Machine Learning for Defense Applications

TECHNOLOGY AREA(S): Info Systems
OBJECTIVE: This is an AF Special Topic in partnership with MD5, please see the AF Special Topic instructions for further details specifically for requirements related to MD5 programs and services and this topic. The objective of this topic is to develop innovative approaches to leverage advances in machine learning technologies to address specific defense challenges. Challenges can include countering machine learning assisted malware, identifying trusted or untrusted behaviors within a network (communications network, social network, etc), supply chain and infrastructure asset management, pattern recognition to discern dangerous behavior in crowds or other related defense-relevant problems. Along with a specific problem, proposals must also identify sources of data that can be used to train and validate the proposed machine learning capabilities. This topic will reach companies that can complete a feasibility study and prototype validated concepts in accelerated Phase I and II schedules. A Phase I award will be completed over 3 months with a maximum award of $75K and a Phase II may be awarded for a maximum period of 15 months and $750K. Proposals that are selected for award under the MD5 Special Topics will need to have participated in an MD5 program or service, or in another technology acceleration program, prior to the completion of the proposed Phase I SBIR project as noted in the AF Special Topic instructions.
DESCRIPTION: The Department of Defense (DoD) is a large and complex organization that consists of many functions that have similar analogies in the commercial sector. We are interested in exploring machine-learning technological areas and solutions that have proven their value in the commercial sector to see if they have applications for an Air Force problem. It is important that any potential solutions have a high probability of keeping pace with the technological change and thus should be closely tied to commercial technologies and solutions that will help support the development of the solution. Proposals must include a description of sources of data that will be used to develop and validate capabilities and should not rely upon the DoD to provide data. Unrestricted or restricted datasets may be utilized but all data must be unclassified and available upon request for review by the government. It is also desired but not required that any potential solutions have a linkage to existing relevant technologies that were developed with defense, other federal, or commercial funding.
PHASE I: Conduct a feasibility study to determine the effectiveness of commercial or defense machine learning technologies or products to solve a defense need. This feasibility study should directly address:
1. Which problem area(s) are being addressed by the solutions
2. How they will apply to the US Government's needs
3. The breadth of applicability of the solution(s) to the US Government
4. Give examples of which government customers would likely be able to utilize the solution(s)
5. The solution(s) should also be evaluated for cost and feasibility of being integrated with current and future complementary solutions
6. How the solution(s) will be able to address potential future machine learning challenges
7. The potential to keep pace with technological change due to things such as other non-DoD applications and customer bases for the solution(s)
The funds obligated on the resulting Phase I SBIR contracts are to be used for the sole purpose of conducting a thorough feasibility study using interviews, analyses, trade studies, experiments, simulations, and/or component testing.
PHASE II: Develop and demonstrate a prototype system determined to be the most feasible solution during the Phase I feasibility study on machine-learning challenges. This demonstration should focus specifically on:
1. A clear and specific government customer that can immediately utilize the solution
2. How the solution differs from any existing technology or product to solve the DoD need (i.e. leverage of new technology or a description of how existing technology has been modified)
3. How the solution can leverage continued advances in technology
4. How the demonstrated capability can be used by other DoD customers
PHASE III: The contractor will pursue commercialization of the various technologies developed in Phase II for transitioning expanded mission capability to a broad range of potential government and civilian users and alternate mission applications.
REFERENCES: 1: Stanford University 2015 Study Panel, One Hundred Year Study on Artificial Intelligence. Sep 2016. https://ai100.stanford.edu/sites/default/files/ai100report10032016fnl_singles.pdf2: 2. National Science and Technology Council, The National Artificial Intelligence Research and Development Strategic Plan. Oct 2016. https://www.nitrd.gov/PUBS/national_ai_rd_strategic_plan.pdfKEYWORDS: Machine Learning, Artificial Intelligence, Trust
CONTACT(S):
Greg Coleman (MD5)
(301) 502-1609
gcoleman@md5.net