4 ms·
From the report: Our data consist of daily transactions for 2,807,476 households from January 1, 2012 to December 31, 2013, and include both banking (i.e., ch
by minouye 12y ago
From the report:
Our data consist of daily transactions for 2,807,476 households from January 1, 2012 to
December 31, 2013, and include both banking (i.e., checking, savings, and debit card) and credit
card transactions. We observe the date, amount, and description of each transaction.
I'm a little disturbed that they had access to complete transaction information for roughly 2% of American households. Is it common practice for banks to share this data? Surely date + amount + description is personally identifiable in many situations.
- gph 12y agoNot to defend the banks, but in what situation could date/amount/description become personally identifiable? I guess it sort of depends on how descriptive the description field is, but if it just lists the product(s) then I don't really know how you'd ever be able to verifiably match it with a name.
- cbhl 12y agoAn '08 paper from Arvind Narayanan and Vitaly Shmatikov at the University of Texas at Austin showed that, in a large dataset, the combination of a large number of transactions could be used to de-anonymize individuals. http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf