Cyber Liability - Insurance Risk Management and Preparation

Cyber Liability Insurance:
A proactive approach to
managing risk
A NEW SAAS MODEL TO ADDRESS CYBER INSURANCE RISK
MANAGEMENT FROM ADAPTIVE SOLUTIONS

Our Presentation on Cyber Risk
• The Adaptive Solutions SaaS model in strategic alliance withWillisTowersWatson
• The Cyber Risk Insurance Market - background and current state
• Cyber Attacks… some current statistics
• Preparation of the “To Be” State: The Adaptive Cyber Security SaaS Platform
• DarkLight – enhanced cyber security effectiveness through ontology driven machine learning

SaaS based cyber liability risk management
• Adaptive Solutions LLC has announced a new cyber risk management program for enterprise
class clients
• Working withWillisTowersWatson, the largest broker of cyber insurance in the US, we will
develop programs for both insured and insurer
• Our solution will let you visualize data governance, lineage, traceability, retention, and
management throughout your organization with the Adaptive Metadata Management™ suite
• We will improve the effectiveness of cyber security efforts through analysis, deconstruction
and prediction of cyber attacks with our strategic partner DarkLight™
• These tools will better address the challenges of your operating environment with targeted
savings in insurance premium throughWillis
• We will provide post-attack analysis for leading insurance carriers and proactively prepare
digital assets to better withstand and recover from cyber attacks and further reduce expense

Key elements of the Adaptive SaaS offering –
what is involved ?
• Adaptive Metadata Manager, highest risk business unit first
• DarkLight Cybersecurity
• Implementation by Adaptive Solutions and Meta Informatics
• In partnership with WillisTowers Watson to design and deliver actual insurance products with
demonstrable effectiveness for Insurance carriers AND clients

The Cyber Risk Insurance Market - background
and current state

A quick background on the pervasive nature of Cyber Risk
• Cyber attacks are a constant threat to businesses around the world with vast sums of money
being spent to protect against them.
• While in 2015, 40 percent of attacks stemmed from ‘outsiders’, a surprising 60 percent were
actually perpetrated by company insiders.
• IBM, who produced the figures based on information from over 8,000 of their clients devices,
revealed that although 15.5 percent of such ‘attacks’ were caused inadvertently, 44.5 percent
were deemed to have been malicious.
• An insider is defined as anyone who has physical or remote access to a company’s assets. IBM
noted that although this would often be an employee, it can also mean business partners or
maintenance contractors – people you trust enough to grant system access to.
• Insiders not only have this access, they may also be aware of your weaknesses and thus exploit
them more effectively than an outside agent might be able to.

Cyber Risk Coverage – Market Players
• U.S. insurers are cautiously underwriting cyber coverage
• The biggest challenge is to understand the true nature of the underlying risk
• While there are about 50 insurers that are writing some cyber coverage, the
market is dominated by five underwriters:
• Ace Ltd.
• American International Group Inc.
• Beazley P.L.C.
• Chubb Corp.
• Zurich Insurance Group Ltd.

Cyber Risk Insurance Market – Background
• Cyber coverage represents a significant area of opportunity for underwriters
• Some analysts predicting that the size of the cyber insurance market will
grow to $10 billion in the next five to 10 years
• Although this market is immature at the moment, there is still value to be
found if insurers properly underwrite risk
• Currently, cyber coverage predominantly is written on a claims-made basis
and primarily covers third-party liability in the United States
• About 90% of the premium volume for cyber — estimated by Lloyd's of
London to be $2.5 billion in 2014 — covers U.S. risks
• The market will increasingly demand tools to mitigate risk and manage
claims adjustment expenses post attack

The Cyber Risk Insurance Market
• Cyber insurance has emerged as a response to growing number of data
breaches worldwide and the extent of damage that they cause to businesses. Data
breaches are perceived as one of the leading risks to businesses as, among other
factors, they can have a huge influence on the company earnings.
• In the United States, the average cost of cyber crime amounted to 12.69 million
U.S. dollars in 2014. As well as financial costs, cyber crime has a negative
impact on employee morale, business reputation and relations with the clients. It
is not surprising, then, that companies have started to look for ways of protecting
themselves against cyber threats.

