Terrorist attack
Parametric Study of the Possible


President  Bush  has  stated  his  support  for a significant increase in  the
number  of  air  marshals and he has asked for $500 million in  appropriations
for  that  program and other air safety measures.  The number of marshals  the
President  requested  is  classified  to  not  reveal  defense  capability  to
terrorists.   Unfortunately,  keeping the number  of marshals secret will  not
keep terrorists from  concluding correctly  that  the  announced  air  marshal
program will provide little security.

In  the  NewsMax  article  Armed Citizens the Best Defense Against Terrorists, 
Sept. 28, 2001, Professor John Lott is quoted to say that 35,000 officers will 
be required to place one air marshal on  each of  the daily flights.   But the 
amount of money requested by President Bush is a small fraction of that needed 
to fund 35,000 marshals.  

To examine  security  implications of  this limited  funding,  we  analyze the
impact on  security of  two levels within the $500 million budget constraints.
The first  analysis  assumes a  level of 3,000 air marshals or approximately a
hundred fold  increase over  current levels.   At an  average  cost for salary
and benefits of  $50,000  per marshal,  3,000 marshals  will cost $150 million
plus  some  administration  and  overhead  (say  another  $50 million).   That
amount is  approximately  two-thirds  of  the cost of  three Boeing 757's (the
type of  airplane  used  in the  Sept. 11 attack)  and two-fifths of 1% of the
cost of clean-up and replacing the World Trade  Towers  in  New York.   So the
cost looks  reasonable  and one hundred fold increase in marshals looks like a
lot of security doesn't it?
 
Well, maybe not so much security  once  we  consider how terrorists might view
this program.   A terrorist planner  could  perform a parametric study just as
we do here.  He will know that adding attack  groups  increases the likelihood
of  finding  planes  without  marshals.   Our goal  is to see  how many attack
groups might be needed and  how that number  is  driven by  the  number of air
marshals employed. 

We assume  for this  analysis  that a  terrorist planner wants to gain control
of  three  aircraft  just  like  the  Sept. 11  attack  where  three  attempts
succeeded  and  the  fourth  was  defeated  by  unarmed ordinary citizens.  He
might assume that a single marshal on a aircraft  could  defeat  the attack on
that aircraft.  Whether or not true, planning  for 100%  marshal effectiveness
will give the  terrorist  one view of the  attack  results.   Assume  that the
terrorist  planner  wants a 95%  chance that  his groups  will be able to gain
control of three planes  and that  the  only barrier  is that  provided by air
marshals (i.e.  without a  marshal  on the  aircraft  the terrorist group will
gain control).  The planner  needs more  than three attack  groups  to account
for some being  defeated  by  marshals.  Statistics  can tell us how many more
will be needed.

The terrorist  planner  is  looking  for the number  of attack  groups, say N,
that results in at least three groups being unopposed  by marshals.   That is,
the number of marshals  encountered  should not be  more than N-3.   We assume
marshals are randomly  assigned to  the flights  with a  equal probability for
all flights (and the terrorists  select  aircraft  randomly  to attack).   The
probability of a marshal being on any particular flight we denote as P.

If the number of marshals is 3000, the  probability  of a  marshal being  on a
particular aircraft is

P = 3000/35000 = .0857

Let Q denote the probability that no marshal is on a particular flight.

Q = 1 - P = .9143

The probability of exactly M marshals on the N flights attacked is

P(M|N) = {N|M}Q**(N-M)*P**(M)

where "**" denotes exponent  (power),  {N|M} = N!/[M!*(N-M)!]  and !  denotes
factorial.

With N groups  (N>3)  attacking, the  probability that  there will be no more
than N-3 marshals  on the  aircraft  (hence three or  more are  uncovered) is
given by

PS(N) = P(0|N) + P(1|N) + ... + P(N-3|N)
i.e.    (Prob. of zero marshals) plus (Prob. of one marshal) plus ... etc.

