Senior Vice President, National Security Services
Redhorse
30
Vincent H. Bridgeman
Research Analyst / Project Manager
United States Department of Defense
Pelayo Fernandez
Transforming Intelligence Analysis with
Knowledge Graphs
Transforming
Intelligence
Analysis with
Knowledge
Graphs
Presented to:
Neo4j Government Conference
Oct 25, 2023
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
• Foundational ideas
̶ Computers can extend human intelligence
̶ Knowledge is networked information
̶ Intelligence analysis is a network problem
• How do graphs transform intelligence analysis?
̶ Knowledge graphs are the right foundation
̶ Mission Graph responds to these imperatives
̶ Pushing the boundaries of graphs
• Case study:
̶ Scientific Knowledge Forecasting with US Air Force
Today’s Agenda
32
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
Extended
Intelligence
Sensor Streams
Core
Computational
Capability
Free association,
logic and ideation
enhance analysis
Validation and
Integration
Human
Observation
All human
insights should
be stored
Human intelligence
focuses on interpretation
& invention
Sensor Data
Sensory Data
CORE
EI
Prediction
Logic
Free Association
Comprehension
Memory
Processing
Analysis
Abstract Thought
Opportunity 1
Opportunity 2
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
data
information
knowledge
wisdom
understanding
relations
understanding
patterns
understanding
principles
connectedness
understanding
Knowledge is networked
information
1. We should
automate this
3. So we can spend
more time on this
2. And we should
accelerate this
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
“There are no single-node targets in intelligence. Even when
we are concerned with a person, organization, object or
location, our target is associated with some network –
usually many networks – that are essential to
understanding the target. These networks have become
more complex over the years, largely because of advances
in communications technologies…….The toughest job for
intelligence analysts in any field….is to deal with networks”
Target-Centric Network Modeling: Case Studies in Analyzing Complex Intelligence Issues
Robert M. Clark (Author), William L. Mitchell (Author)
CQ Press; 1st edition (April 24, 2015)
Intelligence analysis is a
network problem
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
Examples of network problems in intelligence
36
Cyber Security Transnational
Crime
Counter Terrorism Counterintelligence
Military Operations Counter
Disinformation
Technology Watch
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
Manual
Visual
Digital
Canvas
Digital
Networks
We are in the third phase of network analysis tools
37
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
Knowledge graphs are the right foundation
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
What does a knowledge graph do?
39
Digital Asset
Value-added data structure –
connected data
Acts as a “knowledge hub” in a
software application/system
Data Science Enabler
Much of data science is graph-based
math
Network analysis can be applied to
connections in the data
User Interface Enabler
Your mind already works like a graph
– you comprehend the world as
things and their relationships
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
40
Redhorse brings knowledge graphs
to the intelligence mission
Data Lake
Entity Resolution
Data Engr.
Graph Workflows
Graph UI
Graph Database
Other UI
Public
Data
Commercial
Data
Proprietary
Data
Domo, Tableau, D3,
ArcGIS, QGis etc.
Bloom
Senzing
Hume
AWS S3,
Azure, Delta
Neo4j
Data Science
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
Pushing the boundaries of
graphs + AI for intelligence
41
This document is proprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation.
42
Case Study:
Scientific Forecasting with
Temporal Knowledge Graphs
for the
United States Air Force
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
UNCLASSIFIED
Science Knowledge Forecasting
Graph Summit for Government
Oct 2023
Distribution A: Approved for public release. 18 OCT 2023.
Requests for this document must be referred to
SAF/CDM/CDMR, ATTN: ISSO
44
UNCLASSIFIED
UNCLASSIFIED
March 2021
Repeated calls to
discover and
understand
emerging & disruptive
science and technology
Expedite generating insight:
“speed to insight à speed to decision”
NSPM-33
NDAAs FY18-FY24
October 2020
February 2022
“… (B) develop machine learning tools
to identify possible future technologies;
…”
NDAA FY24
45
UNCLASSIFIED
UNCLASSIFIED
How do we enable forecasting of
emerging or disruptive technologies
and their impact?
What and where are technologies emerging globally?
How do they apply to government programs?
What government programs will they impact?
How does the USA compare to others?
Are we influencing their research or vice versa?
