Anonymizing personal information for use in assessing fraud risk

A system for anonymizing data, the system comprising:

• at least one processor;
• and
• memory storing instructions that, when executed by the at least one processor, cause the system to perform a set of operations, the set of operations comprising:
• receiving a set of data, the set of data comprising a first subset of data corresponding to a unique identifier (UID), a second subset of data corresponding to a timestamp, and a third subset of data corresponding to a personal identifiable information (PII) record, the PII record comprising a plurality of fields, one or more of the plurality of fields comprising values;
• pre-processing the set of data to generate a respective embedding for each value of the plurality of fields, wherein the respective embedding comprises one or more randomized values of one or more missing fields in the plurality of fields as one or more randomized parts of the set of data for anonymization;
• anonymizing the pre-processed set of data to generate an anonymized set of data, the anonymizing comprising generating a respective hash for each of the embeddings, and the anonymized set of data comprises embedding of the one or more randomized values;
• and
• providing, via the at least one processor, an output, the output comprising as training data to train a fraud risk model, one or more indications corresponding to the UID, the timestamp, and the anonymized set of data.

A method of anonymizing data to train a fraud risk model, the method comprising:

• receiving a set of data, the set of data comprising a first subset of data corresponding to a unique identifier (UID), a second subset of data corresponding to a timestamp, and a third subset of data corresponding to a personal identifiable information (PII) record, the PII record comprising a plurality of fields, one or more of the plurality of fields comprising values;
• pre-processing the set of data, based on randomization, wherein the randomization comprises inputting a randomized value into a missing field of the one or more of the plurality of fields as a randomized part of the set of data for anonymization;
• anonymizing the pre-processed set of data to generate an anonymized set of data, wherein the anonymized set of data comprises embedding of the randomized value;
• receiving, as training data for training the fraud risk model, via a database, one or more indications corresponding to the UID, the timestamp, and the anonymized set of data;
• and
• training the fraud risk model, based on the database.