A Secure Biotoken for Border Management Steve Barry, Principal Investigator Problems:
RFID may be used to speed processing at U.S. ports of entry by enabling pre-fetch of information about the traveler. However, security, privacy, a way to assure that the user of the RFID device is the person to whom the device was issued, and price are core concerns for this approach to be accepted. This project will build and test a prototype credit-card size device that meets all security, privacy, and authentication objectives for a border management card at a reasonable cost.
Objectives:
This research will enhance security and privacy for users of RFID tags in portal access applications while also raising the level of assurance that an authorized person is using the card. We will provide an active radio device in card form that changes ID numbers regularly without losing the association between the tag and its authorized user. Further, the card will be secured by the authorized user's fingerprint, thus assuring to a high level of confidence that the user of the card is the person to whom the card was originally issued. Among candidates for use of this technology are border management programs such as US-VISIT, Free and Secure Trade, NEXUS, and the Transportation Worker Identification Credintial.
Activities:
We will acquire fingerprint-enabled active radio cards and a host workstation to implement a secure method to associate a radio tag with a singular user. We will test the device under conditions similar to those encountered in the Border Management applications and verify the ability of the fingerprint subsystem to tell when authorized persons use the card.
Impact:
The results of this project will provide a proof of concept for a secure and highly reliable border management identification card that meets the primary known functional requirements. This development will establish a functional, security, privacy, cost, and identity assurance benchmark for fieldable systems that can be used for applications such as high-traffic verification of identity at entry and exit portals.
Approved for Public Release: 08-0525 Presentation [PDF]
Anonymized Target List Expansion for Name Vetting Chris Thorpe, Principal Investigator Problems:
The current state of the art in name-matching techniques is to approach a problem using generic text searches rather than specialized name matching techniques. Because this has led to many false positives, some systems can't be used with high transaction programs. Also, name normalization for both query and target names is not applied uniformly, which often results in inconsistencies across tool sets.
Objectives:
We will study the effectiveness of using specialized tools for target list name expansion. With the expanded target lists, we will also investigate the possibility of using extremely fast exact name matching algorithms, including exact anonymized name matching. We will also study the impact that name standardization has on name matching effectiveness.
Activities:
We will construct a name expansion capability using a number of commercial tools. We will write custom software to interact with these tools for normalizing their input and capturing output-expanded target lists. We will then analyze the effectiveness of the individual tools as well as the name normalization process.
Impact:
Name-matching systems could potentially be made much more effective through the use of expanded target lists. Errors from ineffective name matching have led to much public outcry. Expanded target lists may significantly reduce the cost, time, and inconvenience of misidentifying persons, and in turn decrease the need for human review. More effective name anonymized lists will also better protect the public's privacy.
Approved for Public Release: 08-0008 Presentation [PDF]
Bio-Threat Aircraft Warning System Grace Hwang, Principal Investigator Problems:
The recent Severe Acute Respiratory Syndrome (SARS) virus outbreak demonstrated how commercial air travel can pose a serious threat by spreading infectious agents. Infection need not be deliberate to raise public health concerns; the avian influenza virus has the potential to cause a pandemic.
Objectives:
We will develop a rapid, reliable, reagentless, miniature biosensor system that can be deployed onboard aircraft to detect and limit the spread of infectious diseases and biological contaminants. The envisioned sensor could be strategically positioned on board (e.g., in seat backs, food carts, or environmental control systems) to maximize contact with airborne pathogens.
Activities:
Types I and II error rates will be established for a set of air transportation bio-defense problems. A computational fluid dynamics model was implemented to optimize sensor quantity and placement. A wind tunnel for bio-aerosol testing was designed, built, and calibrated. Sensors employing surface plasmon polaritons and surface enhanced Raman scattering were developed and tested using model analytes.
