
As per claim 12, Taylor discloses, “A computing system comprising” (Pg.2078, particularly section VI; EN: this denotes monitoring load, power consumption, and other aspects of a computer system running programs, which inherently includes some form of computer/processor/memory to run the system).
“one or more processing units coupled to memory” (Pg.2078, particularly section VI; EN: this denotes monitoring load, power consumption, and other aspects of a computer system running programs, which inherently includes some form of computer/processor/memory to run the system).
“one or more computer readable storage media storing instructions that when executed by the one or more processing units cause the computing system to perform operations comprising” (Pg.2078, particularly section VI; EN: this denotes monitoring load, power consumption, and other aspects of a computer system running programs, which inherently includes some form of computer/processor/memory to run the system).
“Detecting attributes” (Pg.2070, particularly C2, section B; EN; This denotes looking at filter sets (i.e. rules) and what attributes they had, such as protocols, port ranges, port pair class, etc). “and values” (Pg.2070, particularly C2, section B; EN: This denotes the values found in these things such as TCP IP, types of port ranges, etc). “in rules contained in a rule set” (Pg.2070, C1, Section III; EN: this denotes looking at premade filter sets to analyze them). “wherein the attributes comprise condition attributes” (Pg.2070, particularly section B; EN: This denotes protocols, port ranges, and port pair class, all of which are conditions of incoming data). “and result attributes” (Pg.2068, particularly C1, introduction section; EN: this denotes using the filter sets to apply security policies, application processing, and QoS guarantees, all of which are examples of results).
“determining definitions of the attributes detected in the rules contained in the rules set from a data model that includes data objects with attributes that map to the attributes in the rules of the rules sets” (Pg.2070, particularly C2, Section B; EN: this denotes the system identifying the different aspects of the rules).
“generating an unlabeled dataset comprising multiple different unlabeled data entries having fields…” (Pg.2074, particularly section V; EN: this denotes creating new filter sets based on statistical values and distributions from a parameter file. This is “unlabeled” because the rules are created via statistics and are not made for a specific purpose). “associated with the condition attributes detected in the rules” (Pg.2073, particularly C1, section Iv; EN: this denotes making the parameter file based off of real filter sets). “and containing data values associated with the condition attributes” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters). “the generating comprising populating the fields associated with the condition attributes with the data values according to the values detected in the rules and definitions of the condition attributes determined from the data model” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters).
“forming a … dataset using the unlabeled data entries and logic contained in the rules set, wherein the … dataset comprises multiple different … data entries comprising fields and data values of the unlabeled data entries and labels” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters based on the parameter set generated from the real filter set). “
However, Taylor fails to explicitly disclose, “forming a labeled dataset using the unlabeled data entries … wherein the labeled dataset comprises multiple different labeled entries comprising fields and data values of the unlabeled data entries and labels, wherein the forming comprises selecting an unlabeled data entry from the unlabeled data entry, executing the rules set on the unlabeled data entry to obtain a result, and using the result as a label for the unlabeled data entry, and the label is derived from the result attributes of the rule”, “forming a training dataset from the labeled dataset”, and “applying the training dataset to a machine learning model during training of the machine learning model”
Bhatia discloses, “forming a labeled dataset using the unlabeled data entries … wherein the labeled dataset comprises multiple different labeled entries comprising fields and data values of the unlabeled data entries and labels” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc). “wherein the forming comprises selecting an unlabeled data entry from the unlabeled data entry” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc).
“forming a training dataset from the labeled dataset” (Pg.4, particularly paragraph 0043; EN: this denotes using the rules to train a recurrent neural network).
“applying the training dataset to a machine learning model during training of the machine learning model” (Pg.4, particularly paragraph 0043; EN: this denotes using the rules to train a recurrent neural network).
Rogers discloses, “executing the rules set on the unlabeled data entry to obtain a result, and using the result as a label for the unlabeled data entry, and the label is derived from the result attributes of the rule” (Pg.2, particularly paragraph 0019; EN: this denotes using rules to label training data).
Taylor and Bhatia are analogous art because both involve network rule creation.
Before the effective filing date it would have been obvious to one skilled in the art of network rule creation to combine the work of Taylor and Bhatia in order to properly label and import rules in order to use them to train a machine learning algorithm.
The motivation for doing so would be to “use[] natural language processing (NLP techniques … to identify and eliminate duplicate rules, combine similar rules together into ‘super rules’ align STEM rules with frameworks and/or standard rules from standard rules repositories, decompose the rules and their conditions into principal components for use in automatically generating new STEM rules, and train a machine learning model, such as a Recurrent Neural Network (RNN) to generate automated rules based on specific threat intelligence and learning of rule components that correspond to threat characteristics” (Bhatia, Pg.7, paragraph 0059) or in the case of Taylor, allow the system to fully parse and understand the pieces of the generated rules in order to improve the system via machine learning for rule generation.