Cyber Insurance Market – current state
• In 2014, 54 percent of global companies were insured against loss of income due to
data breach, while more than half of the companies without cyber liability
insurance considered purchasing it.
• The share of businesses with cyber insurance worldwide increased with company
revenue. Only 3.8 percent of companies with revenues lower than 2.5 million U.S.
dollars owned cyber insurance.
• Among companies with revenues exceeding five billion U.S. dollars, this number
was equal to 25.9 percent.

Cyber Insurance Market - current state
• In the United States, 33 percent of companies owned cyber liability insurance in 2014.
In that year, the U.S. industry sector with highest share of companies purchasing the
insurance was the financial services sector.
• The average limit of purchased cyber liability insurance by the U.S. financial
institutions sector amounted to 23.5 million U.S. dollars. More than 82 percent of U.S.
companies reported that they were able to buy cyber insurance that met their
needs in 2014.
• The companies not protected by cyber insurance cited a lack of insurance fitting their
needs on the market, as well as low policy limits or too high costs, as the reason for
their lack of protection.

Cyber Insurance - Challenges
• S&P said that cyber risk presents a “unique challenge” for underwriters
because neither frequency nor severity is predictable.
• Reliable Actuarial data is also unavailable.
• Metrics for cyber risk also are in the early stages of development, and
probabilistic models pose high levels of uncertainty, mostly because of the
unpredictable human behaviors associated with cyber attacks” the report said.
• Other challenges for underwriters include limited and insufficient disclosures
about cyber attacks, the report said.

Cyber Attacks… some current statistics

Statistics
and facts
about
businesses
and cyber
crime in the
U.S.
• According to the IC3, the monetary damage caused by reported cyber
crime in 2014 amounted to more than 800 million U.S. dollars.
• That year, the U.S. state with the highest amount of losses was California
with over 131 million U.S. dollars in reported cyber crime damages.
• The average cost of a company-directed cyber crime attack in the United
States was 15.42 million U.S. dollars.
• Based on the type of attack, industry figures estimate the number of
days necessary to solve a cyber attack on a company can take up to 62.7
days.
• The most common types of cyber attacks experienced by U.S. companies
as ofAugust 2015 were viruses and malware.
• According to a 2015 survey of U.S. companies, the most popular cyber
securities deployed were advanced perimeter controls, firewall
technologies, and extensive usage of encryption technologies.

Statistics
and facts
about
businesses
and cyber
crime in the
U.S.
• Despite these efforts to protect the company from outside
cyber attacks, there are many employee activities that render a
company vulnerable, such as mobile device usage or remote
work access.
• Other obstacles to implementing more robust cyber security
solutions for businesses are the lack of funds as well as the lack
of clarity regarding best practice.
• Overall, 42 percent of SMB owners in the United States
regarded cyber security expenditure as a cost of business with
36 percent of IT security layer spending being directed towards
the network layer.
• Furthermore, 27 percent of internal costs due to cyber crime
were allocated towards detection.

Total Cost of Cyber Crime
The statistic shows the amount of damages caused
by cyber crime reported to the IC3 from 2001 to
2015. In the last reported period, the annual loss of
complaints referred to the IC3 amounted to 1.07
billion U.S. dollars, up from 781.84 million U.S.
dollars in 2013.
In 2014, the United States accounted for 83.96
percent of complainant losses.
No data available on reported cyber crime losses in
2010.
The numbers refer to internet crimes reported to
the governmental Internet Crime Complaint
Center.
Methodology of evaluating loss amounts: FBI IC3
Unit staff reviewed for validity all complaints that
reported a loss of more than $100,000. Analysts
also converted losses reported in foreign
currencies to dollars. The final amounts of all
reported losses above $100,000 for which the
complaint information did not support the loss
amount were excluded from the statistics.

Type of Cybercrime and Loss
This statistic presents the types of
cyber crime with the highest amount
of victim losses in 2015. During the
reported period, online confidence
fraud accounted for 203.39 million
U.S. dollars in reported victim losses.
In 2014, the United States accounted
for 83.96 percent of complainant
losses.

Types of Cyber
Attacks
This statistic shows the types
of cyber crime attacks most
commonly experienced by
companies in the United
States.
During a 2015 survey of 58 U.S.
companies, it was found that
97 percent of respondents had
experienced malware attacks.
The most common type of
attacks were viruses, worms
and trojans.