Impact:
Rapid, reliable, onboard bio-threat sensing technologies will improve border security, making it possible to respond to threats in real time. This will allow authorities to take appropriate actions to reduce the potential impact of biological threat and pandemic outbreaks caused by naturally occurring pathogens. This effort should be of interest to the FAA, DHS S&T, and CDC
Approved for Public Release: 08-0518 Presentation [PDF]
Closed Loop Link Mining of Textual Data Zohreh Nazeri, Principal Investigator Problems:
Analysts in a variety of domains are faced with the challenging task of identifying and analyzing entity relationships imbedded in large amounts of textual data. Existing link mining techniques cannot be applied to textual data unless desired information (entities and events) is extracted from the text first. Often significant information is lost when information extraction and link mining are performed in isolation from each other. This loss of information in turn impacts the effectiveness of the overall analysis.
Objectives:
This research seeks to improve the effectiveness of the Information Extraction and Link Mining techniques by incorporating them in a unified system. The objective is to develop a system to better exploit disparate sources of structured and textual data, identify and link entities and entity relationships embedded in the data, and present them in the order of their significance to the user.
Activities:
This project will develop a common framework for Information Extraction (IE) and Link Mining (LM). In addition to improving these techniques, a mechanism for providing feedback between the two will be developed. This work is being pursued in collaboration with Arizona State University. A prototype of the system will be developed. Different sources of data in the finance domain will be studied for use in a proof-of-concept demonstration of the system. Through interaction with the sponsors, use cases and criteria for ranking the results will be learned.
Impact:
This system would be of interest to MITRE sponsors, such as the IRS, FinCEN, and DHS, who need to discover the links among entities in disparate sources of data. In the finance domain, it could be used to uncover suspicious patterns of money laundering or other fraudulent activities. In the anti-terrorism domain, the system could help identify social links. Improvements in the Information Extraction and Link Mining techniques would be contributions to the research society.
Approved for Public Release: 08-0439 Presentation [PDF]
Common Ground Agile Information Sharing Kenneth Smith, Principal Investigator Problems:
Data sharing methodologies (e.g., building a data warehouse or corporate web-portal) are often prohibitively expensive in both time and labor. However, many high-value data sharing activities are urgent, or funding-constrained, for example: sharing medical records as patients transfer among hospitals after a hurricane, or case files as criminals cross state and local boundaries. More agile sharing methodologies are needed.
Objectives:
Our objective is to rapidly identify a small, semantically important, core set of attributes from among the data models of sharing partners, thus a) reducing the problem size, b) focusing limited integration resources, and c) building trust and momentum by "getting in the game" sooner. A secondary objective is to rapidly evaluate the sensitivity of each partner's core instance values.
Activities:
We are extending MITRE's Harmony schema matching tool to match very large schemas (typical of sponsor environments) and address N-way matching as a path to determining the core. We are also developing schema clustering techniques to suggest likely groups of common attributes. Finally, we are investigating data sensitivity metrics to assist data releasers once a sharable core is identified.
Impact:
This work will provide MITRE with tools, methodologies, and staff experience to better help sponsors who must share high value-data under pressing time and money constraints. We also anticipate as byproducts: a) better tools for matching, managing, and visualizing the semantic topology of data models; and b) a deeper understanding of sensitivity (e.g., privacy) evaluation in structured data sets.
Approved for Public Release: 08-0522 Presentation [PDF]
Cross Boundary Information Sharing (XBIS) Laboratory Ken Cox , Principal Investigator Problems:
The XBIS Laboratory provides a facility, infrastructure, and the expertise for modeling complex cross-boundary information sharing scenarios and exploring their potential solutions. The laboratory encourages the use of a scenario-based methodology to facilitate deeper analysis and understanding of the enablers of and impediments to effective information sharing. The XBIS Lab is also a collection point for the latest sharing technologies, including information assurance and information management.
