Privacy safe anonymized identity matching

A system comprising:

• one or more processors;
• and
• a non-transitory computer readable storage medium comprising instructions that when executed by the one or processors cause the one or more processors to perform operations comprising:
• on a daily basis, generating salt values and using the salt values to generate anonymized impressions, anonymized conversions and anonymized features, the anonymized impressions comprising anonymized personally identifiable information (PII) from impression records, the anonymized conversions comprising anonymized PII from conversion requests and anonymized features comprising anonymized PII from profiles in a messaging system, the impression records and the conversion requests associated with an external content item displayed by the messaging system;
• generating a salted conversion by anonymizing PI associated with a conversion request received on a conversion date, using a salt value generated on the conversion date;
• providing the salted conversion and the anonymized features as input to a machine learning model (MLM) and generating, by the MLM, an anonymized result indicative of an association between a profile in the messaging system and the conversion request, the anonymized features anonymized using the salt value generated on the conversion date;
• matching the anonymized result with eligible anonymized impressions from the anonymized impressions, the eligible anonymized impressions anonymized using respective salt values generated for the conversion date;
• and
• generating an attribution report based on a result of the matching.

A machine-readable non-transitory storage medium having instruction data executable by a machine to cause the machine to perform operations comprising:

• on a daily basis, generating salt values and using the salt values to generate anonymized impressions, anonymized conversions and anonymized features, the anonymized impressions comprising anonymized personally identifiable information (PII) from impression records, the anonymized conversions comprising anonymized PI from conversion requests and anonymized features comprising anonymized PII from profiles in a messaging system, the impression records and the conversion requests associated with an external content item displayed by the messaging system;
• generating a salted conversion by anonymizing PI associated with a conversion request received on a conversion date, using a salt value generated on the conversion date;
• providing the salted conversion and the anonymized features as input to a machine learning model (MLM) and generating, by the MLM, an anonymized result indicative of an association between a profile in the messaging system and the conversion request, the anonymized features anonymized using the salt value generated on the conversion date;
• matching the anonymized result with eligible anonymized impressions from the anonymized impressions, the eligible anonymized impressions anonymized using respective salt values generated for the conversion date;
• and
• generating an attribution report based on a result of the matching.