Average cost of a breach
The statistic shows the average
organizational cost to business in the
United States after a data breach. In
2016, the average cost to businesses
affected by a data breach in the
United States amounted to 7.01
million U.S. dollars.
Total breach costs include: lost
business resulting from diminished
trust or confidence of customers;
costs related to detection,
escalation, and notification of the
breach; and ex-post response
activities, such as credit report
monitoring.

Average annual costs
related to Cyber Attacks- by
Industry
This statistic shows the average
annualized costs caused by cyber
crimes in the United States as August
2015, sorted by affected industry
sector.
That year, cyber crime caused an
average annualized loss of 16.45
million U.S. dollars in the technology
sector.

Cybercrime Loss Given a
Successful Attack
This statistic shows the estimated
damage a successful cyber attack will
cost a U.S. business.
In 2015, the maximum total annualized
cost of cyber crime committed against
U.S. companies amounted to 65.05
million U.S. dollars.

Number of days to
resolve a Cyber Attack
This statistic shows the average
number of days necessary to
resolve a cyber attack in U.S.
companies as of August 2015,
sorted by type of attack.
That year, U.S. companies need an
average of 41.3 days to resolve
web-based attacks.

IT Environments and
Cyber Attacks
This statistic gives
information on the IT
environments targeted by
cyber attacks worldwide in
2015, sorted by industry.
During the survey period,
it was found that 34
percent of cyber attacks
aimed at the professional
service industry were
targeted at corporate or
internal network
environments.

Largest Data
Breaches revealed
to-date

Cyber Liability - Insurance Risk Management and Preparation

What do these
statistics tell us
about the
evolution of
Cyber Threats ?
• Threats posed by internal actors is the most significant;
in Finance and Insurance, this is effectively 100% of the
source of cyber risk
• E-Commerce is the largest threat to retail and travel;
Point of Sale fraud is largest for Food & Beverage
• Data breaches increasing in size and number of affected
parties
• Time to Resolution has improved, highlighting industry
education and prevention
• Most damaging attacks remain internal
• Preparation is the best policy
• So how to prepare ?

Preparation of the “To Be” State: The
Adaptive Cyber Security SaaS Platform

How does Adaptive Solutions propose to
revolutionize Cyber Liability Underwriting?
• Rudimentary underwriting
• Lack of defined risk metrics
• No means of identifying affected
data
• No traceability or lineage for post-
breach analysis and remediation
“As Is” State
• Identify and measure against key
metrics impacting risk
• Use preventative and analytical
tools to understand depth of event
and remediate/repair
• Create a “data inventory” which
catalogues both data and lineage
Transform the
approach • Inventory of key data assets and
traceability/lineage for breach
analysis
• More effective cyber security
infrastructure
• “Learning” bots to assist with
volume of attacks
“To Be” State

Cyber Liability Insurance – Underwriting
Considerations
• What EXACTLY is being protected ? Or what exactly was affected
by the breach ?
• What are the key underwriting metrics ?
• How is the risk priced ? How is this determined ?
• What are typical policy exclusions ? Retention ?
• What are typical loss scenarios ? Recovery scenarios ?
This all needs to be discussed, documented, and linked to
technology that offers actionable solutions

The End State must specifically address the Threat Matrix
• External actor
• Access through a vendor
• Through ISP
• Through DNS/Brute Force
• Internal actor
• Disgruntled employee
• Actively placed sleeper mole
• Internal incompetence (like passwords in a desktop Folder
labeled “Passwords” – Come on man !)
• Things in Common
• Major Losses
• Lasting Damage
IMPACT
MATRIX
External to
Firewall
Internal to
Firewall
Malicious
Insiders
82 95
Malicious
Code
76 89
Web based
attacks;
phishing;
email fraud
92 84

Manta
Adaptive
Library
Data Model
Databases
Messages
Event Logs
Big Data
Enterprise
Database
“To Be” State - Cyber Risk Management with the Adaptive Repository Orchestration
Transform
Business Glossary
Metric Glossary
BI Objects
Data Objects
Analytic
Software
Reports
BITools
Business
Logic
Model
Data
Logic
Model
Physical
Model
Business
Verticals
NoMagic
Data Object
ETL
Metrics
Business
Process
Model
Terms
Business
Concepts
Integration
Process
People
Data
Quality
Applications
Business
Engagement
Model
DarkLight
Environment