PS(N) is the probability that the  terrorists succeed in gaining control of 3
aircraft by  attacking N.   The  following  table shows  the  probability  of
terrorist success with N groups attacking (N = 4, 5 and 6) if there are 3,000 
air marshals:

    N    PS(N)
    4    .961
    5    .994
    6    .999

An  attack  similar  to Sept. 11 (with four attack groups) could be conducted
with better  than a 95% chance of success of gaining control of three or more
aircraft if the only protection is the 3,000 air marshals.  This hundred-fold
increase  in air  marshals  will not  have much chance (alone) of defeating a 
Sept. 11 attack.  More specifically,  the statistics  tell  us that employing 
3,000 air marshals have no deterrent value against a Sept. 11 style attack.


Note that  we  are not  implying  that  individual  air marshals would not be
effective.  To the contrary, we assume they  are 100%  effective in  stopping
an attack when present.  The problem is  that the  number of marshals must be
sufficient to cover the daily   flights.   Without that  number  of  marshals
there are significant holes in security that terrorists can exploit.

If the full $500 million set aside by President Bush were to  be used for air
marshals and none for strengthening the  aircraft cockpit  or other  security
measures  the money  requested  by  President Bush  could provide  7,500  air
marshals.  What  would  that number of marshals do to prevent a Sept. 11 type
attack?

The probabilities P and Q above  change, but  the methodology  does not.   In
this  new case  the probability  of finding  a  marshal on a particular plane
would be

P = 7500/35000 = .214

and the probability no marshal will be on a particular plane is

Q = 1 - P = .786

Calculating terrorist probability of  success for four, five, and six  attack
groups as before yield for 7,500 air marshals the table:

    N    PS(N)
    4    .797
    5    .931
    6    .978

So, with 7,500 air  marshals the terrorist group  would  have to  launch  six
attacks  before they would have  a 95% level of certainty that at least three
would succeed.   Clearly, even without  knowing  the  "classified number"  of
marshals, a  terrorist  planner  can  estimate  how marshals  will degrade an
attack.  The principal value  in  the  7,500-level  vice  3,000-level of  air
marshals is  to force  the  terrorist organization  to increase  by  50%  the
number of attack groups to accomplish  the same level  of success.   The cost
for the 7,500-level of air marshals is 250% of the 3,000-level.

Terrorists might think  an  attack  to be  a  success even if only one of the
aircraft hit its  target.   The probability that  a terrorist  attack on four
aircraft would  find at least  one uncovered by  a marshal at the 7,500-level
is 0.998 and  that two will  be uncovered is 0.967.   To insure that even one
attack cannot  succeed will require nearly comprehensive coverage of aircraft
by marshals.   As many as 30,000  marshals only  have a .54  probability (one
chance in two approximately) of  being on  all four  aircraft for a  Sept. 11
style attack.

It seems clear that  President Bush's air  marshal program  will not have any
serious deterrent  or  damage  prevention  value  against  another  terrorist
attack similar to  the one made on Sept. 11 and that security rests primarily
with the other measures being taken.

Since he has issued an order permitting two air force generals to decide when
to shoot down civilian airliners, we  conclude that President Bush has little 
confidence  in his other  security measures.   While that  shoot-down step is 
likely to mitigate the  actual damage terrorists  might do using the aircraft 
as bombs, it will  not do much for  those finding  themselves  on an aircraft 
attacked.  

We question the  moral justification  of a  government program  which permits
the United  States Armed Forces to kill innocent civilians while denying them
the ability to save their own  lives when  attacked.   If armed opposition to
terrorists is an  important component to  preventing  terrorist  from gaining
control of  an  aircraft,  the  Federal Government  should provide that armed
guard immediately one way or another.    

Philip F. Lee
12921 Two Farm Dr.
Silver Spring, MD 20904
301-622-0446 Home
703-418-8225 Work


Philip F. Lee has a PhD in Mathematics from Georgia Institute of Technology, 1970.  
He has more than 28 years experience in the application  of statistics to  defense 
and industrial applications.

Added on 10/2/01 by Phil Lee.