46
UNCLASSIFIED
UNCLASSIFIED
What if…
47
UNCLASSIFIED
UNCLASSIFIED
PURPOSE:
“Mining the near future for action” -
forecasting science and technology direction
and determining impact to government
ENVISIONED OUTCOME:
Pre-emptive actions by leadership
CONCEPT:
Innovative application of machine learning
algorithms in combination with temporal co-
occurrence knowledge graphs; integration of
massive complex databases to generate
actionable forecasts and determine implications
48
UNCLASSIFIED
UNCLASSIFIED
Knowledge
Graph
Web of Science
49
UNCLASSIFIED
UNCLASSIFIED
66M
articles
817k
articles
Web of
Science
Initial data set
Started with subsets of Web of Science …
with greater
resources, more
data collections
can be added,
e.g., OpenAlex,
ORCID, patents,
etc.
50
UNCLASSIFIED
UNCLASSIFIED
Examples
Putting
Science Knowledge Forecasting
Into action !
51
UNCLASSIFIED
UNCLASSIFIED
What is Synthetic Biology?
alter genetic code
read and manipulate genetic code
combine genetic codes
build genetic codes from scratch
where biosciences, information sciences
and engineering converge
Control these
processes and you
can control or build
your own life…
52
UNCLASSIFIED
UNCLASSIFIED
Top 50 Projected
Gain Increase
nucleotide resolution ???
Details
Role behavior over time
Other projected
gaining terms
Top 50 Projected
Term clusters, metrics, terms, US
vs foreign pub rates
Example: Synthetic Biology
53
UNCLASSIFIED
UNCLASSIFIED
Exploring
nucleotide resolution
Curious co-occurring clusters
US vs. PRC focus
Comparing forecasts
Revealing the neighborhood
Watching co-occurrences over time
54
UNCLASSIFIED
UNCLASSIFIED
“Nucleotide Resolution” (NR) Forecast Details
• Both USA and others have active NR research and are expected to grow over the near future
• Applications of the NR capability and enabling technologies can be identified
“Nucleotide resolution” linkage
to “pore” technologies
prompted partner discussion to
further exploration on
nanopores and pore technology
2022
forecasted
nucleotide resolution is taking on a predominantly
“interactive, bridging, contributor” function to SYNBIO
55
UNCLASSIFIED
UNCLASSIFIED
SYNBIO S&T R&D
International Overlap
Co-authorships
56
UNCLASSIFIED
UNCLASSIFIED
SYNBIO
and the
HOLSTEIN COW ?
57
UNCLASSIFIED
UNCLASSIFIED
URINALYSIS and QUANTUM SENSING ?
58
UNCLASSIFIED
UNCLASSIFIED
Questions
SAF/CDM/CDMR (ISSO)
Pelayo Fernandez
pelayo.fernandez.1@us.af.mil

Transforming Intelligence Analysis with Knowledge Graphs

  • 1.
    Senior Vice President,National Security Services Redhorse 30 Vincent H. Bridgeman Research Analyst / Project Manager United States Department of Defense Pelayo Fernandez Transforming Intelligence Analysis with Knowledge Graphs
  • 2.
  • 3.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. • Foundational ideas ̶ Computers can extend human intelligence ̶ Knowledge is networked information ̶ Intelligence analysis is a network problem • How do graphs transform intelligence analysis? ̶ Knowledge graphs are the right foundation ̶ Mission Graph responds to these imperatives ̶ Pushing the boundaries of graphs • Case study: ̶ Scientific Knowledge Forecasting with US Air Force Today’s Agenda 32
  • 4.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. Extended Intelligence Sensor Streams Core Computational Capability Free association, logic and ideation enhance analysis Validation and Integration Human Observation All human insights should be stored Human intelligence focuses on interpretation & invention Sensor Data Sensory Data CORE EI Prediction Logic Free Association Comprehension Memory Processing Analysis Abstract Thought Opportunity 1 Opportunity 2
  • 5.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. data information knowledge wisdom understanding relations understanding patterns understanding principles connectedness understanding Knowledge is networked information 1. We should automate this 3. So we can spend more time on this 2. And we should accelerate this
  • 6.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. “There are no single-node targets in intelligence. Even when we are concerned with a person, organization, object or location, our target is associated with some network – usually many networks – that are essential to understanding the target. These networks have become more complex over the years, largely because of advances in communications technologies…….The toughest job for intelligence analysts in any field….is to deal with networks” Target-Centric Network Modeling: Case Studies in Analyzing Complex Intelligence Issues Robert M. Clark (Author), William L. Mitchell (Author) CQ Press; 1st edition (April 24, 2015) Intelligence analysis is a network problem
  • 7.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. Examples of network problems in intelligence 36 Cyber Security Transnational Crime Counter Terrorism Counterintelligence Military Operations Counter Disinformation Technology Watch
  • 8.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. Manual Visual Digital Canvas Digital Networks We are in the third phase of network analysis tools 37
  • 9.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. Knowledge graphs are the right foundation
  • 10.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. What does a knowledge graph do? 39 Digital Asset Value-added data structure – connected data Acts as a “knowledge hub” in a software application/system Data Science Enabler Much of data science is graph-based math Network analysis can be applied to connections in the data User Interface Enabler Your mind already works like a graph – you comprehend the world as things and their relationships
  • 11.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. 40 Redhorse brings knowledge graphs to the intelligence mission Data Lake Entity Resolution Data Engr. Graph Workflows Graph UI Graph Database Other UI Public Data Commercial Data Proprietary Data Domo, Tableau, D3, ArcGIS, QGis etc. Bloom Senzing Hume AWS S3, Azure, Delta Neo4j Data Science
  • 12.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. Pushing the boundaries of graphs + AI for intelligence 41
  • 13.