How does it work ?
• We combine the disciplines of robust data governance and cyber
security through the application of world class technology
• Catalogue the key data assets by business unit and function
• Inventory the data assets, establish lineage and relation
• Implement an integrated cyber security solution
• Our solution lets you understand the key risk metrics BEFORE the risk
is assumed
• How stable and “orderly” is the client data environment ? Data quality ? Points of
access ?You better know before you bind the risk…
• How do we do that ? Adaptive for lineage, governance, security, permissioning,
versioning, and data tracing; DarkLight for cybersecurity enhancement

Cyber Liability
Insurance –
Pre-SaaS
Underwriting
Review
• Understand the data environment
• By Business Unit
• By Data Source
• Understand the network environment
• Number of IP and Email addresses
• Web Sites
• IoT access
• Create enterprise data lineage and traceablity to establish base case
and identify data quality, loss, and retention issues
• Catalogue the data assets being protected and identify the key
stakeholders of each
• Integrate the DarkLight cyber solution with the Adaptive Metadata
Platform to enhance cybersecurity
• Integrate the underwriting review to prepare the SaaS solution for
the specific client

Cloud ServicesBig Data Platform
Metadata Connector
ODBC
RESTful
Custom SDK
Security
Semantic Layer
Templates
Data Store
Scheduler
Templates
- DG Maturity
- DM Compliance
- DQ Maturity
- DA Maturity
- DG Ownership
Data Landscape
* Build customized UI by enhancing
Adaptive’ s UI
Templates
SaaS “Hosted”
Client’s Data Landscape
Data Factory
HostedSolutionExistingDF
Client’s Virtual
Data Excellence
Internal Users
- Data Stewards
- Data Owners
- CDO
- CIO
Reports
- Scheduled Reports
- Monthly DQ/DG Snapshots
- Monthly Data Compliance
- Alerts & Notifications
- On Demand Reports
Functional Data
Architecture Supported
- Data Modeling
- Data Governance
- Data Quality
- Production Support
Key elements of the Cyber Risk Management with the SaaS model
1
2
3
4
5
Client Service Provider Onsite Service Provider Offsite
Data Owners Business Analyst
Data Analyst
Data Analyst
Data Analyst

Legacy systems and cyber risk management
• Legacy systems pose a unique risk to an organization
• Failure to migrate to modern platforms complicate risk
management and recovery post attack
• Most firms delay migration due to theThree P’s:
• pain, personnel, and price…
• We greatly reduce theThree P’s in legacy migration and
management with automated data discovery and documentation
tools
• We offer this on a SaaS basis using open standards

Data governance and lineage tracing –
A live client example

Bank Client – establishing data governance and lineage with
huge amounts of data
• The technical truth of architecture and data flow within a large organization
is nearly impossible to understand for any user without technical
experience.
• We automatically scan the dataflow to identify all of these objects and links.
Because programs, procedures and scripts do refer to exactly these
technical objects, this may cause a huge complexity in the meta data
repository.
• As for example, one Client’s Group Business Intelligence Repository holds over
320,000 tables, columns, views, entities, attributes, report fields and dimensions.
• Approximately 10% are interesting from business view.The others are used for layer
concept, arch. Principles, compliance, performance, Interfaces etc.
Page 37
When scanning the raw technical data automatically to generate data models and data
lineage, one very important aspect are considerations on how to manage the complexity for
different users, and how that data is presented.

In this example, assume for a moment you’ve been hacked… and you need to identify
all the affected data - In this Client case, search for affected data elements with a name
“customer” returns 45,315 Results (across all object types)
Post intrusion
data analysis...
The impact
across an
organization can
be daunting and
difficulty to track
down all of the
affected data

Presentation
Layer
Search
Results II
Page 39
Filter to Relational.Column reduces to 8,243

Presentation
Layer
Search Results
III
Page 40
Classification “Data Store” equal Group Data Pool reduces to 524

Presentation Layer /
Search Results IV
Page 41
Classification “Table Layer” equal Business Data reduces to 123… now its manageable

Presentation Layer / Classifications
Page 42
How to manually classify, track and trace lineage for > 300,00 Objects? You can’t…
The “Rules Engine” can inherit classifications via CWM connection
Type: Relational.Schema
Name: LDDAPPL
Data Store: LDD
Type: Relational.View
Name: ALL_CUST
Data Store: LDD
table layer: LDD Views
Type: Relational.Column
Name: id_customer
Data Store: LDD
Table: layer: LDD Views
Type: Relational.Column
Name: short_name
Data Store: LDD
Table layer: LDD Views

