450 likes | 803 Views
The Problem of Identifying the Terrorists (Bad Guys) Quickly. by Dr. Peter Chen & Cyber Security Group Computer Science Dept LSU Baton Rouge, LA 70803. Outline. What is the Profiling Problem? Why is it so difficult? What are we doing about it?. What is the Profiling Problem?.
E N D
The Problem of Identifying the Terrorists (Bad Guys) Quickly by Dr. Peter Chen & Cyber Security Group Computer Science Dept LSU Baton Rouge, LA 70803
Outline • What is the Profiling Problem? • Why is it so difficult? • What are we doing about it?
Bad guy 1 • black hair • beard/moustache
Bad guy 2 • well dressed • beard/moustache
Bad guy 3 • black hair • beard/moustache
Bad guy 4 • black hair • highly strung • skinny
Attributes of bad guys • black hair • beard/moustache • well dressed • skinny
Using the fewest attributes to catch all the bad guys • black hair • beard/moustache
Good guy 1 • beard/moustache
Good guy 2 • beard/moustache • well-dressed
Good guy 3 • well-dressed
Using the fewest attributes to catch all the bad guys … • black hair • beard/moustache
…also catches some good guys (casualties): • black hair • beard/moustache
…also catches some good guys (casualties): • black hair • beard/moustache
Goal: • Find the smallest number of attributes that will catch all the bad guys, but at the same time• Include as few casualties (good guys) as possible.
Goal (more ambitious):• Find the smallest number of attributes that will catch as many, and preferably the more important bad guys, but at the same time• Include as few, and preferably the less important good guys, as possible.
Why can’t we just try all possible combinations of attributes, and then pick the one that does best?
This problem is complex • Computers are fast, but not for certain problems, the complexity of the problems will make the computation time grow exponentially. The Profiling (Bad Guy Risk Assessment) is such a problem
Unfortunately, the Profiling Problem is one such problem:Adding one more bad guy, or one more attribute may make the problem take twice as long to solve. • Also, adding one more attribute value (for example, adding a new value on possible hair color will increase significantly the computation time.
No use coming up with the best solution many days/weeks/months after we are all dead and gone.
OK. How about a quick method to find a solution that close to the best?
We have developed algorithms to solve this problem that do not take so long, yet produce very good results.
Summary • In the Profiling Problem, we want to catch as many bad guys, especially the real bad ones, and do so by disturbing as few good ones, especially the important ones. • The Profiling Problem takes too long to solve if we want perfect solutions. No one knows if we can get perfect solutions fast. • We have develop a quick algorithm to find the solutions in a reasonably short time.