    This document isproprietary and confidential. No part of this document may be disclosed or reproduced in any manner to a third party without the prior written consent of Redhorse Corporation. 42 Case Study: Scientific Forecasting with Temporal Knowledge Graphs for the United States Air Force
  • 14.
    UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED Science Knowledge Forecasting GraphSummit for Government Oct 2023 Distribution A: Approved for public release. 18 OCT 2023. Requests for this document must be referred to SAF/CDM/CDMR, ATTN: ISSO
  • 15.
    44 UNCLASSIFIED UNCLASSIFIED March 2021 Repeated callsto discover and understand emerging & disruptive science and technology Expedite generating insight: “speed to insight à speed to decision” NSPM-33 NDAAs FY18-FY24 October 2020 February 2022 “… (B) develop machine learning tools to identify possible future technologies; …” NDAA FY24
  • 16.
    45 UNCLASSIFIED UNCLASSIFIED How do weenable forecasting of emerging or disruptive technologies and their impact? What and where are technologies emerging globally? How do they apply to government programs? What government programs will they impact? How does the USA compare to others? Are we influencing their research or vice versa?
  • 17.
  • 18.
    47 UNCLASSIFIED UNCLASSIFIED PURPOSE: “Mining the nearfuture for action” - forecasting science and technology direction and determining impact to government ENVISIONED OUTCOME: Pre-emptive actions by leadership CONCEPT: Innovative application of machine learning algorithms in combination with temporal co- occurrence knowledge graphs; integration of massive complex databases to generate actionable forecasts and determine implications
  • 19.
  • 20.
    49 UNCLASSIFIED UNCLASSIFIED 66M articles 817k articles Web of Science Initial dataset Started with subsets of Web of Science … with greater resources, more data collections can be added, e.g., OpenAlex, ORCID, patents, etc.
  • 21.
  • 22.
    51 UNCLASSIFIED UNCLASSIFIED What is SyntheticBiology? alter genetic code read and manipulate genetic code combine genetic codes build genetic codes from scratch where biosciences, information sciences and engineering converge Control these processes and you can control or build your own life…
  • 23.
    52 UNCLASSIFIED UNCLASSIFIED Top 50 Projected GainIncrease nucleotide resolution ??? Details Role behavior over time Other projected gaining terms Top 50 Projected Term clusters, metrics, terms, US vs foreign pub rates Example: Synthetic Biology
  • 24.
    53 UNCLASSIFIED UNCLASSIFIED Exploring nucleotide resolution Curious co-occurringclusters US vs. PRC focus Comparing forecasts Revealing the neighborhood Watching co-occurrences over time
  • 25.
    54 UNCLASSIFIED UNCLASSIFIED “Nucleotide Resolution” (NR)Forecast Details • Both USA and others have active NR research and are expected to grow over the near future • Applications of the NR capability and enabling technologies can be identified “Nucleotide resolution” linkage to “pore” technologies prompted partner discussion to further exploration on nanopores and pore technology 2022 forecasted nucleotide resolution is taking on a predominantly “interactive, bridging, contributor” function to SYNBIO
  • 26.
  • 27.
  • 28.
  • 29.