Presentation
Layer
-
Virtual business
layers for lineage
Page 43
Although restrictions for data
flow are defined, lineage can
become very big and complex.
The goal is to be able to
automate the lineage tracing
process and narrow presentation
to mazimize effectiveness

Zoom of Lineage
Page 44
And tracing lineage can become overwhelming – the majority of expense post
cyber attack is identifying and tracing affected data…

Managing the Presentation Layer /
Virtual business layers for lineage
Page 45
Adaptive uses smart algorithms to traverse through the lineage and pick only defined
columns to show them in a textual view: “Column is derived from CEE IF Fields”
Rules can be based on
- Classification
- Owning Schema
- OwningTable
Rules allow the user
define and manage their
data environment from
top to bottom

Investigation one – identify the affected data
Page 46
1. Locate the desired data
element (in this case, a
COGNOS Field)
2. Display the reverse lineage
3. Identify the Source Field
4. Investigate single
transformation steps further, if
needed
This allows the client to manage
the complexity and volume of the
data environment

Investigation two /
establish and understand the lineage
Page 47
Source Code: Insert Statement at given line number.
Lineage SVG Graphic
Selfmade “PLSQL” object view of column
PLSQL object view. Gives line number.

Investigation Three:
Repeat for other 3 Targeted Fields
Page 48

Investigating the
interconnection of the
target data element
Page 49
This graphic shows the lineage of
just one of these data source
columns in Reporting.
Failing to understand internal data
lineage is not a good idea. Our SaaS
soultion will link data to business
terms and concepts to trace data.
Post Cyber Attack, this is the major
driver of expense in post attack
investigation, management and
remediation.

As a result of this work, our Client
gained a deeper understanding and
tangible simplification of their data
lineage.
In this example, the Client used
Adaptive to link business terms and
concepts directly to source data to
establish lineage and a governance
framekwork for regulatory compliance
and financial reporting.
This To Be State allowed them to
directly link business concepts and
source data, using automated lineage
tracing and data governance
capabilities.
Report Field Source Table Description
B2 - COLL
before HC:
Resid. Real
Estate
S3_EXPOSURE Basel 2 figure. Collateral value
from SAS BEFORE Haircuts
(deductions) divided to
Exposure-sets; Optimization
after recoverability of collateral;
COLL -
Market
Value
allocated:
Comm. Real
Estate
COAL_COLL_EX
P_CRR
Collateral value that considers
all collaterals independent of B2
eligibilitiy. Collateral Market
Value from SAS; distributed by
SAS Coll Type; capped with
Exposure
B2 - COLL:
Comm. Real
Estate after
Haircut
S3_EXPOSURE Basel 2 figure; Collateral value
from SAS AFTER Haircuts
(deductions);
divided to Exposure-sets;
Optimization after recoverability
of collateral.
COLL –
Accepted
Value
allocated:
Resid. Real
Estate
COAL_COLL_EX
P_CRR
The distributed collateral
acceptable value in the way of
SAS CRR acceptable value
algorithm in EUR.
Page 50

DarkLight Solutions
-
Enhanced Cybersecurity through ontology driven
Artificial Intelligence

AI-Driven Analytics and Automation
► Timeline
 2009: PNNL funded research
 2013: Company founded to transfer technology to market
 2016: Commercial release, deployed in production
► Artificial Intelligence based on Semantic Graph Analytics
 Patented Advanced Reasoning Platform; Two granted, several in process
 Proprietary AI engine captures, automates and scales human expertise
 Applicable to cybersecurity, fraud analysis, and the global movement of money,
etc.