Therefore before the effective filing date it would have been obvious to one skilled in the art of network rule creation to combine the work of Taylor and Bhatia in order to properly label and import rules in order to use them to train a machine learning algorithm.
Rogers and Taylor modified by Bhatia are analogous art because both involve machine learning.
Before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Rogers and Taylor modified by Bhatia in order to use rules to label training data.
The motivation for doing so would be to use “so-called weak supervision techniques … which can be used to concurrently label multiple training data” (Rogers, pg.2, paragraph 0019) in order to provide “accurately labeled data” which can be “daunting, time consuming, and resource intensive task” (Rogers, Pg.2, paragraph 0019) or in the case of Taylor modified by Bhatia, allow the system to have clearly labeled rules from unlabeled rules to perform the supervised learning of the Taylor/Bhatia combination.
Therefore before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Rogers and Taylor modified by Bhatia in order to use rules to label training data.
As per claim 13, Taylor discloses, “wherein the definitions determined from the data model comprise value domains for the attributes detected in the rules” (Pg.2073, particularly C2, the bullet points; EN: this denote show each of the different types of data are determined and what kind of values they hold). “wherein generating the unlabeled dataset comprises determining permissible values for the fields based on the values detected in the rules and the value domains from the data model” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each).
As per claim 14, Taylor discloses, “wherein generating the unlabeled dataset further comprises randomly assigning values to the fields of the unlabeled data entries from the permissible values” (Pg.2074, particularly C2, last paragraph; EN: this denotes using random variables to create the data).
As per claim 15, Taylor discloses, “Wherein generating the unlabeled dataset further comprises assigning values to the unlabeled data entries within the permissible values” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each). “and from existing data with values for the attributes and permissible values” (Pg.2073, particularly C1, section Iv; EN: this denotes making the parameter file based off of real filter sets).
As per claim 17, Bhatia discloses, “wherein the operations further comprise making a prediction using the machine learning model” (pg.12, particularly paragraph 0096; EN: this denotes predicting potential rule components for new rules).
As per claim 18, Taylor discloses, “One or more non-transitory computer readable storage media storing computer executable instructions for causing a computer system to perform operations comprising” (Pg.2078, particularly section VI; EN: this denotes monitoring load, power consumption, and other aspects of a computer system running programs, which inherently includes some form of computer/processor/memory to run the system).
“Detecting attributes” (Pg.2070, particularly C2, section B; EN; This denotes looking at filter sets (i.e. rules) and what attributes they had, such as protocols, port ranges, port pair class, etc). “and values” (Pg.2070, particularly C2, section B; EN: This denotes the values found in these things such as TCP IP, types of port ranges, etc). “in rules contained in a rule set” (Pg.2070, C1, Section III; EN: this denotes looking at premade filter sets to analyze them). “wherein the attributes comprise condition attributes” (Pg.2070, particularly section B; EN: This denotes protocols, port ranges, and port pair class, all of which are conditions of incoming data). “and result attributes” (Pg.2068, particularly C1, introduction section; EN: this denotes using the filter sets to apply security policies, application processing, and QoS guarantees, all of which are examples of results).
“determining definitions of the attributes detected in the rules contained in the rules set from a data model that includes data objects with attributes that map to the attributes in the rules of the rules sets” (Pg.2070, particularly C2, Section B; EN: this denotes the system identifying the different aspects of the rules).
“generating an initial dataset comprising multiple different initial data entries having fields…” (Pg.2074, particularly section V; EN: this denotes creating new filter sets based on statistical values and distributions from a parameter file. This is “unlabeled” because the rules are created via statistics and are not made for a specific purpose). “associated with the condition attributes detected in the rules” (Pg.2073, particularly C1, section Iv; EN: this denotes making the parameter file based off of real filter sets). “and containing data values associated with the condition attributes” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters). “the generating comprising populating the fields according to the values detected in the rules and definitions of the condition attributes determined from the data model” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters).
“forming a … dataset using the initial data entries and logic contained in the rules set, wherein the … dataset comprises multiple different … data entries comprising fields and data values of the initial data entries and labels” (Pg.2074-2075, particularly section V; EN: this denotes filling in the various parts of the newly generated filters based on the parameter set generated from the real filter set).
