BCC (2012): Federal Panel Identifying Future Government Needs
1 like•1,509 views
The document summarizes a workshop hosted by Mitre at the 2012 Biometric Consortium Conference, which focused on identifying unfunded priorities in federal biometric science and technology. Key areas of discussion included fundamental research, biometric capture improvement, better data extraction, the need for trusted systems, privacy considerations, and testing standards. The findings emphasize the importance of collaboration between federal and non-federal entities to address these priorities, as the government may not fully support them due to budget constraints.
BCC (2012): Federal Panel Identifying Future Government Needs
- 1. 1 Duane Blackburn Kim Shepard Elaine Mattair September 2012 Project No: 0712ECSE-CA McLean, Virginia MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise Federal Panel Identifying Future Government Needs The views, opinions, and findings contained in this report are those of the MITRE Corporation and should not be construed as an official government position, policy, or decision, unless designated by other documentation. Approved for Public Release; Distribution Unlimited. 12-4517 ©2012 The MITRE Corporation. All rights reserved.
- 2. 2 Introduction The federal government held its annual Biometric Consortium Conference 18-20 September 2012. MITRE hosted a workshop during this conference to highlight FFRDC support to the federal biometrics enterprise. One panel in this workshop focused on identifying priorities that the federal government will not be able to address and/or sponsor, and that should be considered for attention by non-federal entities. This paper summarizes the priorities identified during this panel. Panel Members Moderator: Duane Blackburn; MITRE John Boyd; Director, Defense Biometrics and Forensics1 Jim Loudermilk; Senior Level Technologist, FBI S&T Branch1 Chris Miles; Program Manager, DHS S&T Directorate.1 Background The National Science and Technology Council’s Subcommittee on Biometrics and Identity Management leads interagency prioritization and coordination of the federal government’s biometrics science and technology (S&T) activities. In 2011, the Subcommittee published, as an update to their highly influential 2006 version, The National Biometrics Challenge (Challenge), which describes biometrics S&T priorities for the upcoming three to five years. The document serves two functions: (1) as an outreach tool to enable public-private discussion; and (2) as internal government policy on where agencies should prioritize biometrics S&T funding. The goal for this panel was to identify the Challenge document priorities that the federal government will not be able to address and/or sponsor, and therefore should be priorities for other entities. The panel was organized around the six areas identified in the 2011 Challenge document, with additional discussion around similar unfunded priorities from a Department- centric standpoint. Panel Feedback Area 1: Fundamental Underpinnings (Miles) Biological Distinctiveness: There is a need to understand scientific distinctiveness. For example: 1 Co-chair of the NSTC Subcommittee on Biometrics and Identity Management.
- 3. MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise U.S. Government Panel: Identifying Future Government Needs 18 September 2012 3 What makes up a fingerprint? What makes a fingerprint (or other biometric) unique by demographic or other feature? Laboratory environments are not always the same as the real world. There is a need to understand how the research expands into the operational environment, and how stable the biometrics (results) are over time. Understanding of Results: The federal government needs to fully understand the results that come back from a system. It is important to fully comprehend the results regarding matches and no matches, as well as likelihood ratios. It is important to understand how the algorithms work. Supporting Comments/Context This area is understood, but considerable core research has not been accomplished. This area is being hit hard from a budget perspective because it is considered basic research. The budget priority movement focuses on operational programs. The line between forensics and biometrics is not clear. Is an analysis aimed to support near-term operational decision-making, or will it be used as evidence in court? Or both? The answer to this question determines how the evidence will be processed, which needs to be reflected in results reporting. Area 2: Biometric Capture (Loudermilk) Better Data: There is a need for better capture devices to get better data. There are three types of subjects: 1) Uncooperative; 2) Cooperative; and 3) Non- cooperative.2 There are no “great” devices for the capture of all three types. An individual agency’s buying power is not big enough to affect this issue (i.e., there is not enough funding or need to significantly improve it). There are valid reasons why the government should not provide direct funding for these devices; these are commercial items. This problem is best solved by academia/industry/FFRDCs. Better and Less Expensive Devices Algorithm improvement: There is room to make improvements in all algorithms (including fingerprint). 2 See definitions in the NSTC’s Biometrics Glossary: http://biometrics.gov/Documents/Glossary.pdf.
