TRAKS
LSDIS Lab at UGA
Ontologies Data Sets
TRAKS Homepage Try TRAKS Templates About TRAKS

Motivation

Iterim Report

Final Report

Team Members

Status

Reference Links



TRAKS: Terrorist Related Assessment using Knowledge Similarity



Chris Halaschek, Boanerges Aleman-Meza, Satya Sanket Sahoo



Department of Computer Science,
University of Georgia Athens,
GA 30602-7404
{ch, boanerg, sahoo}@cs.uga.edu

1 Motivation

Since the terrorist attack on September 11, 2001, homeland security has been a major topic of both research and application. The identification and possible prevention of various contributing factors to terrorist activities, such as money laundering, identity theft, terrorist planning, etc, are in high demand. The terrorist attacks on September 11 have reinforced this area of interest dramatically. These attacks alone caused an estimated “$120 billion of damage” [1] and claimed around 3000 lives [2]. Clearly, something must be done to prevent the types of events from happening in the future.

Currently, automated solutions for detection in this domain primarily include anti-money laundering software which is essentially based on data mining techniques, rule based mechanisms, etc. While this has had benefits, there are several disadvantages. First, for detection the relationships that compose money laundering operations depend on the structure of the database schema being used in a specific financial institution. For the system to be able to find a money laundering operation there must be a structural match between the data within the database and known rules and/or patterns. This lacks the implicit information when the data is semantically annotated. Hence it would be beneficial to find not only structure similarity but semantic similarities as well. Second, if the data is semantically annotated, there are possibilities of further finding interesting connections. For example, given a bank in Afghanistan that has personnel with semantic relationships with terrorists, a system based on semantic similarity will be able to consider this relationship relevant in the domain of potential terrorist activities. In contrast, this type of model would not be able to be automatically discovered in traditional fixed rule based systems. This is because for this to be possible, the database schema of the financial institution would have to model such relationships.


2 Proposal

With the advent of current technology, data is now being semantically annotated. Hence, there exist many sources that describe different characteristics of a given entity in diverse domains. We plan to use (and if needed extract) semantically marked up data along existing data from interested institutions to find potential terrorist activities. Our proposed approach will employ past known money laundering, id theft, and terrorist attack models/templates to discover potential threats in the knowledge base based on novel semantic similarity algorithms that we will develop. An example data source for past real money laundering operations is available at [3]. We aim to showcase these capabilities with a prototype that makes use of data represented in RDF or possibly OWL.


3 References

[1] The Economic Cost of Terrorism
[2] September 11, 2001: A day of terror
[3] Financial Action Task Force on Money Laundering Homepage