|
| Introduction |
 |
 |
This project aims to develop automated deception detection
techniques using data mining approaches. At
current stage, we focus on the problem of criminal
identity deception that is frequently encountered
by police officers during crime investigations.
In the mean time, individual identity is central
to the practice of intelligence communities.
We believe that a solution to that problem will
not only help solving crimes in local law enforcement,
but also benefit a larger domain of the homeland
security.
With
the help of a police detective expert in Tucson
Police Department (TPD) who has served in
law enforcement for 30 years, we conducted
a case study on a sample of deceptive criminal
identity records. We found that identity deception
mostly occurred in the attributes of name,
date of birth, ID numbers, and address. A
taxonomy of criminal identity deception was
built to show different deception characteristics
in different identity attributes. We developed
an automated detection algorithm using string
comparison techniques. Experiments showed
that the proposed algorithm correctly detected
94% of deceptive criminal identities.
|
|
| Screenshot |
|
|
The
following figure shows when an officer enters
a name, date of birth, SSN and address of
a suspect, the Deception Detection algorithm
returns a list of names found in the police
databases that could be the same person. The
officer can have more results returned by
increasing the agreement level using the slider.
|
|
|
|
|
