Situational Awareness in a Heavily Sensored LiveVirtual Facility

1
Situational Awareness in a Heavily Sensored
Live/Virtual Facility

• Ann Speed, PhD
• Sandia National Laboratories
• aespeed_at_sandia.gov

2
Overview

• Quick overview of the Cognitive Framework
• Problem and Solution Definition
• FY05 Progress Overview

3
Conceptual Diagram of Sandias Cognitive
Framework Next Generation Intelligent Systems
Grand Challenge
4
Next Generation Intelligent Systems Grand
Challenge Integrated Demonstration

• This demonstration utilizes a notional problem
  with no bearing on any real facility
• Cannot be a sandbox problem
• Must wed cognitive and perceptive systems
  without violating any of approx. 15 constraints
  imposed by the two systems
• 20 groups, 17 potential problems, 1 finalist
• Situational awareness in heavily sensored
  physical facility
• FY04 Demo was in simulation environment
• FY05 is a live experiment to demonstrate
  effectiveness of integrated system

5
Problem Set Solution

• Humans cannot attend to multiple sensor feeds
  simultaneously
• Humans cannot follow paths of multiple attackers
  through a physical space just by watching sensor
  feeds
• A cognitive model of an expert in tactics,
  techniques, and procedures (TTPs) could augment a
  human officers ability to make sense of alarms
  caused by multiple attackers
• That same system can enable the human officers
  ability to determine intent of attackers

6
Integrated Demonstration FY04

• Goal for FY04 was a simulated demonstration in a
  physical security facility to improve human
  ability to make sense of multiple sensor alarms
• Result was a cognitive model that infers intent
  and an interface that displays the inferred the
  intent of 10 adversaries as they move through a
  heavily-sensored facility
• MIDS (seismic)
• Multiple targets
• 1 scenario with scripted adversaries
• Cognitive model scalable to many attack scenarios

7
Cognitive Model and User Interface
8
FY05 Demonstration

• Executed in a simulated/live hybrid notional
  facility
• Hardware sensors, sensor models integrated into
  3D Umbra model of the hybrid facility
• TTP expert Cognitive model that responds to
  hardware and simulated sensors
• Experimental Stimuli (attack scenarios) include
• No alarms
• False alarms
• Single target attacks
• Multiple target attacks
• Single fence breaches
• Multiple fence breaches
• Scenarios with between 5 and 12 attackers

9
3D Terrain Model of Hybrid Facility
10
FY05 Integrated Demonstration Lifespan
3D Hybrid facility terrain model
Develop experimental Interface
Spring FY05 Data Record with Mock Attacking Force
Conduct experiment
11
Performance Metrics FY05 Live Demo/Experiment

• Metrics focus on determination of guard force
  situation awareness and include
• identify intended target or targets
• determine minimum number of adversaries in
  facility
• accurately identify false alarms
• identify whether one target or multiple targets
  are likely