SLIDE 1
The International Technology Alliance in Network and Information Sciences
ITA Research at Aberdeen Tim Norman Co-Investigators: Nir Oren - - PowerPoint PPT Presentation
ITA Research at Aberdeen Tim Norman Co-Investigators: Nir Oren - - PowerPoint PPT Presentation
The International Technology Alliance in Network and Information Sciences ITA Research at Aberdeen Tim Norman Co-Investigators: Nir Oren & Jeff Z. Pan ITA Project Overview 2006-2016, US DoD/UK MoD, led by IBM, ~$115M White
SLIDE 2
SLIDE 3
§ Trust and reputation are crucial concepts in open,
dynamic systems
§ Problem: How to reliably support trust
assessment of services/information sources?
§ Solution: Exploit feature-behavior correlations to
'bootstrap' trust assessment
Trust Assessment and Decision Making
Burnett et al. “Bootstrapping Trust Assessment Through Stereotypes”, AAMAS-2010; “Trust decision-making in multi-agent systems”, IJCAI 2011; “Stereotypical Trust and Bias in Dynamic Multi-Agent Systems”, ACM-TIST.
Chris Burnett (PhD) (now TRUMP RF)
Effect of stereotyping on existing models
SLIDE 4
§ Surveillance and task-aware decision-support require
knowledge of what individuals/groups are doing
Ø Problem: How to recognise teams and their activity? Ø Solution: Use of linear conditional random fields for incremental team &
activity recognition from position traces
- Masato et al. “Agent-oriented incremental team and activity recognition”, IJCAI 2011
Team, Behaviour and Policy Recognition
Daniele Masato (PhD) (now Qualcomm) § Influencing others behaviour requires knowledge of their
preferences/policies
Ø Problem: How to influence others from sparse evidence? Ø Solution: Learn from dialogical evidence and background knowledge to
inform future strategy
- Emele et al. “Argumentation strategies for plan resourcing”, AAMAS 2011; “Learning
Strategies for Task Delegation in Norm-Governed Environments”, AAMAS Journal.
David Emele (PhD) (now dot.rural RF)
SLIDE 5
§ Need for flexible, policy-aware reasoning mechanisms in SOA § Problem: How can different policy-aware reasoning services be
employed dynamically
§ Solution: Ontological Logic Programming & OWL-POLAR
Reasoning as a Service
O P L
Geeth De Mel (ABDN PhD) IBM internship (now ARL RF) Murat Sensoy (ABDN RF) Visiting Researcher
N2 N3
R1 R4 R5
ITA Sensor Fabric
N4
SAM Portal
Pellet TrOWL Engine N
Sensoy et al. “OWL-POLAR: A framework for semantic policy representation and reasoning”, J. Web Semantics.
SLIDE 6
Exploitation of Distributed, Uncertain and Obfuscated Information
§ Making decisions using large volume of
complex and uncertain data drawn from heterogeneous and distributed sources
§ Sharing information and making decisions
with such information across sources and consumers with varying trusts and policies
§ Constraints and characteristics: Distributed
sources, quality of information, information sharing policies, time to decision, semantic gap, cognitive overload
§ Task 6.1: Balancing the need to share and
need to know via dynamic obfuscation and
- bfuscation-aware fusion
§ Task 6.2: Framework for discovery and
representation of relevant information and reasoning to minimize information overload
COALITION NEEDS PROJECT FOCUS Develop rigorous methods for decision-making and reasoning with uncertain & obfuscated information from diverse distributed sources.
Hard & Soft Information Sources across Coalition
Fusion ¡of ¡Obfuscated ¡ Informa3on ¡ Trust ¡with ¡Obfuscated ¡ Interac3ons ¡ Query ¡& ¡ ¡Informa3on ¡ Retrieval ¡ Presenta3on ¡of ¡ Relevant ¡Informa3on ¡ Reasoning ¡over ¡Uncertain ¡Data ¡
Task ¡6.1 ¡ Task ¡6.2 ¡
SLIDE 7
Key Technical Accomplishments
7
§ Robust combination of information from hard and soft sources in a coalition that
may be subjecting shared information to obfuscations, and where trust between providers and consumers of information may be variable
Ø Estimating trust in the presence of obfuscation Ø Fusing information from obfuscating sources Ø Reasoning about and with obfuscated information Ø Experimental validation of dynamic obfuscation and obfuscation-aware fusion
Task 1: Information Fusion in Coalition Operations under Data Obfuscation and Variable Trust
§ Minimization of information overload and time to decision by efficient querying,
retrieval and presentation of relevant facts from the deluge of data from distributed sources
Ø Framework to query & retrieve relevant facts in support of decision making Ø Framework to represent and aggregate data with uncertainty Ø Techniques to present relevant facts and answers Ø End-to-end demonstration and evaluation of the overall system
Task 2: Querying Over Distributed Coalition Data Network in Support of Decision Making
SLIDE 8
Science of Security: Mathematics, Economics, and Systems
- TSB/RCUK Trust Economics: Modelling behaviour, systems, policy, and
- perations. Just finished, Final Report at
http://www.abdn.ac.uk/~csc335/TE_Final_Report_collated.pdf. About £1.3M.
- HP’s ‘Security Analytics’ service offering: Applies TE to Security
Management.
- TSB/RCUK Cloud Stewardship Economics: understanding agency and
stewardship in the cloud services business context. About £1.4M.
- HP’s Cloud and Security Lab: Applying CSE to major business propositions
- TSB/RCUK Trust Domains: Understanding security, trust, and stewardship
in the whole ecosystem: infrastructure, services, and social layers. About £1.3M.
- FP7 SECONOMICS: mathematics and economics for policy analysis and
design in information and physical security: CNI and transport security. About €3M.
- M. Collinson, B. Monahan, and D. Pym. A Discipline of Mathematical