Which exposures count?
Both rows use all 5,000 people. Each breach record pairs one person’s email with one breach. Repeated copies of the same record count once.
| Records counted | People with a match | Breach records | Records per person |
|---|---|---|---|
| Any recorded exposure | 4,142 (82.84%) | 14,951 | 2.99 |
| Excluding spam lists | 3,589 (71.78%) | 9,304 | 1.86 |
Of the 5,000 people, 858 had no record, 553 appeared only on spam lists, and 3,589 appeared in at least one nonspam breach. Among those 3,589 people, there were 2.59 nonspam breaches per person.
Applying the 2019 analysis to the archived data gives 2.10 nonspam records per person. The calculation kept the 858 people with no records but dropped the 553 with only spam-list records. Restoring their zeros lowers the mean to 1.866. Removing 26 duplicate nonspam records brings it to 1.861. The reported share with any exposure, 82.84%, is the same in both versions.
Exposure counts and denominatorsCite: Gaurav Sood and Ken Cor (2019). "Pwned: The Risk of Exposure From Data Breaches." ACM Web Science Conference, 289–293, 2019. https://doi.org/10.1145/3292522.3326046
@article{sood2019pwned,
author = {Gaurav Sood and Ken Cor},
title = {{Pwned: The Risk of Exposure From Data Breaches}},
journal = {ACM Web Science Conference},
volume = {293},
pages = {2019},
year = {2019},
doi = {10.1145/3292522.3326046},
url = {https://doi.org/10.1145/3292522.3326046}
}