Objectives:
The XBIS Laboratory provides a facility, infrastructure, and the expertise for modeling complex cross-boundary information sharing scenarios and exploring their potential solutions. The laboratory encourages the use of a scenario-based methodology to facilitate deeper analysis and understanding of the enablers of and impediments to effective information sharing. The XBIS Lab is also a collection point for the latest sharing technologies, including information assurance and information management.
Activities:
Using the XBIS Laboratory infrastructure and resources, MITRE has developed a Counterterrorism scenario and a Coalition Warfare scenario. These scenarios model real organizational roles and procedures, and demonstrate new ways of using technology to enable cross-boundary sharing. This year, the Lab is being used as a testbed for a Nuclear Infrastructure Protection scenario showcasing enterprise digital rights management (eDRM) technology to control information sharing.
Impact:
The scenario-based approach stimulates and encourages discussion to understand a range of sponsor challenges, allows for extraction and validation of sponsor requirements, and provides a way for MITRE and its sponsors to explore innovative approaches to solving cross-boundary information sharing problems.
Approved for Public Release: 08-0567 Presentation [PDF]
Data Discovery Using Digests Peter Mork, Principal Investigator Problems:
In this project, we seek to facilitate the discovery of structured data resources relevant to sponsor missions. Current search technology indexes primarily content expressed using textual formats. However, hidden beneath this surface content lurks a vast array of structured content stored in relational or hierarchical data resources. Our research strives to allow potential data consumers to discover this deep content.
Objectives:
Traditionally, data consumers connect with data producers via social mechanisms. We intend to augment this strategy with technology: Using our software, a data producer summarizes its data as a succinct digest. A discovery service aggregates these digests into a repository. Using our search tool, data consumers search the repository for relevant datasets, which they obtain directly from their respective owners.
Activities:
We will first determine how best to summarize data resources in a manner that protects the interests of data producers and meets the needs of data consumers. We will then implement a simple discovery service to support simple ranges (for example, imagery from a geographic region). Finally, we will extend the discovery service to allow more complex searches, including query-by-example.
Impact:
Our work dramatically improves our sponsors' ability to find mission-critical data languishing in hidden databases. Data producers are able to advertise the existence of their datasets to gain visibility, without compromising sensitive data points. We will transition our results to the developers of metadata registries (such as the Department of Defense Metadata Registry) to extend the breadth and depth of these resources.
Approved for Public Release: 06-1398 Presentation [PDF]
Identity Matching Lab Keith Miller, Principal Investigator Problems:
Many MITRE sponsors including the Department of Homeland Security (DHS) and the Terrorist Screening Center (TSC) have an interest in being able to accurately maintain and search databases containing identity data. This data may be either biographic or biometric in nature. The MITRE Identity Matching Lab (IML) has specialized in the development of infrastructure, tools, and data that enable the assessment of COTS and GOTS tools for biographic identity matching.
Objectives:
The Identity Matching Lab is committed to providing data, tools, and infrastructure for objective, user-focused evaluation of identity matching technologies, and to developing technical solutions to maximize sponsors' effective use of these technologies.
Activities:
In addition to continued support of sponsor initiatives, the IML team is researching superior methods for combining identity matching engines in order to further improve identity matching effectiveness. Building on the current data model and tools that make up the IML, the team is also currently considering desiderata for the evaluation of entity resolution tools. This involves consideration of the impact of additional identity attributes, both biographic and biometric, as well as the concepts that must be modeled and tools that must be built in order to effectively and objectively evaluate approaches to identity resolution.
Impact:
Through application of its tools and evaluation methodologies, the MITRE IML team has developed an identity matching "cocktail" that combines three COTS identity matching engines, resulting in a tripling in the effectiveness of identity matching performed by a key national security sponsor. This work will be presented at the 2008 IEEE International Conference on Technologies for Homeland Security. Furthermore, the IML team has been the key technical support to the TSC-sponsored Federal Identity Matching Working Group, assisting the group in defining guidelines and developing tools and a data set for evaluation of name matching technologies across the government. Completion of the beta tests of these tools will prepare them for distribution across the federal government. IML research has also been shared with initiatives across the MITRE Corporation to assist sponsors with interests in intelligence, national security, and defense.