“Wherein populating the fields with data values comprises selecting values randomly” (Pg.2074, particularly C2, last paragraph; EN: this denotes using random variables to create the data).
However, Taylor fails to explicitly disclose, “forming a labeled dataset using the initial data entries … wherein the labeled dataset comprises multiple different labeled entries comprising fields and data values of the initial data entries and labels, wherein the forming comprises selecting an initial data entry from the initial data entries, executing the rules set on the initial data entry to obtain a result, and using the result as a label for the initial data entry, wherein the label is derived from the result attributes of the rule”, “forming a training dataset from the labeled dataset”, and “applying the training dataset to a machine learning model during training of the machine learning model”
Bhatia discloses, “forming a labeled dataset using the initial data entries … wherein the labeled dataset comprises multiple different labeled entries comprising fields and data values of the initial data entries and labels” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc). “wherein the forming comprises selecting an initial data entry from the initial data entries” (Pg.7, particularly paragraph 0060-0062; Figure 3; EN: this denotes parsing in rules from unstructured text, and placing the different pieces in labeled categories as seen in figure 3 such as rule name, tests, enabled, building block, response, etc).
“forming a training dataset from the labeled dataset” (Pg.4, particularly paragraph 0043; EN: this denotes using the rules to train a recurrent neural network).
“applying the training dataset to a machine learning model during training of the machine learning model” (Pg.4, particularly paragraph 0043; EN: this denotes using the rules to train a recurrent neural network).
Rogers discloses, “executing the rules set on the initial data entry to obtain a result, and using the result as a label for the initial data entry, wherein the label is derived from the result attributes of the rule” (Pg.2, particularly paragraph 0019; EN: this denotes using rules to label training data).
Taylor and Bhatia are analogous art because both involve network rule creation.
Before the effective filing date it would have been obvious to one skilled in the art of network rule creation to combine the work of Taylor and Bhatia in order to properly label and import rules in order to use them to train a machine learning algorithm.
The motivation for doing so would be to “use[] natural language processing (NLP techniques … to identify and eliminate duplicate rules, combine similar rules together into ‘super rules’ align STEM rules with frameworks and/or standard rules from standard rules repositories, decompose the rules and their conditions into principal components for use in automatically generating new STEM rules, and train a machine learning model, such as a Recurrent Neural Network (RNN) to generate automated rules based on specific threat intelligence and learning of rule components that correspond to threat characteristics” (Bhatia, Pg.7, paragraph 0059) or in the case of Taylor, allow the system to fully parse and understand the pieces of the generated rules in order to improve the system via machine learning for rule generation.
Therefore before the effective filing date it would have been obvious to one skilled in the art of network rule creation to combine the work of Taylor and Bhatia in order to properly label and import rules in order to use them to train a machine learning algorithm.
Rogers and Taylor modified by Bhatia are analogous art because both involve machine learning.
Before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Rogers and Taylor modified by Bhatia in order to use rules to label training data.
The motivation for doing so would be to use “so-called weak supervision techniques … which can be used to concurrently label multiple training data” (Rogers, pg.2, paragraph 0019) in order to provide “accurately labeled data” which can be “daunting, time consuming, and resource intensive task” (Rogers, Pg.2, paragraph 0019) or in the case of Taylor modified by Bhatia, allow the system to have clearly labeled rules from unlabeled rules to perform the supervised learning of the Taylor/Bhatia combination.
Therefore before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Rogers and Taylor modified by Bhatia in order to use rules to label training data.
As per claim 19, Taylor discloses, “wherein the definitions determined from the data model comprise value domains for the attributes detected in the rules” (Pg.2073, particularly C2, the bullet points; EN: this denote show each of the different types of data are determined and what kind of values they hold)
“determining permissible values for the fields based on the values detected in the rules and the value domains specified in the data model” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each).
“randomly assigning values to the fields of the initial data entries from the permissible values” (Pg.2074, particularly C2, last paragraph; EN: this denotes using random variables to create the data).
As per claim 20, Taylor discloses, “wherein the definitions determined from the data model comprise value domains for the attributed detected in the rules, and wherein generating the initial dataset comprise” (Pg.2073, particularly C2, the bullet points; EN: this denote show each of the different types of data are determined and what kind of values they hold)
“Determining permissible values for the fields based on the values detected in the rules and the value domains from the data models” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each).
“assigning values to the fields of the initial data entries within the permissible values” (Pg.2075, particularly C1, second paragraph; EN: this denotes various information about what values are assigned to each domain, with appropriate ranges (i.e. permissible values) for each). “and from existing data obtained from use of a rule-based system including the rules set” (Pg.2073, particularly C1, section Iv; EN: this denotes making the parameter file based off of real filter sets).
