|
All MITRE Projects (with summaries and presentations where available)
Listing of project titles in alphabetical order
Pages: 1234567891011121314151617
Closing the Semantic Gap
Primary Investigator:Vilain, Marc B.
Problems:
The explosive adoption of language-enabled analytic tools speaks to their abilities to interpolate what people mean from what people say. But much language-enabled analysis has reached a semantic gap; progress on key tasks is stymied because these methods can only approximate what humans mean. This is especially true for the critical unsolved problem of identifying events and their ramifications.
Objectives:
In previous years, this project has pursued three important components of a semantically-enabled approach to event extraction. Lexical hierarchies and grammatical relations serve as the semantic underpinnings of the approach, and a sentence-directed markup scheme provides a "no-sweat" method for designating training instances of events and reporting extracted results. For this continuation year, our goals consist primarily of bringing maturity to the approach. Technically, we intend to strengthen the lexical hierarchies, broaden the scope of the parser, and more tightly integrate the two. In addition, we anticipate performing validation experiments with sponsor and non-sponsor data, so as to grow our body of "best practices" experience with this approach.
Activities:
We are going much farther this year with our efforts to infer lexical hierarchies from machine-readable dictionary glosses. In particular, we are using phrase-oriented tagging models to designate the salient elements of a gloss, and using these elements to infer lexical hierarchies. We are also extracting phrasal constructions from glosses and using these for semantic normalization (e.g., "set fire to" = "cause to burn") and for guidance to the grammatical parser. In addition, we are extending the parser to cover a range of grammatical relations that surface with complex verbs, nominal forms of verbs, and other grammatical constructs. Along with validating these techniques with sponsor data, we are performing limited integration with end-user interfaces (e.g., FAST and other ISR tools).
Impact:
From a language processing standpoint, this work is sharpening the technological repertoire that we have available for event extraction tasks. The improvements in parser and lexicon, and their integration at various points, will provide further degrees of semantic normalization for present and future extraction exercises. From a best practices perspective, the sponsor-data exercises we are engaged in this year are increasing our knowledge base regarding the applications of these techniques to actual sponsor problems. In point of fact, we are presently engaged in a number of transition activities serving other MIP efforts as well as sponsored work. These efforts are already registering significant success, all of which bodes well for future deployment of these techniques.
Public Release No:09-1266
[Presentation]
Exhibit Date(s):May 5
Cognitive Electronic Countermeasures Ad-hoc Team (CECMaTE)
Primary Investigator:Poston, Jeffrey D.
Collaborative Sensing: Tasking, Collection and Mission Management
Primary Investigator:McKillop, Gerard R.
Problems:
This project will explore a mechanism for real time execution-level management of a suite of ISR assets in a manner that directly optimizes meeting commanders-defined operational goals in a dynamic and fluid environment.The initial focus will be on sensors, but longer term could include other areas such as weapon systems and communications. It would do so within the constraints of the tactical situation, capabilities of available assets, and execution guidance (ROEs, etc.). The scope of this effort will focus on integrated ISR support to ground persistent surveillance and Terrorist Network/Link Analysis mission areas.
Objectives:
1. Develop an approach to provide automated support to managing a constellation of tactical sensors in a manner that is explicitly tied to operational goals and measures of effectiveness.
2. Provide a simulation and live-fly demonstration that incorporates these goals to demonstrate the value of dynamic and coordinated sensor tasking.
Activities:
1.Develop a leading-edge, dynamic tasking/re-tasking algorithm for multiple platforms, sensors, and modes using combined linear and integer programs (LP/IP). The LP/IP provides the engine for a Markov Decision Process (MDP).
2. Design a real-time automated UAV mission management demonstration using MDPs and ordinal optimization. The work is in collaboration with the USAF Academy.
