Final Report: PAGE: Policy Analytics Generation Engine
Abstract
SOMA (Stochastic Opponent Modeling Agents) are a paradigm within which it is possible to make statements of the form "If condition C is true in the environment in which a group G operates, then group G will take action A with probability in the range [L,U]". SOME has been used to model the behaviors of various terrorist groups and make forecasts about future attacks by those groups. The PAGE project uses SOMA models of group behavior in order to generate policies that achieve a desired goal. A policy is a way of changing the environment in which the group operates (subject to feasibility constraints). We develop such techniques on sequential machines and then develop a parallel framework for it. We also developed policies and methods by which a group of defensive resources (e.g. checkpoints) could be situated in a given geography in order to minimize attacks by an adversary on both static and moving targets (e.g. convoys). In addition, the project looked at the problem of computing between centrality in hypergraphs also called Group Between-Ness Centrality. GBC describes the probability that a node or a group of nodes lie on a randomly chosen shortest path between two randomly selected nodes. We developed algorithms to scalably compute both BC and GBC in huge networks, outperforming (under certain conditions) all previous results. We also proposed the concept of covertness centrality which identifies how a bad guy can try to hide from good guys in a social network without being easily identified by understanding how such bad guys can hid in social networks, we have learned how to identify them.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 12, 2016
- Accession Number
- AD1053286
Entities
People
- V. S. Subrahmanian
Organizations
- University of Maryland