Claim Rejections - 35 USC § 103
Claim 16 is rejected under 35 U.S.C. 103 as being unpatentable over Taylor et al (“ClassBench: A Packet Classification Benchmark”) in view of Bhatia et al (US 20200272741 A1) and further in view of Korjani et al (US 20160179751 A1).
As per claim 16, Taylor modified by Bhatia fails to explicitly disclose, “wherein the operations further comprise validating or testing the machine learning model using at least a portion of the labeled dataset.”
Korjani discloses, “wherein the operations further comprise validating or testing the machine learning model using at least a portion of the labeled dataset” (Pg.3, particularly paragraph 0025; EN: this denotes breaking up the data used to train the system into training data, validation data, and testing data, and using that to help train the system).
Korjani and Taylor modified by Bhatia are analogous art because both involve machine learning.
Before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Korjani and Taylor modified by Bhatia in order to include validation and test data.
The motivation for doing so would be to use the validating data set to “estimate generalization error” (Korjani, Pg.4, paragraph 0032) or in the case of Taylor modified by Bhatia, allow the training to include a validation set to help estimate generalization errors for the machine learning process.
Therefore before the effective filing date it would have been obvious to one skilled in the art of machine learning to combine the work of Korjani and Taylor modified by Bhatia in order to include validation and test data.
Response to Arguments
In pg.9, the Applicant argues in regards to the rejection under U.S.C. 101,
Further, under Step 2A, Prong 2, even if the claims were somehow found to be directed to an abstract idea, the claims integrate any alleged judicial exception into a practical application. For example, the claim is not merely training a machine learning model with old data. Instead, a dataset is generated in a specific way (i.e., populating the fields associated with the condition attributes that permits generation of training data, which can be useful in domains that do not normally generate a sufficient volume of data as described at least in I [0015] of the Application as published. The details of "populating the fields with the data values according to the values detected in the rules" add additional utility by generating training data that triggers the rules. Such a practical application is a technical solution to a technical problem and is directed to eligible subject matter, not merely an abstract idea. Shortcomings of prior technologies are described at [0003] of the Application:
[0003] Training of machine learning models requires large datasets from which the models can learn. Many systems do not generate sufficient data to create datasets for machine learning. The available data may not have enough examples of rare events to create robust learning in the machine learning model.
In response, the Examiner maintains the rejection as shown above. Merely placing data within data fields within a generic computer system does not denote an improvement to machine learning models or the technology of machine learning models. Manipulation of the data being used by a generic model is an improvement to the abstract idea, not an improvement to the machine learning model. Therefore the rejection is maintained as shown above.
In pg.8 Applicant further argues in regards to the rejection under U.S.C. 101,
Finally, looking at the claims as a whole, there are features that are not found in the art, and such features amount to "significantly more" than any alleged abstract idea. For example, the particular arrangement in which training data is generated (e.g., "populating the fields associated with the condition attributes according to the values detected in the rules") is not well-known, understood, conventional. Such arrangements have been found to be directed to eligible subject matter. McRO Inc. V. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1316, 120 USPQ2d 1091, 1103 (Fed. Cir. 2016) (methods of automatic lip synchronization and facial expression animation using computer-implemented rules were not directed to an abstract idea).
In response, the Examiner maintains the rejection as shown above. The McRo case and associated Diehr case dealt with technological processes. For McRo this was lipsyncing of characters in computer animation, in Diehr it was operation of a Resin Press. There is no associated technology here. The claims at no time describe being used for controlling animated characters or physical machines, and therefore these cases do not apply to the instant claims and the rejection is maintained as shown above.
Applicant's remaining arguments with respect to claims 1-20 have been considered but are moot in view of the new ground(s) of rejection.
Conclusion
The examiner requests, in response to this Office action, support be shown for language added to any original claims on amendment and any new claims. That is, indicate support for newly added claim language by specifically pointing to page(s) and line no(s) in the specification and/or drawing figure(s). This will assist the examiner in prosecuting the application.
When responding to this office action, Applicant is advised to clearly point out the patentable novelty which he or she thinks the claims present, in view of the state of the art disclosed by the references cited or the objections made. He or she must also show how the amendments avoid such references or objections See 37 CFR 1.111(c).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BEN M RIFKIN whose telephone number is (571)272-9768. The examiner can normally be reached Monday-Friday 9 am - 5 pm.
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, Alexey Shmatov can be reached at (571) 270-3428. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BEN M RIFKIN/ Primary Examiner, Art Unit 2123