Approved for Public Release: 07-0901 Presentation [PDF]
Method for Prioritizing Suspicious Behavior Jianping Zhang, Principal Investigator Problems:
Many government agencies, including the IRS, FinCEN, SEC, CBP, and their private sector partners, such as PCAOB, are faced with the challenge of conducting large volumes of in-depth examinations of suspicious activities with a limited number of examiners. Predictive models generated with the help of data mining techniques can provide the examiners with a rank-ordered list of these suspicious activities. This list can, in turn, facilitate the prioritization of investigative efforts, optimizing the allocation of limited resources by focusing examiners on cases deemed as highly suspicious
Objectives:
The objective of the proposed research is to design and develop novel data mining techniques to address the above challenge. Specifically, we will: Develop novel rule induction algorithms to generate human comprehensible predictive models from data sets with unbalanced class distributions; Develop effective rule scoring techniques; Develop mechanisms for optimizing the performance of rule induction and scoring algorithms on top-ranked activities; Conduct experiments to validate the developed data mining methods.
Activities:
The result of this research will be a fully functional product based on the proposed rule induction and scoring algorithms as well as publishable research papers and filed patents. Specific tasks include: a literature review on mining with imbalanced data and ranking with rules; design and implementation of an innovative rule induction and scoring algorithms; and an evaluation of these algorithms on their efficiency and accuracy, using real world data.
Impact:
The proposed research is another step toward an effort to assist our sponsors (i.e., the financial intelligence community) in tackling their challenging problems and to position MITRE as a leading R&D organization in this community. It expands on previous research on identifying suspicious activities by developing a novel approach to prioritization
Approved for Public Release: 08-0465 Presentation [PDF]
Miniaturized Hybrid Sensor for Multiple Threat Detection Samar Guharay, Principal Investigator Problems:
Detection of threats has become increasingly difficult due to continuous changes of threat characteristics; for example, the use of new explosives and the construction of improvised explosive devices (IEDs). To achieve a highly effective solution, sensors must be widely deployable and capable of combating multiple threats simultaneously. Rugged, miniaturized sensors with good probability of detection and low false alarms are needed.
Objectives:
The overall objective pertains to rapid detection of chemical, biological and explosive/IED "fingerprints," and forensic analysis. The new hybrid sensor uses the best features of two complementary technologies to achieve greater detection sensitivity and higher selectivity than other currently available sensors.
Activities:
The key activities include designing and developing two orthogonal subcomponents and integrating them to build a new hybrid sensor. Systematic modeling studies are under way, and collaborative experiments with academic institutions are being pursued. This study examines the performance of an integrated sensor for a wide range of threats, especially those due to explosives, in the presence of environmental non-threat substances that can appear as interferences. Ruggedness, miniaturization, and lifetime of the system components are critical in this effort.
Impact:
Development of this new sensor will cut across MITRE mission areas and be especially relevant in obtaining significant performance improvement for explosive/IED detection, force protection, and above all, medical applications. Potential impacts include increased support to the Department of Homeland Security, Department of Justice, Special Operations Command, and the Department of Health and Human Services.
Approved for Public Release: 08-0462 Presentation [PDF]
Natural Language Processing for Anonymization John Aberdeen, Principal Investigator Problems:
De-identification can address a number of problems. For example, medical records are largely unavailable to researchers due to privacy constraints. Because of this, medical record processing is labor-intensive and automated techniques have been slow to develop. Removal of personally identifiable information from such records makes them available for research and applications.
Objectives:
Our objective is to create a de-identification system for free text that can be rapidly tailored to multiple domains and record types. The system should be embedded in an interactive interface to allow end-users to refine the performance and inspect the results. We plan to develop metrics for evaluating the effectiveness of de-identification in various real-world settings.