Impact:
This work will provide a framework to integrate disparate airbreathing platform sensors. It facilitates synthesis of tactical and operational objectives. Battlespace knowledge improvements are potentially many fold. This work also pushes the research envelope with use of Markov Processes to assign sensors to targets. It has applicability to existing and developmental sensing programsincluding MIST, UAVs, JSTARS, and others.
Public Release No:09-0789
[Presentation]
Combinatorial Auctions for Emission Permit Markets
Primary Investigator:Olson, Matthew E.
Problems:
Global climate change represents an increasingly urgent concern, one the federal government will address by limiting net greenhouse gas emissions. Multiple means have been proposed, but market-based "cap-and-trade" schemes have garnered the widest political support.
Under such a scheme, the initial allocation of permits into the market represents a transfer of wealth that is ripe for political manipulation and economic inefficiency. A well-designed auction is the best means of forcing transparency into the process and reducing inefficiency.
Objectives:
Our goal is to develop an auction for use in a U.S. emission market that has been expanded to include multiple, correlated pollutants. We look to offer improvements to the state-of-the-art and then assess our designs in a laboratory environment.
Activities:
FY09 begins with an experiment to determine the value of bid privacy in a single commodity, single vintage auction. This will be followed by the determination of a combinatorial design appropriate for the assumed greenhouse gas market. The intent is a theoretical understanding of market performance and participant behavior. A future effort will include an experimental assessment of the design.
Impact:
Should auctions be implemented in future emissions market, they will affect the efficacy of the entire market. Our work will provide a design that has been experimentally confirmed to enable market efficiency while satisfying environmental need.
Public Release No:09-1151
[Presentation]
Exhibit Date(s):May 7
Commercial Comms for Tactical Use
Primary Investigator:McKee, Robert A.
Exhibit Date(s):May 6, May 7
Common Ground Agile Information Sharing
Primary Investigator:Smith, Kenneth P.
Problems:
The primary goal of this research effort is to enable "agile data sharing" for structured datasets. For example, consider a group of federal, state, local, commercial, and non-profit agencies suddenly brought together to provide humanitarian relief in a crisis. It would be valuable to be able to quickly exchange, compare, and assess resources and important records each party could provide. For example, local hospitals might want to form a joint view of available beds and medications, and share medical records when patients are moved due to power outages. However, two key technical challenges hinder such rapid interoperation:All these agencies did not expect to share their structured data with each other, and thus their data models are almost certainly heterogeneous.Once other barriers to sharing (such as heterogeneity) have been removed, it is still often very difficult to determine the sensitivity of local data prior to sharing it with new partners.
Objectives:
Goal 1 - Produce "The Common Ground Workshop": Given N partners, each with a data model, quickly determine the intersection of these N data models. While an exhaustive integration of all N sources is valuable, their intersection contains semantic elements each party felt were valuable (and useful join keys), providing a focal point to begin data sharing. The Common Ground Workshop (CGW) is a set of tools and functionalities enabling the rapid production of such a starting-point vocabulary from a set of sharing partners' data models. Specifically, the CGW will extend binary matching of simple schemas (as done by MITRE's Harmony 1.0 tool), to N-way matching of large "industrial scale" (e.g., 1000's of elements each) schemata. CGW will also include an associated schema repository which can identify clusters of similar schemas as input to the N-way matching process. Finally, CGW will include a visualization of a vocabulary derived from an N-way match.
Goal 2- Use CGW to identify sensitive attributes: High-level data sharing policy is typically set by organizational authorities. However, in an urgent sharing scenario, the person who clicks "send" must make a run-time evaluation of data sensitivity with respect to existing policy. Current COTS PET (privacy enhancing technology) tools do little more than apply masking scripts to databases, and have not yet been able to utilize published data sensitivity metrics (e.g., K-anonymity) to evaluate sensitivity and to suggest appropriate remedial actions to decrease sensitivity. However, these metrics cannot be applied by future tools unless we can first identify "quasi-identifiers" attributes. We will use CGW to rapidly identify the quasi-ids in any schema in a repository.
