2. Centrifuge Corporate Overview
Visual data analysis pioneer since 2007
Genesis in US National Intelligence Sector
Funded by Novak Biddle Venture Partners
Used by thousands of analysts in both public
sector and commercial organizations
Headquartered in McLean, VA
Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 2
3. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 3
Social Network Analysis (SNA)
Sentiment Analysis
Claims Fraud
Telephony / Text Messaging
Gang Activity
Evidence Discovery Analysis
Financial Money Laundering
Serial Crimes
Cyber & Computer Network Analysis
Analyzing large volumes of diverse data
Targets trying to avoid detection
4. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved.
Commercial Examples
Problem: Criminal networks arrange procurement of medication through
low-cost government contracts, then sell on open market for high profit.
Problem: Sophisticated fraud rings of patients, providers and pharmacies
conspire to circumvent law . Difficult to detect with via traditional analytics.
Drug Profiteering
Prescription Fraud
4
BIG “Pharma”
Healthcare
5. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved.
Third Party Recognition
5
Honored as a 2012 FinTech Innovation Lab participant
for Wall Street-focused startup companies
Recently chosen by Gartner as a 2013
“Cool Vendor” for Security Intelligence
6. Centrifuge Systems
Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 6
Data Remains at the Source
• No need for redundant storage silos or client-side resources
Client Access via Lightweight Browser
• Browser-based clients promotes both collaboration and security
Easy and Secure Collaboration
• Collaboration performed at server
• Data does not transfer ownership during analysis
Modern Server-based Java Architecture
• No feature lock-in, Interact with other best-in-class applications
• Highly scalable with no client-side resource requirements
7. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 7
No Data Preprocessing or Redundant Data Storage
• Avoid the time & cost of preprocessing / prepositioning target data sets and
the resulting storage management overhead (hardware, security, backup, etc.)
• Data remains on its original server and native format allowing quick access
to newly discovered or updated data;
• Templates can created and saved for ease of future analysis.
8. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 8
Client Access through Lightweight Browser
• Browser-based clients can connect from anywhere, promote
ad hoc collaboration
• Client issues eliminated, extreme server scalability
• (RAM, storage, conflicts, etc.)
• No data touches the client, greatly reducing privacy concerns
9. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 9
Collaboration constraints removed
• Collaboration occurs on a secure server, NO data touches a client
• Data is not in jeopardy if client is “compromised”
• Collaborators can share data without transfer of ownership, greatly reducing
security and privacy concerns
• Two ways to share
• Static PDF image
• Fully functional Centrifuge Dataview, each collaborator can contribute
Static, or Live
and Interactive
10. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 10
Leverage other best-in-class capabilities via standards-based
Java/JavaScript and REST APIs, plus the included Solution Factory
• Early adopter of Ozone Widget Framework (OWF)
• Our focus is visual data analytics, not user lock-in
• We readily interface with best-in-class partners as determined by your needs
Unstructured data
Entity extraction
Fuzzy matching
Web harvesting
Geo-tagging
Statistical Algorithms
NLP
Embedded
11. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved.
VISUAL DATA ANALYTICS
See Clearly Now May 6, 2013
11
A Powerful Lens for Data Discovery
13. Copyright 2012 Centrifuge Systems Inc 13
Scenario: Health care insurance provider looking for prescription fraud
Dataset: Claims data relating drug, pharmacy, prescribing doctor and consumer
Initial Graph Using Opacity
Using Size
Applying iterative
filtering
14. Copyright 2012 Centrifuge Systems Inc
Highlight Paths and Degrees of Separation
14
Scenario: Cyber Security – investigating network traffic and identifying potential threats
Dataset: NetFlow data with originating and destination IP addresses and users/orgs
1. Identify all paths
Visually highlight all paths that connect subjects of interest
2. Identify & isolate path of
interest
3. Iteratively explore suspicious elements
1. Start with suspects 2. Reveal nodes associated
with suspect organizations
3. Reveal where traffic is
being directed
4. Reveal users associated
with suspect traffic (targets)
15. Copyright 2012 Centrifuge Systems Inc
Choose Link Layouts & Placement
15
Force-Directed
Radial
Circular
Algorithmic layouts allow placement of nodes to be based on different criteria allowing
natural patterns to emerge
Hierarchical
16. Graphically Select or Filter
Select and filter from one visual to another
Narrow results, highlight subsets
17. Query Build Out
As “link-ups” are run, the new “associates” are brought into the picture – building out
our relationship graph based on added connections. “Link ups” can be added for as
many associates as desired.
18. Copyright 2012 Centrifuge Systems, Inc. All Rights reserved. 18
Contact us for more information
Russ Holmes
Centrifuge Systems, Inc.
7926 Jones Branch Dr., Suite 210
McLean, VA 22102
(571) 830-1300
www.centrifugesystems.com
Hinweis der Redaktion
Centrifuge beganby addressing the needs of the Intelligence CommunityLarge volumes of seemingly unrelated dataJoin together disparate sources quickly and without regard to the original formatExtreme time sensitivityLooking for activities or relationships that were actively trying to avoid detection
We’ve now been embraced across both national and local governmental organizations, as well as within the commercial sector.
We’ve now been embraced across both national and local governmental organizations, as well as within the commercial sector.
BYOD: Bring Your Own Data
BYOD: Bring Your Own Data
BYOD: Bring Your Own Data
BYOD: Bring Your Own Data
BYOD: Bring Your Own Data
In this simple example of a two-variable relationship graph involving warranty claims data, we display the relationships or connections between customers (green) and warranty claims (blue). As is typically the case, the vast majority of relationships are “normal”, in this case meaning a one-to-one relationship between claims and owners. These types of data connection are often referred to as “noise” in the term “signal to noise ratio”. Most data is pedestrian and useful to an analyst for nothing, other than masking, through layers of “routine activity” noise, those activities that are truly novel.One relationship of interest in this data set is the magnified one in the top left that shows 3 customers, two of which have an unusually large amount of claims associated to them, and who both share a single claim with each other and a third customer. This would certainly warrant closer inspection.
Now we jump to a more complex graph. This is based on a use case for a health insurance company what wanted to analyze the relationship between doctors, pharmacies and consumers to identify anything out of the ordinary – doctor-shopping, excessive prescriptions etc.Size can be used to visually represent the node based on any metric - either native data or derived - Opacity – individual icons for nodes can be brighter or lighter based on any metric – simplest metric to use is the number of neighbors.Also, the entire graph can be adjusted in layers so that either nodes, labels are emphasized.