► Scarcity/high turnover of seasoned cyber analysts
 Outnumbered and overworked - projected shortfall of 1.5 million Analysts by 2019(1)
 Enterprise knowledge leaves with analyst – 18 mo. ramp to get “proficient”
► Staggering volume of cyber attacks creating “big data” issue
 Existing technology investments are underutilized while threats persist
 Staff is “drowning in data”
Challenges in the Market
(1) http://www.csoonline.com/article/2953258/it-careers/cybersecurity-job-market-figures-2015-to-2019-indicate-severe-workforce-shortage.html

Challenges in the SOC
 Not enough analysts to address the volume of alerts  New or junior Analysts not productive enough
 Not enough analysts to address the volume of events  Can’t fill open requisitions for SecurityAnalysts
 Knowledge leaves org when Analysts leave
 Too much time spent monitoring rather than
responding
 Incident response time takes too long
 No centralized process or tools OR
 Too many tools, not enough coordination between
 Analysts waste time manually attributing and
documenting incident response
 Analysts waste time chasing down false positives  No Analysts dedicated to hunting
 Existing alerts or select security feeds are ignored  Other – domain specific

Drowning in Data vs. DarkLight
Source: “The Cost of Malware Containment,” Ponemon Institute, January 2015 Survey of 630 IT / IT Security Practitioners in US responsible for detecting, evaluating and/ or containing malware infections.
17,000
ONLY
19%
Malware alerts
received on average
by an organization in
a typical week
Alerts investigated; Only 705
Exposed to risk of remaining 15%
Deemed “reliable”
4%WASTED: 395 Hours/week due to False
positives/false negatives
LOSTVALUE: $25K/week or
$1.27 million/year/org
Typical Industry Experience Customer, deployed in production
100%
Alerts Examined n=9500
1,816
additional alerts/wk
investigated
Lowered Risk by
investigating
previously ignored
alerts
Improved IRR and
increased
utilization of
existing security
investments
With 30%fewer staff,

A Force-Multiplier for your Analysts

The Cybersecurity “Big Data” Problem
Wisdom
Operational
Cybersecurity
Knowledge
StructuredCybersecurity
Information
Cybersecurity Data
Science of Security &
Semantic Infrastructure
Cybersecurity Measurement
and Management
Cyber Ecosystem
Technology
& Data
Human Intelligence and Reasoning

DarkLight™ - Human-quality analytics, at scale
Fuses data from disparate intelligence sources
Unifies network sensors + threat intelligence + enterprise context
Improves IRR on existing security investments
Captures analyst knowledge for retention by the enterprise
Augments deductive and investigative skills
Prevents “brain-drain” while accelerating training of new staff
Force-multiplier which enhances human reasoning
Acts as aVirtual Analyst, improving performance by 10X to 100X+
Advanced, AI-based reasoning able to infer conclusions

DarkLight Reference Model
Alerts Events
Adversarial Knowledge
Threat Intelligence - Internal/External
Feeds Incident Response System
Cyber
Ecosyste
m
Security:
Firewall,
Proxy, AV,
IDS/IPS,
Network
Devices
(Sensors)
Enterprise Knowledge:
AD, Legacy Data Sources, HRIS,
etc.
NotifiesAnalyst Directly
TriggersOrchestration / Action
in other product

► Results
 Improved ROI: Doing more with 30% fewer analysts by reducing false positive alerts
 Improved Situational Awareness: Now analyzing 220 previously ignored data streams
 Expanding to threat hunting
Customer Success
5000 employees, One of 17 DoE National Labs; Performs classified and unclassified research for
DoE, DoD, DHS and other government agencies.
100K+ alerts per day, 2.5B events/week through Splunk
“Increased our effectiveness from 5-15% to 90-95%”
BJ Stephan, Deputy CISO

DarkLight PROs (Programmable Reasoning Objects) analyzing data
Once data is ingested into DarkLight, the Programmable
Reasoning Objects (PROs) go to work – in real time –
analyzing thousands or tens of thousands of events in
seconds.
The purpose of these are to make inferences on sets of
data whether contextual, working, or both. To put their
use into perspective, each PRO acts something like an
analyst assigned to finding correlations between different
data sets and records ranging from thousands if not tens
of thousands of logs.
In order to find any disruptive or dangerous activity
analysts must spend hours or days searching for these
patterns. DarkLight alleviates this by incorporating PRO
reasoners to do this daunting task for the analysts,
leaving them with a condensed data set to work with.

Results: Summary and Graph View
Several views work together to provide
the full picture about a single event. The
Working Memory view contains lists of
PRO Output Types and indicates how
many items are in each type. Clicking on a
working memory type loads those events
into the Events view where they can be
sorted by date. Clicking on a single event
populates the tabular Results view, the
graphical Results Graph view, and the
Processors View.
Since different users prefer to see
information in different ways, the user
may select which views are shown – and
save them as a perspective.