Activities:
In FY09 we plan to:
1) Complete and release "Harmony 2.0", a binary schema matching tool for industrial-scaled (e.g., thousands of attributes) schemata. Harmony 2.0 will be able to operate in "local mode", as a desktop application, or in "enterprise mode" utilizing an integrated schema repository which may reside on a separate server.
2) Design the Common Ground Workshop, an environment for exploring, matching, and producing a comprehensive vocabulary from N (where N can be greater than 100) large schemata in a schema repository. This design will include the workflow steps, algorithms, and interfaces & visualizations.
3) As the first piece, we will perform schema clustering research and prototyping to develop well-tested and understood methodology for a) rapidly clustering a large set of schemata in a repository into semantically-related clusters, and b) visualizing those clusters (which may be defined in a high-dimensional space) on a 2D grid.
4) A related and important research project is the development and testing of N-way matching algorithms. That is, given a cluster of N (perhaps around 5-10) semantically related schemata, rapidly produce a "comprehensive vocabulary" consisting of all (matched & unified) terms, and the subset of the N schemata which share them. As a baseline, an all-pairs binary match will be used, however this scales with the square of N and more efficient approaches will be investigated.
As application scenarios, we will:
1) Work to integrate the biosecurity investment portfoloio through several activities. A) Interviewing each project to determine information produced, information consumed, and a general information flow model. B) Designing a central information model (e.g., a Biosecurity Activity Message Format) to faciliate the production of information by various sensors, field technicians, and sponsor reports (with appropriate provenance information) for subsequent consumption by analysis and visualization tools. C). We will also seek to utilize project-developed tools to help integrate portions of the biosecurity information flow determined in A.
2) Continue and strengthen an FY08 engagement (1.25 FTE) with the DCGS-A project as they seek to use semi-automated tools to rapidly build "comprehensive vocabularies" in order to achieve enterprise information interoperability.
3) Work with 13 laboratories involved in data sharing to support a dynamic brain atlas project at Johns Hopkins University, in a "nip it in the bud" approach using Harmony to control heterogeneity of data models while they are being designed.
4) Continue involvement with other potential deployments of Common Ground-developed tools such as a) JFCOM and COI vocabularies, b) DHS I&A and enterprise information asset awareness, c) clustering of USMTF (message formats) to evaluate their relationships, d) supporting the DoD-VA healthcare records exchange.
Impact:
All of MITRE's sponsors face the challenges of heterogeneous data, and how to interoperate in the face of it. The lightweight techniques being developed by Common Ground (i.e., Harmony and Unity), which enable rapid solutions for sharing structured data in large collaborations, are a growing neccessity as we find more ways & reasons to collaborate in this highly-networked era. Research addressing the uses & forms of "industrial scale" schema matching, as pursued by this project, is acknowledged as the cutting edge of that field.
An ongoing transition example is the DCGS-A project, which is attempting to rapidly assess the commonalities and distinctiveness of large data models. In FY08 we helped reduce a task that would have taken months (matching two schemata having approximately one million possible element matches to consider) to one that took 3 days. Given that success, larger examples and bigger challenges are being attempted in FY09.
Another example impact is the collaboration with Google, IBM, and many academic partners (see the related OpenII project) which is utilizing the schema repository-based architecture designed and implemented by Common Ground and the Flexi-MOIE projects in FY08 as the basis for a novel open source data integration toolkit.
As an integrating force in the Biosecurity portfolio, it is the goal of this research to bring together systems such as the DNA Order Tracking System (DOTS) that can monitor incoming DNA synthesis requests and identify hazardous DNA sequences, and sensors which can detect triggers such as bioterrorism agents with the visualization tools such as Bioscope and Healthmap to effectively present and analyze the comprehensive views possible through agile information sharing.
Public Release No:09-0903
[Presentation]
Exhibit Date(s):May 6, May 7
Compact coaxial dual-band imaging system
Primary Investigator:Euliss, Gary W.