Activities:
We are identifying partners in relevant domains to understand their de-identification requirements. This will allow us to develop a de-identification system for each domain, including interface and workflow management using partner-defined metrics. We will establish a prototype system at a partner site, evaluate the system based on partner-defined metrics, and publish the results.
Impact:
This work will enable MITRE to break through a critical bottleneck in handling records with privacy constraints. It will result in a new set of utility-based metrics to assess emerging technology for privacy protection, enabling MITRE to provide system engineering advice to sponsors. It will build partnerships with key stakeholders in the medical community and surveillance community.
Approved for Public Release: 07-1316 Presentation [PDF]
Pervasive Personal Navigation Thom Bronez, Principal Investigator Problems:
Determining the physical position of dismounted personnel is a critical capability for a wide variety of missions. The Global Positioning System (GPS) has proven indispensable for this, yet GPS reception can be inaccurate, unreliable, or denied in many environments, including natural or urban canyons, heavy foliage, and inside structures. Traditional dead reckoning approaches are limited by rapid error accumulation.
Objectives:
The objective of this project is to develop novel methods and equipment that dismounted personnel can use for three-dimensional position determination and navigation. This system will provide absolute positioning through high-sensitivity GPS when available and also provide high-accuracy relative positioning during GPS outages through novel dead-reckoning techniques. The system will include navigation applications and be suitable for dismounts.
Activities:
We will develop innovative body-worn sensors along with corresponding position estimation algorithms to develop dismount position in three dimensions. We will integrate this dead-reckoning subsystem with high-sensitivity GPS. We will collaborate with an academic human motion laboratory for early virtual sensor design studies and later experimental sensor evaluations. We will develop and field-test a navigation prototype with low-power, wearable form-factor.
Impact:
Successful development of our novel techniques will enable dismounted ground forces and personnel to determine their physical position and navigate with high reliability and accuracy in many modern GPS-impaired environments. Personal Blue Force Tracking and other location-aware applications will improve the situational awareness and effectiveness of dismounted teams. Our field-tested prototypes will aid transition to sponsors.
Approved for Public Release: 06-1407 Presentation [PDF]
Protected Sharing of Controlled Information Rich Pietravalle, Principal Investigator Problems:
In homeland security applications, sensitive but unclassified (SBU) information sharing among federal, state, local, and private entities requires additional technology-assisted controls. As the sharing exchanges carry the information further from the originator, securing the information consistent with the originator's constraints presents increasing challenges. Current technical implementations make it difficult to ensure that policies and regulations concerning SBU information are followed.
Objectives:
The project will implement a prototype approach to secure automated information sharing that supports fine-grained access and usage controls. We will incorporate policies and rules for accurate sharing of controlled, unclassified information, basing them upon operational scenarios from the Department of Homeland Security (DHS) and State and Local Fusion Centers. We will validate the prototype, scenarios, and policies and rules.
Activities:
Research activities include modeling the sponsor environment, information flow, and CONOPS and building an initial scenario based on a subset of that model. We will create a prototype based on COTS software, augmented by needed information sharing functions; test, validate, and demonstrate the prototype using the XBIS (Cross-Boundary Information Sharing) lab; and iterate the process as time and resources allow.
Impact:
The research will help form the requirements for the next iteration of information sharing systems for DHS and other SBU environments. These requirements will assist in focusing sponsor and COTS supplier dialogue for future acquisitions and information system planning, especially for those users with complex cross-domain needs.
Approved for Public Release: 06-1516
Risk Model for Dynamic Aviation Security Alfred Anderegg, Principal Investigator Problems:
The Department of Homeland Security (DHS) has concluded that mitigation measures should be dynamically applied in proportion to the risk of terrorist actions using an algorithm-based risk assessment. Airport security coordinators need a support environment that can account for shifts in threat that follow security adjustments.
Objectives:
A model will be developed to gauge how adjustments in a layered defense shape the future risk based on adversary reaction to the adjustments. It will supplement red team exercises, providing the security coordinator feedback in the form of dynamic threat in response to security refinements of resource allocations and rules of engagement.
