Weitere ähnliche Inhalte Ähnlich wie Towards the Next Generation Financial Crimes Platform - How Data, Analytics, & ML Are Transforming the Fight Against Fraud, AML & Cybersecurity -Nadeem Asghar (20) Mehr von Molly Alexander (14) Kürzlich hochgeladen (20) Towards the Next Generation Financial Crimes Platform - How Data, Analytics, & ML Are Transforming the Fight Against Fraud, AML & Cybersecurity -Nadeem Asghar1. TOWARDS THE NEXT GEN
FINANCIAL CRIME PLATFORM
Nadeem Asghar
VP Solutions Engineering & Field CTO
2. © 2019 Cloudera, Inc. All rights reserved. 2
CLOUDERA IN FINANCIAL SERVICES
82%Share of Top 100 Global
Banks
> 525
Customers
Global Financial Services:
25/29
Globally
Systemically
Important
Banks (G-SIB)
12/15
Top
Insurance
Firms
12/15
Top
Credit Card
Issuers
4/5
Top
Stock Exchanges
8/10
Top
Wealth
Management
Firms
4/4
Top
Card Networks
3. © 2019 Cloudera, Inc. All rights reserved. 3
INDUSTRY TRENDS THAT REQUIRE NEW APPROACHES
OPEN BANKING
AND FINTECH
• Open banking and fintech enables
payments from any digital device.
• Provides new financial crime
opportunities.
SOPHISTICATED
CRIME NETWORKS
• Criminals are well coordinated,
creative, and technologically
sophisticated.
• Any product, device, or
communication is a point of entry.
ELEVATED CUSTOMER &
INVESTOR EXPECTATIONS
Generating profits via rapid product
development can be at odds with risk
mitigation.
Reputational loss can devastate stock
valuation & revenue.
New crime typologies are enabled by new technology products demanded by customers
SOPHISTICATED
CRIME NETWORKS
4. The State of Combating Financial Crime
THE STATE OF COMBATING FINANCIAL
CRIME
5. © 2019 Cloudera, Inc. All rights reserved. 5
THE STATE OF COMBATING FINANCIAL CRIME
High costs, large losses, and little progress towards solving the problem.
Losses over 12 months
It is estimated that the combined
revenue lost due to financial crime
is $1.45 trillion over 12 months.
Spent over 12 months
Global financial institutions have
spent $1.28 trillion over 12
months combating financial crime.
Criminal proceeds confiscated
Despite tighter regulation and major
investment, only 1% of criminal
proceeds are confiscated by
authorities in the EU.
Source: Refinitiv - The true cost of financial crime — a global report (2018).
Fraud
($239 Billion)
Bribery and
Corruption
($309 Billion)
Money
Laundering
($267 Billion)
Theft
($209 Billion)
Cybercrime
($241 Billion)
Human
Trafficking
($188 Billion)
6. © 2019 Cloudera, Inc. All rights reserved. 6
“PERVASIVE” – ONE WORD DESCRIBES FINANCIAL CRIME
Financial crime is not by line of business and it is funded by victims.
GANG MASTERS ITALYVICTIMS FROM INDIA AND BANGLADESH
Legitimate Employer
Mr. A
ATM
$
$$$
$
$
Mr. B Mr. C
$
$
$
$
$
Loan
Loan
$
Investments
$
$
$
$
$
$$
$
$
HUMAN TRAFFICKING IDENTIFY THEFT FRAUD A MONEY LAUNDERING MACHINE
INDIVIDUAL VICTIMS RETAIL BANKING IMPACT COMMERCIAL BANKING IMPACT INVESTMENT BANKING IMPACT
Individual Account
Individual Account
Individual Account
Fake Company
Fake Company
Fake Company
Source: FATF REPORT Financial Flows from Human Trafficking July 2018
FROM ASIA
Individuals
TO EU
$2.5M in 2.5 years“Recruited” by
CRIMES across sectors
Impacts in all sectors
7. © 2019 Cloudera, Inc. All rights reserved. 7
NEW APPROACHES TO COMBATING FINANCIAL CRIME
Collective and collaborative ownership through public-private partnerships.
PRIVATE
UK Joint Money Laundering
Intelligence Taskforce
US
Joint Statement on
Innovative to Combat
Money Laundering
ASIA AML/CFT Industry
Partnership (ACIP)
8. © 2019 Cloudera, Inc. All rights reserved. 8
NEW APPROACHES TO COMBATING FINANCIAL CRIME
Innovation is a critical aspect in battling financial crime and regulators encourage it.
Joint Statement on Innovative Efforts to Combat Money Laundering and Terrorist Financing (December 2018)
• Encourage banks to consider, evaluate, and, where appropriate, responsibly implement innovative
approaches to meet AML and other illicit financial crime obligations.