- 4. MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise U.S. Government Panel: Identifying Future Government Needs 18 September 2012 4 Supporting Comments/Context J. Loudermilk used fingerprints as an example to illustrate the problem/need. He encouraged the audience to expand the illustration to other modalities. The Federal Bureau of Investigation (FBI) has achieved 99.6 percent reliability of identifying a person already enrolled in their system. While that sounds impressive statistically, this could still result in missing 720 identifications a day in the current Next Generation Identification (NGI) system because of the sheer volume of matches being performed. Clearly, there is room for improvement. Capture devices need to be improved to provide better data no matter the enrollment circumstance (e.g., when an individual has oil on their hands). Similar problems are found in other modalities. Area 3: Extraction and Representation (Boyd) EXTRACTION Data: There is a need for databases/datasets that can be used for testing. Databases should include both operational datasets collected under realistic conditions and controlled datasets. There is a need for research datasets that can be shared, as well as sequestered data not previously seen. There is a need for datasets with multiple modalities of interest, especially rapid capture and low quality facial images. Approaches for Robust Segmentation and Exploitation of Information Exploitation of information includes approaches from a range of characteristics and environments. Segmentation and exploitation needs to work in real-time from video, while resolving off-angle pose and low-resolution constraints. Invariant Representations of Individuals Across Multiple Sensors and Modalities While Maintaining Uniqueness. This includes taking samples upon encounter, crunching them into an abstracted feature vector, and generating an all-encompassing identity through the fusion of multiple modalities (possibly at the template level). REPRESENTATION Open representation (or templates) of features
- 5. MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise U.S. Government Panel: Identifying Future Government Needs 18 September 2012 5 This should result in a larger, more open market rather than smaller, proprietary ones. Leveraging these features may require redesign in matchers. Subsequently, matchers may need to be re-tuned to maintain and/or improve accuracy. Develop algorithms that support large scaling, as “big data” becomes a very real concern. The biometric records of the Department of Justice, Department of Homeland Security, and the Department of Defense (DoD) combined will soon reach half a billion. Supporting Comments/Context For fingerprints, the DoD is interested in sensors with reduced costs, speedier acquisition, and improved reliability and accuracy. Area 4: Trusted Systems (Loudermilk) Cancellable3 Biometrics Research This is important to the nation as commerce certainly needs it. The federal government will not be at the forefront of funding it because it’s not as important to our national security missions. Creation and adoption of this capability will open biometrics to online commerce, among other things. Supporting Comments/Context This is not Information Assurance; it is public confidence to trust the system for transactions. The biggest problem in this space is liveness detection, which can be solved. (Liveness detection is not common in products today.) There is a lot to be done in this area; however, it is outside the federal government’s purview. 3 Cancelable (or revocable) biometrics is an intentional process where a biometric is repeatably distorted at enrollment and subsequent usage. If this information becomes compromised, the distortion characteristics can be changed. The concept allows biometric-level personalization while minimizing the risk of a system becoming unusable should data become compromised. Any compromised data would also have significantly less negative privacy ramifications due to the distortion.
- 6. MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise U.S. Government Panel: Identifying Future Government Needs 18 September 2012 6 Area 5: Privacy (Miles) Ask the Right Questions; as a community, we must ask the correct questions about what information our technology collects, how it is maintained, and how that might impact people. Recognize the risk of negative impacts from biometrics/forensics (e.g., a “Dad” turning away from a child because a DNA analysis showed that he is not the biological father). Supporting Comments/Context Technology cannot simply be provided in a vacuum. It is important to provide protection around the information that goes with the technology, and to take responsibility for overseeing that information’s use. Cancellable Biometrics: it is important to advance this concept as it will enable enhanced privacy-protection in operational systems. Area 6: Standards and Testing (Boyd) STANDARDS Reduce the Number of Standards: Collapse to use of fewer standards. This will reduce costs associated with proprietary template generation and algorithms. Consider Revocable (Cancellable) Biometrics. This can mitigate risks associated with cybersecurity and spoofing. TESTING Utilize the Biometric Interagency Test and Evaluation Schema (BITES). BITES is intended to facilitate coordination and collaboration for test information and to promote the consistent implementation of testing standards and methodologies. At a high-level, BITES provides a structure that any entity can use to conduct operationally relevant testing. Adoption of this schema then enables the government to trust that entity’s results. This effort was chartered by the NSTC Subcommittee on Biometrics and Identity Management.
- 7. MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise U.S. Government Panel: Identifying Future Government Needs 18 September 2012 7 Supporting Comments/Context Reference the registration of recommended standards compiled by the NSTC at http://www.biometrics.gov/standards.
- 8. MITRE Biometrics – FFRDC Support to the Federal Biometrics Enterprise U.S. Government Panel: Identifying Future Government Needs 18 September 2012 A-1 APPENDIX A: Departmental-Specific Feedback Department of Defense (Boyd) Access Control Example: Office of the Secretary of Defense (OSD) is staffing a memorandum from leadership that would allow installation commanders to collect biometrics from personnel attempting to access DoD installations for initial vetting. Big Data There is increasing concern about performing 1:N and N:N matches, not just 1:1. The storage and processing requirements (e.g., template generation, matching, and linkages to intelligence and/or law enforcement information) are likely to highly stress the current architectures and infrastructures. There are scenarios that have the potential for billions of images/matches. Department of Homeland Security (Miles) Biometric Fusion There is an overwhelming amount of data. Analysts need a “thumbs up/down” result. Complexities and number of comparisons rise much more quickly in 1:N fusion applications than in 1:1 applications. Creative strategies are needed to take advantage of all sources of data. Department of Justice (Loudermilk) Education and Policy to address the Hollywood Effect Shows such as CSI and NCIS are incorrectly affecting public perception - and courts’ – about what is possible biometrically. Data The question to ask when considering biometric technology is, “Where will the data come from?” Positive funding streams do not solve all problems. There must be data for a system to be effective (e.g., there are practically no iris databases in the criminal justice arena).