Activities:
Using building blocks created in FY07, we will begin with a rapid development effort involving airport security coordinators. A working tool set with one or more airports will be developed, and a definition of measures and algorithms will also be completed.
Impact:
The dynamic threat interaction concept was presented at the Aviation Security Technology Symposium, DHS Science and Technology Showcase, and the American Association of Airport Executives Airport Security Summit in 2007. Completion of a working version of the model will prepare MITRE for a technology transfer of the tools to the aviation community. The research has been shared with similar special initiatives for DHS to broaden thinking on proportional response.
Approved for Public Release: 08-0010 Presentation [PDF]
The 3x2 Challenge: Improve Fingerprint Recognition Speed by Two Orders of Magnitude While Decreasing Cost by Three Orders of Magnitude Steve Barry, Principal Investigator Problems:
Fingerprint matching uses exhaustive comparisons of a sample fingerprint (the probe) against a database of fingerprints (the gallery). As database size increases, computational load and cost increase unacceptably. Current matching systems perform 1.5-2.5 matches/second/acquisition dollar. To meet projected needs, matching must be done at 250,000 matches/second/dollar. This project will meet this challenge by using commodity hardware and innovative software.
Objectives:
We will distribute the fingerprint matching application onto parallel hardware to realize orders of magnitude better cost/match. Hardware includes the Cell BE processor, clustered general-purpose CPUs, vector processing units, and high-speed graphics processors. We will also use advanced algorithms, indexing, and pattern-based classification to reduce the search set size to improve matches per second by two orders of magnitude.
Activities:
Our initial work will establish the feasibility of our goals through simulation based on a theoretical model and through incremental targeted studies to implement and port image processing functions on specialized hardware. We will examine database scalability and classification properties for binning and filtering and for synchronizing single instruction, multiple data functions. Select hardware platforms will be clustered and individually researched.
Impact:
Our program will develop the core elements of a high-volume fingerprint matching system and demonstrate practical ways to improve the price: performance ratio of such systems. Work will incorporate promising research and integrate a demonstration system. Our goal is to demonstrate the value of our approach to four U.S. government departments: DHS, DoD, DoJ, and DoS.
Approved for Public Release: 08-0519
Understanding (Arabic) Nonverbal Behavior Dan Loehr, Principal Investigator Problems:
A ubiquitous communication channel of interest to the national security community is under-exploited: nonverbal behavior. Current successful use (such as the denial of U.S. entry to the alleged "20th hijacker") is based primarily on intuition. Relevant knowledge is largely confined to islands of specialized research focusing on Western culture, with no bridge from the laboratory to sponsor applications.
Objectives:
The project's primary objective is to enable the national security community to recognize, interpret, and exploit information embodied in nonverbal communication (beyond intuition). To achieve this, there are two supporting objectives: (1) provide enabling technology for analyzing nonverbal behavior, and (2) use that enabling technology to understand and exploit nonverbal behavior for specific cultures, starting with the Arabic culture.
Activities:
The approach for enabling technology is fourfold: refining a methodology for nonverbal analysis, investigating tools for such analysis, creating further tools for sharing analyses, and devising a knowledge base for storing and sharing nonverbal analyses. For the culture-specific objective, the approach involves collecting videos showing Arabic speakers in scenarios tailored to sponsor needs, and performing micro-analysis using the enabling infrastructure.
Impact:
This work will increase the safety and effectiveness of U.S. forces by providing a clearly documented understanding of Arabic nonverbal behavior and the first instance of an infrastructure that allows this knowledge to be used in real-world situations. No other organization has the combined understanding of technology, social sciences, and sponsor needs; hence, MITRE is uniquely positioned to move this field forward.
Approved for Public Release: 08-0440, 06-0514 Presentation [PDF]
Last Updated:05/05/2008 | ^TOP | 