• Pilot programs that expose gaps in a compliance program will not necessarily result in supervisory action.
• The Agencies will establish projects or offices that will work to support the implementation of responsible
innovation and new technology in the financial system.
9. © 2019 Cloudera, Inc. All rights reserved. 9
FINES ADD UP WHEN REGULATORS COLLABORATE
Source: DLA Piper: AML Bulletin Regulatory News Update – May 2019, p.9
OFAC
$657,016,283
+ + +
Federal
Reserve
$164,000,000
Department of
Justice
$150,129,359
Financial
Conduct
Authority
$128,854,358
Global criminal
networks.
Poor processes
and non
disclosure of risk.
Fraud
Ignored non-
financial data.
Did not
collaborate with
other banks.
One investigation can launch a chain of events of more investigations.
= $1.1B
$2-2.5
Billion
(pending)
~~
$15
Million
(2018)
=
Did not
collaborate with
other banks.
10. The State of Combating Financial Crime
CHALLENGES > SOLUTIONS
11. © 2019 Cloudera, Inc. All rights reserved. 11
PERVASIVE
CRIME
ENTERPRISEDATAPLATFORMENHANCE FINANCIAL CRIME SOLUTIONS TO BE HOLISTIC
“Mind the Gap” between siloed solutions with advance data analytics from the Edge to AI.
Edge Datacenter Private Cloud Public Multi-Cloud Hybrid Cloud
Catalog Schema Migration Security Governance
Streaming Ingest &
Analytics
Data
Engineering
Operational
Analytics
Data Science &
Machine Learning
Analytics &
Reporting
Generate synthetic data of
simulated fraud scenarios to
proactively test and improve
detection systems.
Industrialize machine learning
to tighten the loop between
detection and new money
laundering patterns
Adapt to evolving normal and
malicious actions by
simultaneously analyzing live
and historic behavior.
Apply alternative data
(eComms, voice, media,
machine logs, etc.) to link event
with malicious intent.
Flag suspicious activity that
deviates from card owners’
historic behaviors.
FRAUD
PREVENTION
Mitigate financial losses due
to fraudulent activities
Mitigate known crime patterns
with rules-based transaction
monitoring.
ANTI-MONEY LAUNDERING
Track movement of illicit
funds through the financial
system.
Analyze network telemetry to
flag abnormal behavior
(anomaly detection).
CYBER
SECURITY
Protect the organization from
digital intrusion.
Catch suspicious cross-border
transaction flows with market
surveillance systems.
INTERNAL
SURVEILLANCE
Monitor employee activity to
deter rogue behavior.
G
A
P
G
A
P
G
A
P
12. © 2019 Cloudera, Inc. All rights reserved. 12
Fraud
Prevention
Anti-Money
Laundering
Cyber
Security
Internal
Surveillance
SILOED ENTITIES AND GAPS (DATA, TECHNOLOGY, TOOLS, AND PROCESSES)
KEY DATA AND ANALYTICS CHALLENGES
VOLUMES OF
DATA
NEW STREAMING
DATA SOURCES
ANALYTICS
ARMS RACE
DISJOINTED SECURITY AND
GOVERNANCE
• All data needs to be
evaluated.
• Biometrics (voice,
facial, behavior)
• Video
• Call centers
• Legacy integration
• Reactive approach
and dependency on
historic data.
• Data security,
management, and
governance policies
are often• By 2025, worldwide data will
grow 61% to 175 zettabytes,
residing in the cloud and the
data centers*.
*Source: IDC. The Digitization of the World From Edge to Core (2018).
duplicative and inconsistent
across the enterprise.
• Inflexibility to scale, test, and
deploy.
• False positives
13. © 2019 Cloudera, Inc. All rights reserved. 13
PROCESS MODELSTOREINGEST ANALYZE DEPLOY
ENTERPRISE DATA PLATFORM
Best-of-Breed
financial crime
applications
Industrialization of
machine learning &
AI
Real-time,
streaming ingest
and analytics
Hybrid on premise
and public cloud
architecture
The latest open
source advances
Data Security, Governance & Compliance
Ingest
Streaming
Data
Ingest Data at
Rest
Ingest
Stream or
Batch Data
ALTERNATIVE DATA
ENTERPRISE DATA
TRADING DATA
BANKING DATA
EDGE DATA
Data Science
Tools
Data
Scientists
Business
AnalystsAnalytical
Tools
BI and
Visualization
Enable Collaborative
Innovation
KEY ATTRIBUTES TO ENHANCE FINANCIAL CRIME SOLUTIONS
14. The State of Combating Financial Crime
KYC AS A KEY ENTERPRISE STRATEGY
15. © 2019 Cloudera, Inc. All rights reserved. 15
CRIMINALS ARE CONSTANTLY INNOVATING
Synthetic Identify Fraud: The fastest-growing type of financial crime in the United States.