Problems:
The advantages of imaging in multiple spectral bands have been widely acknowledged, and dual-band capability has been or is being incorporated in a number of soldier systems. Typical dual-band designs tend to be undesirably bulky, particularly for soldier-based application, and require complicated post processing.
Objectives:
DARPA has sponsored the development of a lightweight, compact imaging system based on a novel, folded lens design. The objective of this project is to perform the systems engineering ultimately leading to a compact, pixel-registered, dual-band imaging system composed of two folded imagers stacked on a common optical axis and pre-registered at the pixel level.
Activities:
Build and demonstrate a proof-of-concept VNIR/LWIR dual-band imaging system on a common optical axis and demonstrate pixel-registered alignment of collected images. Develop a transition strategy aimed at applications for the dismounted soldier.
Activities can be divided into three primary components. The first component is the fabrication of a hollow-core, broad band, folded annular optic suitable for integration into a thermal imaging system. The second component is the system engineering and integration of the thermal folded optic with the thermal focal plane array and associated electronics. The third component is then the system engineering and integration of the thermal imaging subsystem with a visible imaging inset to achieve dual bands on a common optical axis.
Impact:
Successful completion of the proposed effort will provide a viable, advantageous alternative to dual-band imaging that will have the most immediate and tangible impact on soldier systems where the reduction in size and weight compared to conventional approaches to dual-band imaging are important.Thus, the impact is in reducing the size and weight of a soldier’s equipment. We believe that the common optical axis, unique to the proposed approach, will potentially have an additional impact. It will enable dual-band capability to be integrated into systems and applications that do not currently allow for such capability because it is too difficult to achieve the necessary optical alignment using the conventional approach with offset optical apertures.
Public Release No:09-0862
[Presentation]
Exhibit Date(s):May 6
Computational Imaging for Persistent Pervasive Surveillance
Primary Investigator:Stenner, Michael D.
Problems:
Computational Imaging (CI) is the design of imaging systems by jointly optimizing optical and post-processing degrees of freedom. By contrast, traditional imager design optimizes the optics to achieve clear raw image, and post-processing is used to further improve that result. The additional degrees of freedom afforded by jointly optimizing both stages (and allowing the raw image to be unusable in its raw state) have led to improved depth of field, resolution, size, weight, etc.
The CIPPS program aims to bring CI techniques to bear on the problem of Persistent Pervasive Surveillance of large urban environments from an airborne platform. This challenging imaging task both pushes the limits of current imaging technology and suffers from the problem of communicating vast amounts of data in an efficient manner.
Objectives:
Identify key CI techniques applicable to the PPS problem space.
Consult with related projects and programs to tailor solutions to current and future PPS needs.
Establish collaboration with the University of Arizona to explore novel tracking technologies.
Create device simulation to demonstrate tangible benefit from CI techniques.
Activities:
We intend to explore several CI techniques, including aperture coding, exposure coding, and multiplexed measurement for change detection and tracking. These approaches have been shown to extract task-oriented information with a dramatically reduced number of measurements compared the traditional approach of processing high-resolution video.
After identifying promising technologies and assessing their combined use, we will design and simulate a proof-of-concept PPS system which demonstrates tangible benefit from CI techniques.
Impact:
By reducing the data volume while optically improving the relevance of PPS data, crucial tracking and surveillance information can be made available to the warfighter immediately. By efficiently using bandwidth and power resources, these technologies may allow next-generation PPS systems to be smaller and cheaper while providing faster and better information.
Public Release No:09-0822
[Presentation]
Exhibit Date(s):May 6
Continuity in the Cloud
Primary Investigator:Goldberg, David S.
