360 likes | 505 Views
Aggregating T&E Data for CCoD Trust in Operational Implementation. Presented to: 27 th International Symposium on Military Operational Research 1 September 2010. Suzanne M. Beers, Ph.D. The MITRE Corporation Operations Research & Systems Analysis Department (E525)
E N D
Aggregating T&E Data for CCoD Trust in Operational Implementation Presented to: 27th International Symposium on Military Operational Research 1 September 2010 Suzanne M. Beers, Ph.D. The MITRE Corporation Operations Research & Systems Analysis Department (E525) Colorado Springs, Colorado
Background • Composable Capability on Demand (CCoD) • MITRE research focus area • Provides means to develop agile information architectures • Rapidly meet warfighter mission needs • Implementation challenge: Trust in composed capability • Fuzzy Logic • Multi-valued logic using fuzzy sets • Successfully used in control and analysis applications • Allows humanistic reasoning, gradual transition between states • Decision support system: Aggregate component T&E data
Overview • What’s Composable Capability on Demand (CCoD)? • Implementation problem statement • CCoD trust taxonomy • What’s fuzzy logic? • Fuzzy logic-based CCoD Decision Support System (DSS) • Structure • Example DSS and results • Conclusion
What’s CCoD? Producers Broker Consumers Loosely Coupled Services Describe & Publish Discover & Subscribe Composable Architectures for Net-centric Operations Rapidly Changeable to Meet Mission Information Needs
One instantiation of CCOD: Mashups • Mashup is a web application that brings together several sources of data to form a unique new combination of information • EMML, Enterprise Mashup Markup Language • Language for creating enterprise mashups • Software applications that consume and mash data from variety of sources • Output presented in graphical user interface, widgets, gadgets
…for example CCoD: Satellite tracking Mashup: Chicago crime
Composed Capability Trust Problem Take the components “off the shelf” Compose a CCoD C2 capability Do I believe this info? Should I act on it? What should I DO??? Tell the user WHAT to give him confidence to use the system To know he’ll make the right moves based on the information
CCoD Implementation Challenge • Lives could depend upon the answer… • How to build trust without slowing down the process? • Use existing component T&E data • Software quality metrics • Aggregate using fuzzy logic • Provide trust measure at composed capability level
Software Trust Metrics • Component Quality Models (CQM) • Evaluate the quality of reusable software modules • Typically start with ISO/IEC 9126 standards • Define characteristics, sub-characteristics, attributes/metrics • CCoD Trust Taxonomy developed by MITRE • Trust: to have confidence in; depend on • Formal acquisition: requirements development then T&E process • CCoD: collect metrics during private to public state transition • Categories • CCoD environment • Component • Composition • Component Developer & Composer: proficiency • Conducted decision-maker survey – ID’ed important factors
CCoD Trust Taxonomy – Top Level Component Quality Characteristics Provide required services under specified conditions Protect info and data; unauthorized cannot read/modify; authorized not denied access Maintain a specified level of performance Usable by someone other than developer Provide appropriate performance relative to amount of resources used Can be modified
Down-sized CCoD Decision Hierarchy Provide correct results w/ needed precision Provide required services Provide appropriate set of functions Interact w/ other components Prevent information disclosure Protect info & data Only modifiable by authorized users Maintain performance Re-establish performance & recover data Understand if suitable / how used Non-creator usable
Why fuzzy logic? • Multi-valued logic based upon fuzzy set theory • Each element in a fuzzy set has membership value • A(u) [0,1] • Linguistic variables related in if – then rule bases • Ideal for representation and analysis of imprecise dependencies • Lowers the cost of products by simplifying programming • Define fuzzy sets, define rule bases relating the sets • Captures human-reasoning • Wide variety of successful control and analysis applications • Control: vacuum cleaners, video cameras, subway systems • Analysis: optimize manufacturing lines, predict insurgent network behavior, aging
Fuzzification Fuzzy Input Inference Engine Fuzzy Output Defuzzification Crisp Output Fuzzy Inference Structure Crisp Input
Fuzzification Turning a Crisp (Numerical) Value Into a Degree of Activation of a Fuzzy Set
SMALL MED Fuzzy Inference -- Rule Base IF Height is MED AND Weight is LIGHT THEN Build is SMALL IF Height is SHORT AND Weight is MED THEN Build is MED max
q yCOA = i=1 yi C (yi) q i=1 C (yi) 2/3 1/3 11 3 4 10 1 2 5 6 7 8 9 yCOA = 1*0 + 2* 1/3 + 3* 2/3 + 4* 2/3 + 5* 2/3 + 6* 2/3 + 7* 2/3 + 8* 1/3 + 9* 1/3 + 10* 1/3 +11*0 yCOA = 5.642 0+ 1/3 + 2/3 + 2/3 + 2/3 + 2/3 + 2/3 + 1/3 + 1/3 + 1/3 +0 Center of Area Defuzzification
Example Fuzzy-Logic CCoD Decision Hierarchy Provide correct results w/ needed precision Provide required services Provide appropriate set of functions Interact w/ other components Prevent information disclosure Protect info & data Only modifiable by authorized users Maintain performance Re-establish performance & recover data Understand if suitable / how used Non-creator usable
Fuzzy Logic DSS – Test Data to Component Trust: Sample Fuzzy Sets
Fuzzy Logic DSS – Test Data to Component Trust: Sample Rule Base (Functionality)
Fuzzy Logic DSS – Test Data to Component Trust: Sample Results
Fuzzy Logic DSS – Test Data to Component Trust: Sample Results
Fuzzy Logic DSS – Component Trust to Composition Trust: Sample Fuzzy Sets
Fuzzy Logic DSS – Component Trust to Composition Trust: Sample Rule Base
Fuzzy Logic DSS – Component Trust to Composition Trust: Sample Results
Conclusion • CCoD provides a means to rapidly develop agile information architectures • Warfighter needs trust in composed capability • Fuzzy logic provides easily-built decision support • Future research/work focus areas • Incorporate real-time decision-maker risk preferences • Refine metrics, definitize data sources, construct automated data collection/cataloging • Develop Web 2.0-like tool to operationalize