The largest synthetic ID ring detected so far racked up losses for banks of $200
million from 7,000 synthetic IDs and 25,000 credit cards. (2013)
Largest Credit Card Fraud
Synthetic ID fraud accounts for 10 -15 percent of charge-offs in
a typical unsecured lending portfolio.
Synthetic ID Fraud Charge-offs
• Synthetic IDs are created by applying for credit using a combination of real and fake, or
sometimes entirely fake, information.
• The act of applying for credit automatically creates a credit file at the bureau in the name of
the synthetic ID.
• Fraudster can now set up accounts in this name to build credit.
• The fraud is nearly impossible to detect because the credit file looks identical to real people
who have limited or no credit history.
Source: McKinsey & Company - Fighting Back Against Synthetic Identity Fraud (2019)
16. © 2019 Cloudera, Inc. All rights reserved. 16
SYNTHETIC IDENTIFY FRAUD
Why Machine Learning is NOT sufficient to significantly enhance Synthetic ID detection?
Lack of Sufficient Training Data - few synthetic ID fraud cases exist, so you can’t
train models and therefore deep neural network techniques that find associated
patterns are of little use.
KYC Data Is Incomplete - one needs a holistic (KYC) view of each customer.
Synthetic IDs look real - Unsupervised machine-learning techniques that look for
anomalies in data struggle, because there are few differences between real and
synthetic IDs at the time of application.
Source: McKinsey & Company - Fighting Back Against Synthetic Identity Fraud (2019)
17. © 2019 Cloudera, Inc. All rights reserved. 17
WHY YOU NEED THE NEXT GEN PLATFORM)
Third-party data sources allow banks to deepen their understanding of their customers.
Matrix for scoring profile
depth and consistency
From nine sources of external data, McKinsey researchers determined
the likely authenticity of identities based on data depth and consistency.
ü Holistic KYC
22,000 unique fields of information
Source: McKinsey & Company - Fighting Back Against Synthetic Identity Fraud (2019)
18. © 2019 Cloudera, Inc. All rights reserved. 18
SYNTHETIC IDENTITY FRAUD – SUPPORTED BY THE DARK WEB
Sources:
Experian – “Here’s How Much Your Personal Information is
Selling for on the Dark Web”
The Federal Reserve - Synthetic Identity Fraud in the U.S.
Payment System (2019)
19. © 2019 Cloudera, Inc. All rights reserved. 19
CUSTOMER USE CASE – STORE CREDIT CARD ISSUER
• Deepen fraud detection and
prevention mechanisms.
• Enhance the credit
underwriting process.
• Improve the customer
experience through
personalization.
CHALLENGE SOLUTION OUTCOMES
Leveraged the Hortonworks
Data Platform (HDP) as the
foundational data infrastructure
with Cloudera Data Science
Workbench to build high
precision fraud detection
algorithms.
• Seconds vs. days to uncover
fraudulent behavior.
• Credit line assignments
improved by 20% to 30%.
• Product offerings better
reflect individual needs.
High precision fraud detection
algorithms discern synthetic
identities and provide a holistic and
accurate view of consumers.
Improved credit line assignments
21. © 2019 Cloudera, Inc. All rights reserved. 21
NEXT GEN PLATFORM FOR TACKLING FINANCIAL CRIME
Ingest
Streaming
Data
BANKING DATA
Data Flow (CDF)
Data Science
Workbench
FINANCIAL CRIME APPLICATIONS
Data
Scientists
Data
Processing
Data Eng.
Data
Warehouse
Operational
DB
Catalog | Schema | Security | Governance
Business
Analysts
EDGE DATA
ALTERNATIVE DATA
Enterprise Data Store
ML
Cyber Security AMLFraud Surveillance
Analytical
Tools
BI and
Visualization
Ingest
Data at
Rest
Deploy Models
Ingest Stream or
Batch Data
Teams
speaking
the same
language
ENTERPRISE DATA
TRADING DATA
`
Ingest
1
3
4
2
22. © 2019 Cloudera, Inc. All rights reserved. 22
THE NEXT GENERATION FINANCIAL CRIME PLATFORM
The Enterprise Data Cloud from the Edge to AI
A Unified View - improve financial crime performance by eliminating the
constraints and costs of siloed business, analytic and data environments.
• Real time streaming, embedded analytics and dynamic transaction scoring.
• Break down siloed divisions between AML, Fraud, Cybersecurity & Surveillance.
• Industrialization of ML & AI to support dynamic model updating and monitoring.
• Latest open source advances ensures adoption of latest technologies and algo’s.
• Ensures an agile environment for development and deployment of applications
• Contributes to improved behavioral insights and Customer 360 capabilities.
