The problem of using a search set of fingerprints to identify an individual as a member of a very large existing repository has long been of interest to law enforcement agencies. Partially automated approaches have been used that require significant search space reduction before automated techniques can be used. The time required to perform such an identification becomes onerous as the repository population size increases. An architectural approach to fully automating this one-on-many identification problem is presented and described. The use of automated search space reduction techniques combined with repository distribution across the architecture results in rapid identification response times by implementing a fully parallelized search. The architecture supports scalability in both workload and repository population. A detailed architectural simulation model was created and results were generated to predict performance. These results are presented and substantiate rapid identification response to one-on-many searches of very large fingerprint feature repositories.

The Identification Tasking and Networking (ITN) Federal Bureau of Investigation system will manage the processing of more than 70,000 submissions per day. The workflow manager controls the routing of each submission through a combination of automated and manual processing steps whose exact sequence is dynamically determined by the results at each step. For most submissions, one or more of the steps involve the visual comparison of fingerprint images. The ITN workflow manager is implemented within a scaleable client/server architecture. The paper describes the key aspects of the ITN workflow manager design which allow the high volume of daily processing to be successfully accomplished.

The Federal Bureau of Investigation (FBI) is currently developing a new system to provide timely criminal and civil identities and criminal history information to the nation's local, state, and federal users. The Integrated Automated Fingerprint Identification System (IAFIS), an upgrade to the existing Identification Division Automated Services (IDAS) System, is scheduled for implementation in 1999 at the new FBI facility in Clarksburg, West Virginia. IAFIS will offer new capabilities for electronic transmittal of fingerprint cards to the FBI, an improved fingerprint matching algorithm, and electronic maintenance of fingerprints and photo images. The Interstate Identification Index (III/FBI) System is one of three segments comprising the umbrella IAFIS System. III/FBI provides repository, maintenance, and dissemination capabilities for the 40 million subject national criminal history database. III/FBI will perform over 1 million name searches each day. Demanding performance, reliability/maintainability/availability, and flexibility/expandability requirements make III/FBI an architectural challenge to the system developers. This paper will discuss these driving requirements and present the technical solutions in terms of leading edge hardware and software.

This paper describes the technical and programmatic plans for moving and checking certain data from the IDentification Automated Services (IDAS) system to the new Interstate Identification Index/Federal Bureau of Investigation (III/FBI) Segment database--one of the three components of the Integrated Automated Fingerprint Identification System (IAFIS) being developed by the Federal Bureau of Investigation, Criminal Justice Information Services Division. Transitioning IDAS to III/FBI includes putting the data into an entirely new target database structure (i.e. from IBM VSAM files to ORACLE7 RDBMS tables). Only four IDAS files were transitioned (CCN, CCR, CCA, and CRS), but their total estimated size is at 500 Gb of data. Transitioning of this Very Large Database is planned as two processes.

The high incidence of recidivism (repeat offenders) in the criminal population makes the use of the IAFIS III/FBI criminal database an important tool in law enforcement. The problems and solutions employed by IAFIS III/FBI criminal subject searches are discussed for the following topics: (1) subject search selectivity and reliability; (2) the difficulty and limitations of identifying subjects whose anonymity may be a prime objective; (3) database size, search workload, and search response time; (4) techniques and advantages of normalizing the variability in an individual's name and identifying features into identifiable and discrete categories; and (5) the use of database demographics to estimate the likelihood of a match between a search subject and database subjects.

This paper will describe the software architecture of the Interstate Identification Index (III/FBI) Segment of the FBI's Integrated Automated Fingerprint Identification System (IAFIS). IAFIS is currently under development, with deployment to begin in 1998. III/FBI will provide the repository of criminal history and photographs for criminal subjects, as well as identification data for military and civilian federal employees. Services provided by III/FBI include maintenance of the criminal and civil data, subject search of the criminal and civil data, and response generation services for IAFIS. III/FBI software will be comprised of both COTS and an estimated 250,000 lines of developed C code. This paper will describe the following: (1) the high-level requirements of the III/FBI software; (2) the decomposition of the III/FBI software into Computer Software Configuration Items (CSCIs); (3) the top-level design of the III/FBI CSCIs; and (4) the relationships among the developed CSCIs and the COTS products that will comprise the III/FBI software.