Problems:
Continuity of operations (COOP) in the face of hostile action or natural disaster is a key mission-assurance requirement for MITRE and its sponsors. With many enterprises vitally dependent on information technology (IT) infrastructures, COOP planners must prepare to keep the business afloat even if its primary IT infrastructures are degraded or destroyed. Cloud computing, offering low-cost application deployment, dispersed and secured datacenters, redundant storage, and metered (pay-as-you-go) billing, promises significant advantages in terms of business continuity. But cloud computing is not free, and its risks and rewards are not easily calculated and compared. For example, an enterprise must accept the risk of outsourcing protection of its data to another organization, and may need to adapt its business practices to ensure smooth transition to cloud-based services. Modified business practices may include adopting cloud-based applications for document processing, workplace collaboration, and database management. This project seeks to assess whether the business continuity rewards of external cloud computing services sufficiently exceed their risks to justify their use.
Objectives:
This project has three primary goals: (1) Determine whether commercial cloud computing services provide sufficient value in terms of business continuity to justify making the necessary changes in risk management decisions and business practices to take full advantage of the cloud; (2) Document the risks, including security concerns and necessary operational changes, which must be understood and mitigated to ensure that deployment in the cloud is successful; (3) Demonstrate at least one mission function running in the cloud space and document the technical requirements for deploying in the cloud.
Activities:
We will choose a set of critical functions to attempt to deploy in commercial space. The IT and data dependencies will be documented. We will then attempt to procure resources in one or more commercial cloud spaces such as Amazon's Elastic Computing Cloud (EC2) and model those dependencies there. We will track metrics such as how long it takes to set things up. We will also simulate usage of those reconstituted functions and measure performance and usability. In addition, we will conduct studies to assess the current state of the cloud service industry, meet with key players and document our findings as to their service level agreements, security practices and usability.
Impact:
This project will build MITRE expertise in cloud computing, and position MITRE to help its customers develop business continuity plans that take appropriate advantage of commercial cloud-service offerings. We envisage that such plans will be less expensive in the long run and more resilient to a wider array of disruptions.
Public Release No:09-0941
[Presentation]
Exhibit Date(s):May 5
Cyber Mission Impact Assessment and Response
Primary Investigator:Musman, Scott A
Problems:
Currently, when a cyber attack affectspart of the IT infrastructure, it is nearly impossible to determine the impact that the attack might have onattaining mission goals. When an incident occurs do we know which mission elements are affected? Can we continue to operate and fulfill the mission? Will the mission fail? Can we reconfigure? Can we salvage part of the mission? If so, which are the right parts to salvage?
Mission Assurance means designing and operating systems in a way that minimizes mission impact in spite of adverse factors. The goal is to be able to maintain critical capability, even if in a degraded capacity, to best accomplish mission objectives. To achieve this, it is necessary to encapsulate knowledge of the mission(s) as a component of the proposed defensive solution. This is difficult because it requires understanding how system components support mission capability, and developing a way to compute the mission impact of adverse events based on system design and operational decisions. Our project is about providing the "mission knowledge" for Mission Assurance.
Objectives:
Demonstrate the application of modeling and reasoning techniques to be able to characterize and assess the mission impact of cyber events
Activities:
Explicitly capture and reason mission and system characteristics and cyber incidents
–Avoid implicit encoding of knowledge/relationships
Map the relationships between mission functions, mission assets, cyber assets, and mission related value metrics
–Our core approach uses influence diagrams
Develop techniques to compute Mission Impact of Cyber Events
–Also use this to assess response actions that would mitigate/minimize mission impact
Evaluate the techniques using before/after and with/without comparisons
–Different modeling choices, a variety of use-cases
Generalize the results across multiple mission systems
–Understand the implications of modeling detail
–Comprehend generality of mission models and modeling elements
Develop a working prototype and supporting tools that can make this widely practical when applied to several sponsor missions
Impact:
Significant Improvements in Cyber SA
Provide the capability to quickly identify the mission tasksaffected by an attack
Help prioritize the relative importance that the attack has on each task
Public Release No:09-0860
[Presentation]
Exhibit Date(s):May 6, May 7
|