Graph View of Event with full attribution
In this example, DarkLight correlates a FireEye event with a vulnerable host, attributing CVE, device and employee details.
The ResultsGraph view is
a node/link graph that
describes the selected
item in the Events view.
It contains all of the
properties and objects
that have been attached
to the event as it works
its way through ingestors
and PROs.
Each new object gets a
different color.

DarkLight Event Orchestration
PROs do the heavy lifting to reason and analyze, saving time
…and based on the results, can alert, notify or orchestrate other systems to take action

Semantic Technology 101
• Semantic Graph Databases
• Description Logic Reasoners

Graph Databases – Big Data
• A graph is a data
structure
• A graph holds data
• Schema (ontologies)
• Facts (assertions)
King
Line Manager
192.168.5.164
Edwards
Jones
Employee
Employee
Employee
Log On
Event
rtedward
rtjones
hasAccountName
Project Manager
hasRole
imking
4624

Automated Reasoning
• Also known as an Inference Engine
• DARKLIGHT is a framework for
supporting multiple reasoners
• Each DARKLIGHT Reasoner (called a
PRO) examines the known facts
and asserts new facts based on the
axioms of cybersecurity.
Karen
Ryan
Roger
Known Fact
Inferred Fact

Old Ineffective Method:
• Read all facts into a single
monolithic graph
• Manage the logical consistency
of the large graph
• Use a single Reasoner over the
entire graph
Monolithic Reasoner
Monolithic Graph
WARNING:
All facts asserted into a graph
MUST be logically consistent or
the Reasoner will not function.
- AND -
The larger the graph the harder
it is to keep it logically consistent.

Our Innovative Method:
• Read all facts into a single
monolithic graph
• Manage the logical consistency
of small subgraphs as they need
to be reasoned over
• Use MULTIPLE Reasoners over
the graph, not just one
Monolithic Graph
PRO PRO
INNOVATION:
It is easier to maintain
consistency in many
smaller graphs than
one large graph.

Hierarchy of PROs
Contextual
Memory
Graphs
DARKLIGHT Configuration
Working Memory
(Main Semantic Graph)
Known Facts Known & Inferred Facts
DARKLIGHT is a
Framework for
Reasoners

Contextual Memory
Working Memory
1. Trigger
2. Collect
3. Reason
4. Publish
The PRO Lifecycle
PRO Memory
5. Clear

DarkLight PROs in Use
False-Positive
Reduction
Insider Threat
Alerts
Data Enrichment
& Enhancement
Data Exfiltration
Suspicious
Command
Execution
Multiple Sensor
Correlation
Contextual Knowledge Maintenance
AnalysisHeartbeat FilteredFEIPSAlert
AttributedFEWebMalwareObject IPUserRecordCleanup
AttributedFEEMPSAlert MaliciousDomainMatch
AttributedFEEMPSAlertNotification MaliciousDomainMatchNotification
AttributedFEMalwareCallbackAlert NonNameServerFEDomainMatchAlert
AttributedFEMalwareCallbackAlertNotification NotifiedOnlyFEIPSAlert
*AttributedFEWebInfectionAlert Attributes SuspectPing
*AttributedFEWebInfectionAlertNotification SuspectPingNotification
AttributedFEWebMalwareObjectObjectNotificatio
n
TypeCountNotification
AttributedMaliciousProcess UnattributableEmailAddress
AttributedMaliciousProcessNotification UnattributableIPAddress
AttributedNetcat UnattributableUsername
AttributedNetcatNotification UnattributedEventNotification
BlockedFEIPSAlert WorkingTypeCountReport
BlockedFEIPSAlertNotification 1102 – The audit log was cleared.
ContextTypeCountReport 4672 – Special privileges assigned to new logon.
DHCPRecordCleanup 4798 – A user’s local group membership was enumerated
FEIPSAlertForVulnerableHost 4799 - Security-enabled local group membership enumerated
FEIPSAlertForVulnerableHostNotification 5156 - Windows Filtering Platform has allowed a connection
FEIPSAlertReport 5140 - A network share object was accessed
FEIPSAlertReportNotification 7045 - A service was installed in the endpoint
FEIPSAlertWithHostVulnerabilities 4624 - An account was successfully logged on
FEIPSAlertWithVulnerability 4663 - Attempt was made to access an object, File or Registry
Key
Force-Multiplying “Virtual Analysts”

