Private Record Linkage is a technique for using the content of multiple fields
from two private datasets to determine high-probability record connections
between the two datasets.  This technique is tolerant of inexact data matches,
missing data, and various types of "fuzzy" conditions.  Ultimately, it should
produce matches that a human would see but which aren't clear using traditional
matching techniques like an SQL join.


For example, imagine looking for matches in these two simple datasets:

Dataset 1:
   ID        Name           Address          Age     Gender
   1         S Smith        3614 East 23rd   54      M
   2         W Smith        3614 East 23rd   18      M


Dataset 2:
   ID        Name           Address          Age     Gender
   22        William Smith  3614 E 23 St     18      M
   23        Steve Smith    3614 E 23 St     54      M
   24        Sarah Smith    3614 E 23 St     54      F

Humans can easily spot the two pairs of records that match -- IDs # 1 & 23,
and 2 and 22.  But no one piece of information is adequate to establish these
as good matches, and each match uses different bits of information to refine
the pairing. A combination of close names, matching (although different)
addresses, matching ages and matching gender to sort out which should and which
shouldn't be matched.  Performing this is difficult on raw data, but the problem
is even more difficult when the information being used to establish these
matches must remain completely private!

TripleBlind's Private Record Linkage uses a combination of cryptography and
machine learning to identify these matches without leaking the content of either
dataset to the owner of the other dataset.

The example code in this folder shows how simply this complex action can be
achieved using TripleBlind's SDK.
