Ballistic Imaging: Not Ready for Prime Time
Wednesday, April 30, 2003
by David B. Kopel, J.D., & H. Sterling Burnett, Ph.D.
Table of Contents
Over the last decade the BATFE has built the National Integrated Ballistic Information Network (NIBIN), a database of ballistic images from bullets and cases associated with crimes. The NIBIN is accessible to 235 forensic laboratories throughout the United States.12 This system is sometimes useful to criminal investigators. For example, if a firearm is recovered from a suspected criminal's home, the NIBIN can sort through the ballistic images of bullets found at recent crime scenes in the city. If the NIBIN provides some close matches, a firearms examiner can study the crime scene bullet and the recovered gun to determine if the gun fired the bullet found at the crime scene.
"Of a total of 166,672 bullet entries collected by labs, queries to the bullet image produced only 264 'hits."
Matching Images. The test begins with a cartridge case or bullet recovered from a crime scene or test fired from a gun recovered from a suspect. The image of the cartridge case and/or bullet is entered into the computer database; the computer then compares the new images to its existing image database. The database query produces a list of possible "matches," giving them a "match score" and ranking them in order of their similarity to the bullet or cartridge case at issue. According to Frederic A. Tulleners, Director of the California Bureau of Forensic Services Laboratory, because "[a]utomated computer matching systems do not provide conclusive results," the potential candidates must be "manually reviewed" by an expert ballistics examiner.13
National Database Hits. The NIBIN Web site reports "success stories" for its automated system in suggesting preliminary matches. For the last quarter of 2002, NIBIN reported 10 cases in which NIBIN was used to provide evidence against a particular criminal or to alert investigators that a single perpetrator might have committed two or more crimes.14
Statistics about NIBIN's performance thus far were released at the November 6, 2002, meeting of the Southwest Association of Forensic Scientists in Scottsdale, Ariz., based on data supplied by 206 labs using NIBIN.15 According to the statistics released:
- Of a total of 166,672 bullet entries collected by the labs, queries to the bullet image database had produced 264 "hits." [See Figure II.]
- In other words, 0.16 percent of the bullet entries were associated with a hit - a confirmed link between two different bullets or between a bullet and a gun. (No data were released on the number of hits that led to solving a crime, making an arrest or mounting a prosecution.)
- The NIBIN system had 351,194 cartridge case entries and had produced 4,395 cartridge case hits - a rate of 1.25 percent.
"It costs about $12,000 for a cartridge case hit and about $195,000 for a buller hit."
Cost of Database Matching. These data illustrate that cartridge case ballistic identification is much more productive than bullet identification, although neither system has a very high rate of ballistic matches. The data also show that successful matches using the NIBIN system are expensive:
- At a cost of about $250,000 per site for equipment - not including operator training, system maintenance and operator hours - the 206 labs had spent about $51,500,000 for equipment acquisition.
- Thus the equipment costs alone have amounted to about $12,000 for a cartridge case hit and about $195,000 for a bullet hit. [See Figure III.]
The software currently used by NIBIN is the Integrated Ballistic Identification System (IBIS), manufactured by Forensic Technology, Inc., a private company. At present, IBIS is the only relevant technology available.16 However, a report commissioned by the California Attorney General notes that "IBIS has not been designed for operating with large databases such as the ballistic fingerprint database."17
"Investigators with only a dozen bullets can not specify which brand or model of firearm fired them."
Police sometimes suspect that bullets found at two different crime scenes belong to a single criminal. If so, the firearms examiner can compare the bullets directly, without the NIBIN database. This is what was done in the capital-area sniper case: firearms examiners studied bullets from various murder scenes and concluded that the bullets came from the same gun. When the killers bragged in phone calls about a robbery-murder in Alabama, bullets from an unsolved liquor store robbery in that state were microscopically compared to the Maryland-D.C.-Virginia bullets and found to match. The Maryland sniper case also highlights the limitations of ballistic imaging: investigators with only a dozen bullets were unable to specify which brand or model of firearm had fired the bullets. Indeed, the make and model of the firearm used by the snipers was unknown until the suspects were captured with the gun in their possession.
Automated preliminary ballistic imaging is especially useful when a bullet or cartridge case must be searched against a small comparison set - such as the "open case files" of unsolved crimes in a particular city. For this reason, overly large databases can actually be a hindrance, as we detail below. Currently, the NIBIN databases being created by local forensics labs contain only images of bullets or cartridge cases found at crime scenes or bullet/cartridge images created from test firing guns seized from criminals.
The current selectivity creates a high concentration of guns likely to be involved in unsolved crimes. A database that also included guns belonging to law-abiding citizens would be orders of magnitude larger and would produce many more false positives, which firearms examiners would have to spend many hours disproving.
Incomplete Databases. Any imaging database of new guns would be incomplete for whole classes of firearms, including shotguns and revolvers, that might be used in crimes.18 The databases also would be incomplete because they would not include ballistic images for the estimated 200-million-plus guns privately owned in the United States. Costs, time constraints and privacy concerns have made retroactive gathering of ballistic data on existing guns impractical at present. Also, ballistic traces would be of very limited use if the gun matched to a crime had been sold, traded or stolen.