# US 12,602,907 B2 · Data sensitivity estimation > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,602,907 B2 - Document ID: US12602907B2 - Status: Granted - Kind: B2 - Application number: 17/728,045 - Filed: 2022-04-25 - Published: 2023-10-26 - Granted: 2026-04-14 - Earliest claimed: 2022-04-25 - Assignees: Microsoft Technology Licensing, LLC - Applicants: Microsoft Technology Licensing, LLC - Inventors: David Trigano; Andrey Karpovsky; Basel Shaheen - Web page: https://patentgrep.com/patent/US12602907B2 ## CPC - G06N 20/00: Machine learning - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30) - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06F 40/295: Named entity recognition - G06F 40/30: Semantic analysis - G06N 20/20: Ensemble learning - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract The disclosed technology is generally directed to data classification. In one example of the technology, training data and a ground truth that indicates sensitive data within the training data is received. Based at least on the training data, natural language processing is used to learn features. The features include a naming feature that is associated with names of data resources in the training data. Based at least on the training data and the ground truth, using supervised learning, a model that is a heuristic model and/or a machine learning model is created. Input data information that is associated with input data is received. The model is used to determine a data resource sensitivity estimator (DRSE) value for each portion of the input data. The determination is based on the combination of features for the input data. Potentially sensitive data within the input data is flagged based on the DRSE values. ## Claims ### Claim 1 An apparatus, comprising: a device including at least one memory having processor-executable code stored therein, and at least one processor that is adapted to execute the processor-executable code, wherein the processor-executable code includes processor-executable instructions that, in response to execution, enable the device to perform actions, including: receiving training data and a ground truth that indicates sensitive data within the training data; based at least on the training data, using natural language processing to learn a plurality of features, wherein the plurality of features includes a naming feature that is associated with names of data resources in the training data; based at least on the training data and the ground truth, using supervised learning, creating a model that is at least one of a heuristic model or a machine learning model; receiving input data information that is associated with input data; using the model to determine a data resource sensitivity estimator value for each portion of a plurality of portions of the input data, wherein the determination is based on a combination of features for the input data; flagging potentially sensitive data within the input data based on the determined data resource sensitivity estimator values; identifying, using the model, resources that share the combination of features; performing actions on the resources improving data security of the resources; limiting access to the input data flagged as potentially sensitive data; periodically using the model to determine potentially sensitive data added to the input data and flagging new input data as potentially sensitive data; and limiting access to the new input data flagged as potentially sensitive data. ### Claim 2 (depends on 1) The apparatus of claim 1, wherein the model is a machine learning model that uses neural networks. ### Claim 3 (depends on 1) The apparatus of claim 1, wherein the model is a heuristic model that is a probabilistic model. ### Claim 4 (depends on 1) The apparatus of claim 1, wherein the model is a heuristic model that uses at least one of decision trees, logistic regression, or support vector machines. ### Claim 5 (depends on 1) The apparatus of claim 1, wherein using the model to determine the data resource sensitivity estimator value for each portion of the plurality of portions of the input data is accomplished without accessing the input data itself. ### Claim 6 (depends on 1) The apparatus of claim 1, wherein the combination of features further includes at least one access policy that is associated with the plurality of portions. ### Claim 7 (depends on 1) The apparatus of claim 1, wherein the combination of features further includes at least one resource owner property that is associated with the plurality of portions. ### Claim 8 (depends on 1) The apparatus of claim 1, wherein the data resource sensitivity estimator value is at least one of a predicted probability that a portion of the input data is sensitive or a confidence score indicating a predicted relative confidence that the portion of the input data is sensitive. ### Claim 9 (depends on 1) The apparatus of claim 1, wherein the model further determines whether each portion of the plurality of portions of the input data is sensitive for a plurality of categories of sensitivity, and wherein flagging the potentially sensitive data includes indicating, for each category of sensitivity the plurality of categories of sensitivity, whether the portion of the input data is predicted to be sensitive according to that category of sensitivity. ### Claim 10 A method, comprising: receiving training data and a ground truth that indicates sensitive data within the training data; based at least on the training data and the ground truth, using supervised learning, creating a model that is at least one of a heuristic model or a machine learning model; receiving input data information that is associated with input data; using the model to determine a data resource sensitivity estimator value for each portion of a plurality of portions of the input data, wherein the determination is based on a combination of features for the input data; flagging potentially sensitive data within the input data based on the determined data resource sensitivity estimator values; identifying, using the model, resources that share the combination of features; performing actions on the resources improving data security of the resources; monitoring traffic to the input data flagged as potentially sensitive data; receiving a request for a data sensitivity estimation for a portion of input data; using the model to determine the data sensitivity estimation for the portion of the input data; and monitoring the traffic to the portion of the input data flagged as potentially sensitive data by the model. ### Claim 11 (depends on 10) The method of claim 10, wherein the model is a machine learning model that uses neural networks. ### Claim 12 (depends on 10) The method of claim 10, wherein the model is a heuristic model that is a probabilistic model. ### Claim 13 (depends on 10) The method of claim 10, wherein the model is a heuristic model that uses at least one of decision trees, logistic regression, or support vector machines. ### Claim 14 (depends on 10) The method of claim 10, wherein the data resource sensitivity estimator value is at least one of a predicted probability that a portion of the input data is sensitive or a confidence score indicating a predicted relative confidence that the portion of the input data is sensitive. ### Claim 15 A processor-readable storage medium, having stored thereon processor-executable code that, upon execution by at least one processor, enables actions, comprising: receiving training data and labels that indicate sensitive data within the training data; based at least on the training data, using natural language processing to learn a plurality of features, wherein the plurality of features includes a naming feature that is associated with names of data resources in the training data; using supervised learning to generate a model, based at least on the training data and the label, such that the model that is at least one of a heuristic model or a machine learning model; receiving input data information that is associated with input data; using the model to calculate a score for each portion of a plurality of portions of the input data, wherein the calculation is based on a combination of features for the input data; flagging, within the input data, data that is predicted to be sensitive based on the calculated scores; identifying, using the model, resources that share the combination of features; performing actions on the resources improving data security of the resources; performing actions improving data security for the data predicted to be sensitive; receiving a request for a data sensitivity estimation for the input data; using the model to determine new potentially sensitive data in the input data since a prior use of the model; and performing the actions improving the data security for the new potentially sensitive data identified in the input data by the model. ### Claim 16 (depends on 15) The processor-readable storage medium of claim 15, wherein the model is a machine learning model that uses neural networks. ### Claim 17 (depends on 15) The processor-readable storage medium of claim 15, wherein the model is a heuristic model that is a probabilistic model. ### Claim 18 (depends on 15) The processor-readable storage medium of claim 15, wherein the model is a heuristic model that uses at least one of decision trees, logistic regression, or support vector machines. ### Claim 19 (depends on 15) The processor-readable storage medium of claim 15, wherein the score is at least one of a predicted probability that a portion of the input data is sensitive or a confidence score indicating a predicted relative confidence that the portion of the input data is sensitive.