Private cross-media measurement using HMAC and bloom filters
A cross-measurement system gathers data from a set of publishers to determine cross-publisher statistics about content provided on multiple online platforms associated with different publishers. The cross-measurement system receives a set of publisher-specific Bloom filters associated with content exposure of a content set by each of the multiple publishers. using the received publisher-specific Bloom filters, the cross-measurement system generates a cross-publisher Bloom filter, for example, by combining the publisher-specific Bloom filters. Then, the cross-measurement system analyzes the cross-publisher Bloom filter to determine cross-publisher statistics reflecting the content exposure of the content set across the set of publishers, such as a cross-publisher reach of the content set.
1. A method comprising:
receiving, at a cross-measurement system, a set of publisher-specific Bloom filters, each publisher-specific Bloom filter comprising a combination of a hash function applied to an item of personally identifying information of each user of a cohort of users that were exposed to a content item by a particular publisher of a plurality of publishers;
generating, based on a selected set of two or more of the publisher-specific Bloom filters, a cross-publisher Bloom filter by combining the selected set of two or more of the publisher-specific Bloom filters, the cross-publisher Bloom filter containing a group of users that were exposed to the content item across any of the publishers associated with the selected set of publisher-specific Bloom filters;
determining a number of users represented by the cross-publisher Bloom filter; and
analyzing the cross-publisher Bloom filter to determine one or more cross-publisher statistics reflecting the content exposure of the content item across the plurality of publishers based on the determined number of users represented by the cross-publisher Bloom filter.
2. The method of claim 1 , further comprising:
sending, to each publisher of the plurality of publishers, a comparison key; and
wherein each of the publisher-specific Bloom filters are based on content exposure data encrypted using the comparison key.
3. The method of claim 1 , further comprising transmitting, from the cross-measurement system to a content provider associated with the content item, the one or more cross-publisher statistics.
4. The method of claim 1 , wherein each publisher-specific Bloom filter comprises a set of bits representing a set of users exposed to the content item by the publisher and the cross-publisher Bloom filter comprises a set of bits representing an combined set of users exposed to the content item by the plurality of publishers.
5. The method of claim 4 , wherein generating the cross-publisher Bloom filter comprises combining the selected set of two or more publisher-specific Bloom filters by performing a bitwise OR operation on the selected set of two or more publisher-specific Bloom filters.
6. The method of claim 4 , wherein generating the cross-publisher Bloom filter comprises combining the selected set of two or more publisher-specific Bloom filters by performing a bitwise AND operation on the selected set of two or more publisher-specific Bloom filters.
7. The method of claim 1 , wherein determining a number of users represented by the cross-publisher Bloom filter comprises estimating the cardinality of the cross-publisher Bloom filter.
8. The method of claim 7 , wherein estimating the cardinality of the cross-publisher Bloom filter comprises estimating the cardinality of the cross-publisher Bloom filter based on the density of the cross-publisher Bloom filter.
9. The method of claim 1 , wherein the cross-publisher statistics comprise a cross-publisher reach representing the total number of unique users exposed the content item by the plurality of publishers.
10. The method of claim 1 , wherein the cross-publisher statistics comprise a cross-publisher frequency representing an average number of times a user exposed to the content item was exposed to the content item across the plurality of publishers.
11. A non-transitory computer-readable storage medium comprising instructions which, when executed by a processor, cause the processor to perform the steps of:
receiving, at a cross-measurement system, a set of publisher-specific Bloom filters, each publisher-specific Bloom filter comprising a combination of a hash function applied to an item of personally identifying information of each user of a cohort of users that were exposed to a content item by a particular publisher of a plurality of publishers;
generating, based on a selected set of two or more of the publisher-specific Bloom filters, a cross-publisher Bloom filter by combining the selected set of two or more of the publisher-specific Bloom filters, the cross-publisher Bloom filter associated with exposure of the content item across the publishers associated with the selected set of publisher-specific Bloom filters;
determining a number of users represented by the cross-publisher Bloom filter; and
analyzing the cross-publisher Bloom filter to determine one or more cross-publisher statistics reflecting the content exposure of the content item across the set of publishers based on the determined number of users represented by the cross-publisher Bloom filter.
12. The non-transitory computer-readable storage medium of claim 11 , wherein the instructions further cause the processor to perform the steps of:
sending, to each publisher of the plurality of publishers, a comparison key; and
wherein each of the publisher-specific Bloom filters are based on content exposure data encrypted using the comparison key.
13. The non-transitory computer-readable storage medium of claim 11 , wherein the instructions further cause the processor to perform the step of transmitting, from the cross-measurement system to a content provider associated with the content item, the one or more cross-publisher statistics.
14. The non-transitory computer-readable storage medium of claim 11 , wherein each publisher-specific Bloom filter comprises a set of bits representing a set of users exposed to the content item by the publisher and the cross-publisher Bloom filter comprises a set of bits representing an combined set of users exposed to the content item by the plurality of publishers.
15. The non-transitory computer-readable storage medium of claim 14 , wherein generating the cross-publisher Bloom filter comprises combining the selected set of two or more publisher-specific Bloom filters by performing a bitwise OR operation on the selected set of two or more publisher-specific Bloom filters.
16. The non-transitory computer-readable storage medium of claim 14 , wherein generating the cross-publisher Bloom filter comprises combining the selected set of two or more publisher-specific Bloom filters by performing a bitwise AND operation on the selected set of two or more publisher-specific Bloom filters.
17. The non-transitory computer-readable storage medium of claim 11 , wherein determining a number of users represented by the cross-publisher Bloom filter comprises estimating the cardinality of the cross-publisher Bloom filter.
18. The non-transitory computer-readable storage medium of claim 17 , wherein estimating the cardinality of the cross-publisher Bloom filter comprises estimating the cardinality of the cross-publisher Bloom filter based on the density of the cross-publisher Bloom filter.
19. The non-transitory computer-readable storage medium of claim 11 , wherein the cross-publisher statistics comprise a cross-publisher reach representing the total number of unique users exposed the content item by the plurality of publishers.
20. The non-transitory computer-readable storage medium of claim 11 , wherein the cross-publisher statistics comprise a cross-publisher frequency representing an average number of times a user exposed to the content item was exposed to the content item across the plurality of publishers.