Data de-identification using semantic equivalence for machine learning
An approach is provided in which the approach detects a set of personal information data corresponding to a set of users in a set of training data. The approach transforms the set of training data into a set of semantically equivalent training data by replacing the set of personal information with a set of semantic equivalent data. The approach then trains a machine learning model using the set of semantically equivalent training data.
1. A computer-implemented method comprising:
detecting a set of personal information data corresponding to a set of users in a set of training data;
transforming the set of training data into a set of semantically equivalent training data by replacing the set of personal information data with a set of semantic equivalent data, wherein the set of semantic equivalent data contains de-identified data with dimension retention;
training a machine learning model using the set of semantically equivalent training data, comprising:
loading metadata mapper objects into an object cache;
loading a user identity and associated access permissions into the metadata mapper objects; and
responsive to a polled thread receiving a response, requesting access to the semantically equivalent training data based on the cached metadata mapper objects; and
responsive to a runtime query from a client device, transforming personal information within the runtime query into semantic equivalencies, wherein the personal information is replaced with a semantic proximate associated with the set of semantic equivalent data and transmitted to the trained machine learning model.
2. The method of claim 1 further comprising:
retrieving a record corresponding to a selected one of the set of users, wherein the record comprises a plurality of attribute values;
determining that a subset of the plurality of attribute values corresponds to a personal information data policy; and
replacing the subset of attribute values with a subset of the set of semantic equivalent data in response to determining that the subset of the plurality of attribute values corresponds to the personal information data policy.
3. The method of claim 2 further comprising:
accessing a policy map comprising a plurality of policies corresponding to a plurality of machine learning models; and transforming the set of training data into a set of semantically equivalent training data in response to determining that the machine learning model corresponds to a semantic access policy in the policy map.
4. The method of claim 3 wherein at least one of the plurality of policies is selected from the group consisting of an allow access policy, a deny access policy, and the semantic access policy.
5. The method of claim 1 further comprising:
inputting the set of training data into a semantic engine; and
generating, by the semantic engine, the set of semantically equivalent training data.
6. The method of claim 1 further comprising:
receiving, by a client demon, a request comprising a set of feature values;
determining a personal information attribute indicator of each of the set of feature values; and
invoking a semantic engine to transform at least one of the set of feature values based on their corresponding personal information attribute indicator.
7. The method of claim 6 further comprising:
receiving a set of semantic proximate values from the semantic engine that corresponds to the at least one feature value;
merging the set of semantic proximate values with the set of feature values; and
sending the merged set of semantic proximate values and the set of feature values to the trained machine learning model.
8. An information handling system comprising:
one or more processors;
a memory coupled to at least one of the processors;
a set of computer program instructions stored in the memory and executed by at least one of the processors in order to perform actions of:
detecting a set of personal information data corresponding to a set of users in a set of training data;
transforming the set of training data into a set of semantically equivalent training data by replacing the set of personal information data with a set of semantic equivalent data, wherein the set of semantic equivalent data contains de-identified data with dimension retention;
training a machine learning model using the set of semantically equivalent training data, comprising:
loading metadata mapper objects into an object cache;
loading a user identity and associated access permissions into the metadata mapper objects; and
responsive to a polled thread receiving a response, requesting access to the semantically equivalent training data based on the cached metadata mapper objects; and
responsive to a runtime query from a client device, transforming personal information within the runtime query into semantic equivalencies, wherein the personal information is replaced with a semantic proximate associated with the set of semantic equivalent data and transmitted to the trained machine learning model.
9. The information handling system of claim 8 wherein the processors perform additional actions comprising:
retrieving a record corresponding to a selected one of the set of users, wherein the record comprises a plurality of attribute values;
determining that a subset of the plurality of attribute values corresponds to a personal information data policy; and
replacing the subset of attribute values with a subset of the set of semantic equivalent data in response to determining that the subset of the plurality of attribute values corresponds to the personal information data policy.
10. The information handling system of claim 9 wherein the processors perform additional actions comprising:
accessing a policy map comprising a plurality of policies corresponding to a plurality of machine learning models; and transforming the set of training data into a set of semantically equivalent training data in response to determining that the machine learning model corresponds to a semantic access policy in the policy map.
11. The information handling system of claim 10 wherein at least one of the plurality of policies is selected from the group consisting of an allow access policy, a deny access policy, and the semantic access policy.
12. The information handling system of claim 8 wherein the processors perform additional actions comprising:
inputting the set of training data into a semantic engine; and
generating, by the semantic engine, the set of semantically equivalent training data.
13. The information handling system of claim 8 wherein the processors perform additional actions comprising:
receiving, by a client demon, a request comprising a set of feature values;
determining a personal information attribute indicator of each of the set of feature values; and
invoking a semantic engine to transform at least one of the set of feature values based on their corresponding personal information attribute indicator.
14. The information handling system of claim 13 wherein the processors perform additional actions comprising:
receiving a set of semantic proximate values from the semantic engine that corresponds to the at least one feature value;
merging the set of semantic proximate values with the set of feature values; and
sending the merged set of semantic proximate values and the set of feature values to the trained machine learning model.
15. A computer program product stored in a computer readable storage medium, comprising computer program code that, when executed by an information handling system, causes the information handling system to perform actions comprising:
detecting a set of personal information data corresponding to a set of users in a set of training data;
transforming the set of training data into a set of semantically equivalent training data by replacing the set of personal information data with a set of semantic equivalent data, wherein the set of semantic equivalent data contains de-identified data with dimension retention;
training a machine learning model using the set of semantically equivalent training data, comprising:
loading metadata mapper objects into an object cache;
loading a user identity and associated access permissions into the metadata mapper objects; and
responsive to a polled thread receiving a response, requesting access to the semantically equivalent training data based on the cached metadata mapper objects; and
responsive to a runtime query from a client device, transforming personal information within the runtime query into semantic equivalencies, wherein the personal information is replaced with a semantic proximate associated with the set of semantic equivalent data and transmitted to the trained machine learning model.
16. The computer program product of claim 15 wherein the information handling system performs further actions comprising:
retrieving a record corresponding to a selected one of the set of users, wherein the record comprises a plurality of attribute values;
determining that a subset of the plurality of attribute values corresponds to a personal information data policy; and
replacing the subset of attribute values with a subset of the set of semantic equivalent data in response to determining that the subset of the plurality of attribute values corresponds to the personal information data policy.
17. The computer program product of claim 16 wherein the information handling system performs further actions comprising:
accessing a policy map comprising a plurality of policies corresponding to a plurality of machine learning models; and
transforming the set of training data into a set of semantically equivalent training data in response to determining that the machine learning model corresponds to a semantic access policy in the policy map.
18. The computer program product of claim 17 wherein at least one of the plurality of policies is selected from the group consisting of an allow access policy, a deny access policy, and the semantic access policy.
19. The computer program product of claim 15 wherein the information handling system performs further actions comprising:
inputting the set of training data into a semantic engine; and
generating, by the semantic engine, the set of semantically equivalent training data.
20. The computer program product of claim 15 wherein the information handling system performs further actions comprising:
receiving, by a client demon, a request comprising a set of feature values;
determining a personal information attribute indicator of each of the set of feature values;
invoking a semantic engine to transform at least one of the set of feature values based on their corresponding personal information attribute indicator;
receiving a set of semantic proximate values from the semantic engine that corresponds to the at least one feature value;
merging the set of semantic proximate values with the set of feature values; and
sending the merged set of semantic proximate values and the set of feature values to the trained machine learning model.