One of the unique approaches Science Applications International Corporation took to meet performance requirements was to start the modeling effort during the proposal phase of the Interstate Identification Index/Federal Bureau of Investigations (III/FBI) project. The III/FBI Performance Model uses analytical modeling techniques to represent the III/FBI system. Inputs to the model include workloads for each transaction type, record size for each record type, number of records for each file, hardware envelope characteristics, engineering margins and estimates for software instructions, memory, and I/O for each transaction type. The model uses queuing theory to calculate the average transaction queue length. The model calculates a response time and the resources needed for each transaction type. Outputs of the model include the total resources needed for the system, a hardware configuration, and projected inherent and operational availability. The III/FBI Performance Model is used to evaluate what-if scenarios and allows a rapid response to engineering change proposals and technical enhancements.

In business and technical text there is a considerably amount of data, factual knowledge that provides specific details about a topic. Because of the exponentially increasing availability of text, there is therefore vast amounts of data to be gleaned. The problem is that the gleaning tools are insufficient, missing a lot of the relevant data and burying the user in irrelevant returns. The paper describes the EXTRACT tool which is specifically designed to dig data out of all its text crevices, doing so with great accuracy and ease-of-user.

A face detection system that automatically locates faces in gray-level images is described. Also described is a system which matches a given face image with faces in a database. Face detection in an image is performed by template matching using templates derived from a selected set of normalized faces. Instead of using original gray level images, vertical gradient images were calculated and used to make the system more robust against variations in lighting conditions and skin color. Faces of different sizes are detected by processing the image at several scales. Further, a coarse- to-fine strategy is used to speed up the processing, and a combination of whole face and face component templates are used to ensure low false detection rates. The input to the face recognition system is a normalized vertical gradient image of a face, which is compared against a database using a set of pretrained feedforward neural networks with a winner-take-all fuser. The training is performed by using an adaptation of the backpropagation algorithm. This system has been developed and tested using images from the FERET database and a set of images obtained from Rowley, et al and Sung and Poggio.

The combination of virtual reality and digital camera techniques when applied to the field of law enforcement explosive ordnance (EO) has demonstrated a significant savings: in training time, personnel awareness, production costs, and in providing a means of rapidly distributing high quality EO technical material to widely diverse law enforcement organizations over a personal computer network. Highly detailed primary source data was not required, nor in most cases available.

Automatic fingerprint classification, automatic fingerprint identification, and latent or partial fingerprint matching each continue to present significant computational challenges despite the rapid improvements in the speed of digital computers. The new emerging technology of optical processing promises to alleviate this computational roadblock and bring fingerprint classification and identification to the everyday user. Several areas of research utilizing optical processors for fingerprints are currently being conducted. In this paper we will discuss the se of optical correlation for whole and partial fingerprint matching with known fingerprints from a preselected set of candidates. Initial optical correlator simulations and breadboard tests indicate that the hybrid processor provides significant capability increases in speed and throughput over a pure digital system. Results showing good correlation of matches and good discrimination from non-matches are presented. Additionally, correlation of partial prints is demonstrated with a strong degree of discrimination.

One of the biggest issues facing the law enforcement community today is the integration of multiple disparate legacy systems with new systems and Commercial-Off-the-Shelf (COTS) products. It is an inherent characteristic of the law enforcement process that multiple agencies and multiple systems are involved in the processing of a single individual from the time of booking to final incarceration, and eventual release. A standard life cycle could be described as follows: arrest/booking, court arraignment, court trial and/or warrant generation, jail/probation, and release/parole. The processing associated with each of these events is often the responsibility of different agencies using different systems but they all must maintain the same information about the individual and the arrest (including photographs, fingerprints, data about the individual, and data about the arrest). Today's difficult is that in many jurisdictions, the same data must be captured multiple times, as the individual passes through different agencies and phases. This replication of the data entry process introduces inaccuracies and inconsistencies into the databases, as well as, increases the manual labor associated with the processing of an individual through the criminal justice system. The challenge for the future is to capture this data once and disseminate this data to all `interested' parties. Interested parties could be other applications within the same agency, local court systems, local prosection systems, state systems, and Federal systems. The other challenge is to architect a solution where multiple disparate legacy systems, newly developed systems, and COTS systems can be integrated together into an architecture which will provide the ability to easily incorporate new technologies or new systems as they become available. This paper will present the integrated solution developed by SAIC where digital photographs, fingerprints, and identification and booking data are collected during the arrest/booking process and passed to other interested parties as the individual travels through the criminal justice process.