DarkLight
-
Description
• DarkLight is the only patented system that embraces the human
decision making process and knowledge to combat cyber threats.
DarkLight was created, tested, and proven at one of the nation's most
advanced research laboratories, spanning more than four years of
R&D.
• DarkLight intelligently processes the massive data streams from a
current network and security appliances through a patented formal
Description Logic Reasoning Framework and Semantic Graph
Analytics.
• Unlike all other workflow-driven or machine learning-based
automation tools, this approach more effectively models normal
and abnormal user and network behavior.
• DarkLight’s Reasoning Engine is used to interpret and analyze facts
using an analyst’s unique knowledge of cybersecurity and the
enterprise, including the policies and compliance requirements of the
organization they are protecting. By utilizing the analyst rather than
black box or statistical models, the system becomes a true force
multiplier of expert experience and knowledge.

Applying DarkLight PROs to Detect Insider Threat
• By representing common sense knowledge from the cybersecurity
community and the knowledge from your enterprise's cybersecurity analysts,
tasks and data interpretation can be efficiently and intelligently automated.
• Because the DarkLight PRO (Programmable Reasoning Object) is created by
the security analyst themselves, it thinks and works like a human, and it can
be created to find any correlations and patterns between data sets.
• This gives your analyst the ability to create custom PRO's to track whatever
activity they deem necessary to keep your enterprise secure.

Ontologies and Threat detection
• Thought leaders at the CERT InsiderThreat Center at Carnegie Mellon's Software
Engineering Institute (SEI) have recently released new model concepts to help insider
threat programs to implement more effective controls.
• Based on cases from more than 1000 organizations, the research paper and models
have been several years in the making and provides a standardized method of
expression for indicators of potential malicious insider activity.
• They have identified an ontological approach to the problem and have provided the
industry with an InsiderThreat Indicator Ontology (ITIO).
• An ontological approach provides a standard common language with which to
represent and share knowledge, a factor they have identified as currently lacking
within the threat intelligence community.

Applying DarkLight PROs to Detect Insider
Threat
Examples of InsiderThreat PROs:
• Track group membership over time
• Detect off-hours system usage
• Detect uploading to known file-storage locations
• Detect unusual program execution
• Detect unusual printing activity
• Correlate when a member of a group decimated by layoffs uploads to a known location
• The InsiderThreat Indicator Ontology

Superior performance through reasoning
• DarkLight approaches the Cyber Security problem by allowing analysts to
explicitly establish what is the "normal" user behavior baseline in the context
of the enterprise business model and operations.
• For example, a compensation specialist working in HR should not be downloading customer
data; that is not part of the employee's normal user profile or approved behavior or UEBA.
• Understanding every employee, vendor and customer profile and behavior is
at the heart of what DarkLight gives the internal enterprise cyber security
analysts, followed by the ability to alert and act quickly.
• DarkLight offers the user a means to perpetuate their know-how via our
exclusive PROs. Other UEBA providers require an entity to use their machine-
learned models of the user and/or its peers.

Operationalizing the ITIO
• You can put CERT’s ITIO (as well as models like STIX,CybOX, OpenIOC and others) to work today with
DarkLight, for a force-multiplying, cyber analytic and automation platform.
• Import the InsiderThreat Indicator Ontology to DarkLight and the general concepts of the ontology are
mapped to real-time data of your organization.
• As an example, data of the “Actors” are mapped to “People & Organizations” of the company,
immediately leveraging the ITIO. Once this mapping has occurred, the hard problem of InsiderThreat--
identifying the subtle changes in an employee's behavior--can be identified much more easily.
• DarkLight helps you:
• Find the indicators
• Identify exfiltration
• Identify I.D. theft and fraud
• Collect the intelligence needed to allow efficient forensic investigations of affected assets.

Next Steps – when
do we start ?

Thank you for your
attention!

Cyber Liability - Insurance Risk Management and Preparation

Related slideshows

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

Recommended for you

More Related Content

What's hot

What's hot (20)

Similar to Cyber Liability - Insurance Risk Management and Preparation

Similar to Cyber Liability - Insurance Risk Management and Preparation (20)

Recently uploaded

Recently uploaded (20)

Cyber Liability - Insurance Risk Management and Preparation

Editor's Notes