1
Introduction
MOTIVATION
The terrorist bombing of Pan Am Flight 103 over Lockerbie, Scotland, in December 1988, led to an extensive reexamination of procedures in place for airline security.1 A number of government and nongovernmental organizations, including the Federal Aviation Administration, the Office of Technology Assessment, and the National Research Council, assessed procedures in place for airline security, including luggage and passenger screening, and looked at new bomb detection methodologies.2 Following the attacks of September 11, 2001, and the attempted shoe bombing of American Airline Flight 63 in December 2001, the issue of airline security was again examined and additional detection and screening procedures were implemented at airport checkpoints.3
Although these terrorist acts were different in nature and execution, the increased security response following each event was based on a common procedure, airport passenger and baggage screening through a portal. In a portal system, each passenger and all luggage receives screening with detectors that are in close contact, at most 1-2 feet away.
As important as airline security and portal screening are, many terrorist threats fall outside the realm of airport portal security. Table 1.1 lists examples of these types of terrorist bombings that have occurred over the last 20 years. For a number of reasons (practicality, lack of awareness of a threat, insufficient checkpoints), screening systems were not in place or were easily defeated in those instances. All of the events in Table 1.1 are different, and none fit into the airport security scenario described above. In each of these events—as opposed to an airport portal screening—individuals or vehicles appear from an open environment and a se-
TABLE 1.1 Selected Terrorism Attacks Outside the Realm of Portal Screening Security
Location |
Year |
Type of Bombing |
Deaths |
Israel |
2000-present |
Suicide (individual) |
>200 |
Najaf, Iraq |
2003 |
Car |
~100 |
UN Headquarters Baghdad, Iraq |
2003 |
Suicide (truck) |
~20 |
U.S. Military Checkpoint Najaf, Iraq |
2003 |
Suicide (car) |
4 |
Mumbai, India |
2003 |
Car (multiple bombings) |
45 |
Jakarta, Indonesia |
2003 |
Car |
14 |
Bali, Indonesia Nightclubs |
2002 |
Car (multiple bombings) |
202 |
USS Cole |
2000 |
Suicide (boat) |
19 |
U.S. Embassies Kenya and Tanzania |
1998 |
Truck |
223 |
U.S. Military Housing Dhahran, Saudi Arabia |
1996 |
Truck |
19 |
Murrah Federal Office Building Oklahoma City |
1995 |
Truck |
168 |
World Trade Center New York City |
1993 |
Car |
6 |
U.S. Marine Barracks Beirut, Lebanon |
1983 |
Suicide (truck) |
242 |
U.S. Embassy Beirut, Lebanon |
1983 |
Suicide (truck) |
63 |
curity decision about the individual or vehicle must be made at a distance. Sampling and sensing in these situations are made difficult by dynamic backgrounds, standoff considerations, and in some instances, pace.
A key component of identifying and responding to an impending terrorist attack that utilizes chemical explosives is to have methods and systems in place to protect military and civilian personnel without the benefit of portal control. This is a daunting challenge. In addition to the challenges of detection in environments with dynamic backgrounds, the variety of different explosives with an array of different chemical structures further complicates this task.
For these reasons, a single type of detector is unlikely to be applicable to all situations in which detection at a distance may be required. Because the problem of explosives detection encompasses so many different potential environments and situations, an essential component of any effective strategy is consideration of different scenarios. These can be broadly divided into two general categories, suicide bombers and wide-area surveillance. Both of these are addressed in this report.
FOCUS OF THE STUDY
The purview of the committee was to consider detection of chemically based explosives in the two basic scenarios of a suicide bomber (e.g., at a military checkpoint) or wide-area surveillance.4 These scenarios were outlined to the committee by the sponsor at their first committee meeting.5 Nuclear explosives were not considered, nor were explosives designed to deliver biological agents. The committee received briefings from outside experts and developed a knowledge base on existing chemical explosives and existing detection methods, without constraining itself to standoff detection techniques. The committee then examined potential new methods for detecting existing threats as well as methods for detecting potential new threats related to the two basic scenarios. The goal was to develop recommendations for research to further the development of novel standoff explosives detection methods. The committee was not tasked with and did not attempt to recommend new detection techniques.
The statement of task for the committee emphasizes identification of research that could reasonably lead to new approaches to standoff detection with good sensitivity and few false interpretations. The committee
4 |
The committee’s Statement of Task is given in Appendix A. |
5 |
See “Standoff Explosives Detection Study,” Appendix C. |
quickly realized that no single detector for all scenarios and devices is likely to be found. Rather, a systems approach incorporating orthogonal detection methods is necessary. Research is needed not only in the science of sensors but also in the systematic incorporation of multiple, orthogonal sensors leading to a sound decision-making process. Therefore, future research programs should develop the science and systems engineering of detection in parallel.