The Regional Information Sharing Systems (RISS) program was formed as a partnership for information exchange between the federal government and state and local law enforcement. The six regional projects provide member law enforcement agencies in all 50 states with a broad range of intelligence and investigative support services. Recently, the existing RISS databases were redesigned to allow for connectivity among projects and the capability of a nationwide search of over 450,000 suspects. This relational database of intelligence information, along with a photographic imaging system, an operational `critical event' database, and GIS mapping are integrated components of RISSNET. The Geographical-Regional Information Sharing System (G-RISS) application is being prototypes by Graphic Data Systems Corporation at one RISS site, the Western States Information Network in Sacramento, California. G-RISS is a tool that will combine information from various law enforcement resources, map criminal activities to detect trends and assist agencies by being proactive to combat these activities.

Many forces are driving states to improve their current Criminal History and On-Line Criminal Justice Information Systems. The predominate factors compelling this movement are (1) the deterioration and cost of supporting older legacy systems, (2) current generation high performance, low cost hardware and system software, and (3) funding programs, such as the National Criminal History Improvement Program, which are targeted specifically at improving these important systems. In early 1996, SAIC established an Internal Research and Development project devoted to Computerized Criminal History Systems (CCH). This project began with an assessment of current hardware, operating system, and relational database technology. Application software design and development approaches were then reviewed with a focus on object-oriented approaches, three tier client server architectures, and tools that enable the `right sizing' of systems. An operational prototype of a State CCH system was established based on the results of these investigations.

There is an existing, low cost, and widely used framework in place for both the public distribution of law enforcement information, and the secure and restricted distribution of sensitive data. That is, of course, the Internet. Already, hundreds of law enforcement agencies around the world, at all levels, are utilizing this most cost effective medium for a variety of tasks. In the public mode, now with 21 - 35 million individuals in the U.S. having access, agencies typically make available contact information, Community Oriented Policing (COPS), employment, crime prevention, DARE, police explorer, and other helpful information. Most often this information is available via WWW page, or a local BBS. Other public access is available to thousands of specialized sites, such as forensics, training, narcotics, firearms, terrorism and hate crimes, K9, police supply, traffic related, crime prevention, most wanted, missing persons, etc. Public newsgroups provide a forum for local, national, and international law enforcement issues. In the private mode, there is a wide variety of restricted mail lists providing for the exchange of information on narrowly defined topics including forensics, firearms, COPS, officer survival, and other related areas. Traditional EMail provides another cost effective method for the exchange of information, either to a specific point, or broadcast to an explicit wide audience. As a secure method of quickly exchanging information in a most cost effective way, encrypted data, typically text, files, or images, can be instantly transmitted between individuals or agencies. Commonly available encryption technology (the most commonly used is PGP, a public key encryption utility), is freely or inexpensively available. An additional Internet benefit available to the law enforcement community, is the availability of software. Currently available is a variety of accident investigation, crime scene, dispatch, maintenance, evidence tracking, and other useful software. This software can typically be downloaded and updated for free, or for a nominal cost. It is imperative that more agencies make use of this valuable resource; while hundreds are, thousands are not.

This paper looks specifically at computers and information systems and their effects on the patrol force. Current mobile computing solutions provided to patrol do very little to enhance their productivity. Empowering police officers in the field with mobile computing software that enables them to access critical secondary data improves effectiveness and increases productivity.

Law enforcement personnel need to capture and retrieve quality multimedia data in `real time' while in the field. This is not done today, for the most part. Most law enforcement officers gather data on handwritten forms and retrieve data via voice communications or fax. This approach is time consuming, costly, prone to errors, and may require months before some data are entered into a usable law enforcement database. With advances in the computing and communications industries, it is now possible to communicate with anyone using a laptop computer or personal digital assistant (PDA), given a phone line, an RF modem, or cellular capability. Many law enforcement officers have access to laptop computers within their vehicles and can stay in touch with their command center and/or retrieve data from local, state, or federal databases. However, this same capability is not available once they leave the vehicle or if the officer is on a beat, motorcycle, or horseback. This paper investigates the issues and reviews the state of the art for integrating a PDA into the gathering and retrieving of multimedia data for law enforcement.

This presentation addresses the information continuum from local through federal systems to provide a perspective for local through regional needs. The collection, organization, management, and exploitation of information as we move along this span provide distinct, but related, criteria for systems at the various levels within an overall framework. The requirements for local, state, and regional systems stand out in contrast. The experiences at the Washington/Baltimore HIDTA in dealing with the definition, development, and implementation of capabilities which have been based on modularity, interoperability, supportability, and affordability are used to demonstrate issues and approaches.