Although the cost of a detection system is certainly a factor in the deployment decision, it was not considered by the committee. For one thing, the economics of system and product development change substantially with time. For another, deployment of a detection system to very few wide-area surveillances has different cost parameters than providing each soldier in the field with a detection system. Similarly, the impact of detection on civil liberties was outside the scope of the committee’s study.
STANDOFF DETECTION
Standoff detection involves decision making at a distance within a certain time frame. To focus its task, the committee developed the following definition:
Standoff explosive detection involves passive and active methods for sensing the presence of explosive devices when vital assets and those individuals monitoring, operating, and responding to the means of detection are physically separated from the explosive device. The physical separation should put the individuals and vital assets outside the zone of severe damage from a potential detonation of the device.
This definition is necessarily situational. The key words “zone of severe damage” and “vital assets” must be specified. Vital assets are defined by cost and replacement, and depend on scenario and deployment. For example, the sensing elements of the detection system could be within the zone of severe damage, while the human and physical assets related to interpretation and decision making might be outside the zone. Defining the zone of severe damage is more problematic because it depends on the nature of the explosive—its power and collateral debris including shrapnel. For a pedestrian suicide bomber, the zone is taken to be 10 m, while for a vehicle suicide bombing the zone could be 100 m.6 For widearea surveillance—monitoring a large area for the presence of explo-
sives—the zone of severe damage is defined by the location of civilians rather than by the individuals and vital assets associated with detection.
Challenges in Standoff Detection
All explosives detection methods generate alarms in the absence of a true threat (“false positives,” “false alarms”) and fail to generate alarms in the presence of a true threat (“false negatives,” “misses”). The associated frequencies (probabilities) of false negatives and false positives offer a means for comparison of proposed detectors. Although these two probabilities are not entirely comparable (one assumes the presence of a true threat, while the other assumes the absence of a true threat), they are related since adjusting a detection method to limit false positives typically results in an increase in false negatives and vice versa. A graph comparing the false-negative and false-positive probabilities for a variety of settings for a particular detection methodology yields the “receiver operating characteristic (ROC) curve,” providing an approach for comparing the relative performance of proposed detection approaches.
In the absence of a single satisfactory approach, a system composed of multiple detectors may provide better overall performance (lower false-positive and false-negative rates, better ROC curve) than any of its component detectors. The multiple detectors may be redundant (i.e., measuring the same signal to reduce measurement error) or orthogonal—measuring different aspects of the same potential threat (e.g., chemical composition of a concealed package, detection of the presence of shrapnel). An effective detection system will integrate raw data, data summaries, or decisions (threat yes-no) from the individual detectors to arrive at an overall assessment of the threat situation.
In addition to the performance of proposed detection methods (individual detectors or detection systems), certain implementation issues considerably complicate standoff detection. First, successful implementation involves detection of a weak signal in a noisy environment. Second, the noisy background is often dynamic (e.g., changing humidity and ambient light, presence of non-threat-associated compounds triggering false alarms). Single-detector systems with exemplary performance in controlled laboratory settings often exhibit considerably poorer performance when applied in the field.
Several practical constraints also hamper the development of effective standoff detection techniques. Cost and expendability of detectors can limit deployment. Ease of transport, setup, and operation impacts the feasibility of any proposed detection methodology (e.g., portability by foot soldiers may be required in a military setting).
Finally, many attack scenarios provide a limited time for detection
and decision; hence, any feasible standoff detection methodology must collect and interpret data as well as provide recommendations to operators in time to allow an appropriate response to an impending threat.
This report emphasizes the need to integrate basic science (e.g., biology, chemistry, physics) with systems thinking and decision making. Chapter 2 lays out the basic principles for viable standoff detection. A framework for developing detection systems, which identifies a decision process based on disparate data inputs, is described in Chapter 3. Chapter 4 summarizes the chemical characteristics of known explosive devices that could be exploited in a detection scheme, while Chapter 5 describes detection methods currently used or under development. In Chapter 6, the potential of exploiting biological markers associated with bomb makers and bombers is discussed. Examples of ideas for detection that have not been fully exploited to date are presented in Chapter 7. Recommendations for further research in specific areas can be found in the various chapters. The purpose of this report is to identify potentially useful research areas deserving more attention to advance the state of the art, rather than proposing best methods or specific solutions. The appendixes include the committee’s statement of task, a glossary of terms used in the report, and a brief summary of the open session presentations to the committee.