ü Data From the Edge
ü Enterprise Wide Platform
ü Data Science & Simulation
ü Open Source Innovation
ü Hybrid/Multi-Cloud
ü Enhanced KYC
23. © 2019 Cloudera, Inc. All rights reserved. 23
REFERENCE ARCHITECTURE
Example for holistic financial crime platform
Tx
Party
Account
Alternatives
Streaming Layer
Batch Layer
Ingest
Serve
Hive Spark Flink
Spark
Streaming Kafka Streams Flink
Real-time Layer Operational Store
Historical Store
HBase
Solr
Kudu
HDFS
Object Store
CDSW
Endpoint Phoenix
Impala
HiveNiFi
Kafka
BI &
Visualization
Analytics
Operations &
Governance Control Plane Atlas Ranger HMS SMM
CDSW
Hue
25. © 2019 Cloudera, Inc. All rights reserved. 25
MASTERCARD
The search capabilities for a
customer facing fraud
prevention platform didn’t
satisfy increasingly complex
customer queries on
hundreds of millions of
fraudulent businesses.
CHALLENGE SOLUTION OUTCOMES
Delivered dynamic scalability
and improved performance to
accelerate searches, enrich
searching capabilities and
increase search accuracy.
• +5X annual searches
supported.
• +25X daily searches per
customer.
• Expanded business model
to new markets.
Creating new revenue streams with an
advanced anti-fraud solution.
INCREASE IN CUSTOMER INTERACTION
26. © 2019 Cloudera, Inc. All rights reserved. 26
WESTERN UNION
New digital money transfer
services required a variety of
new and more sophisticated
authentication techniques.
CHALLENGE SOLUTION OUTCOMES
Enterprise Data Hub to deliver
an omni-channel customer
experience.
Machine learning models
managed in Cloudera Data
Science Workbench
• Real-time risk decisioning,
fraud alerts, and compliance
enhancements.
• Multi-channel customer
insights that drive consumer
engagement.
A multi-prong approach to countering
fraud with machine learning has
brought dividends for Western Union
in the form of dramatically reduced
fraud rate.
FASTER DATA LOADING
27. © 2019 Cloudera, Inc. All rights reserved. 27
UNITED OVERSEAS BANK
Legacy databases restricted
the amount as well as the
variety of data that can be
used.
CHALLENGE SOLUTION OUTCOMES
• Enterprise Data Hub as
the principal data
platform.
• Cloudera Data Science
Workbench as the
enterprise data science
platform.
• Reduced from 3 months to 3
weeks the time to identify
suspicious relationships.
• 1,000+ hours saved
evaluating global client
networks and spotting new
opportunities.
“With Cloudera and machine learning
technologies, we were able to
enhance AML detection and reduce
the time to identify new links from
three months to three weeks.”
PERSONALIZED RECOMENDATIONS
28. © 2019 Cloudera, Inc. All rights reserved. 28
Comprehensive customer
insights was impossible due to
“multiple versions of the
truth” resulting from
inconsistent data sources and
processes.
CHALLENGE SOLUTION OUTCOMES
Implemented a single data
platform that supports all
workloads, including self-
service analytics, operational
analytics, and data science.
95 new proactive financial
crime control alerts protect
3.7 million individual
customers from poor
outcomes.
Unlocking the power of data for
comprehensive customer insights,
millions in savings, and better crime
investigation.
REDUCED INFRASTRUCTURE COSTS
31. © 2019 Cloudera, Inc. All rights reserved. 31
WHY IT MATTERS - UNIQUE CAPABILITIES OF CDP
Cloud
Optimized for IT & LoB
(hybrid, multi-function, SDX,
open, container-based cloud
experiences)
Cloud Burst
(supplement on-prem
capacity)
Intelligent Replication
(data, users, workloads)
Best of CDH & HDP
(Cloudera Runtime)
ENTERPRISE DATA
CLOUD
34. 34© Cloudera, Inc. All rights reserved.
KEY DIFFERENTIATORS
Comprehensive streaming platform – Only big data vendor to offer a comprehensive streaming platform from
real-time data ingestion, transformation, routing to descriptive, prescriptive and predictive analytics.
100% open source technology – Only vendor with this strategy; prevents vendor lock-in
300+ pre-built processors – Only product to offer such comprehensive connectivity from edge to enterprise
Built-in data provenance – Only product in the market to offer out-of-the-box data provenance on data-in-motion
3 Streaming analytics engines – Only vendor to offer a choice of three streaming analytics engines to customers
for all their streaming architecture needs
36. © 2019 Cloudera, Inc. All rights reserved. 36
Manage data, pipelines, models
Automate pipelines
Deploy models anywhere
Monitor performance
DEPLOYDEVELOP
Make teams more productive
Explore shared data
Open, extensible workspaces
Collaborate with peers
TRAIN
Scale resources efficiently
Train models
Tune parameters
Track performance
MANAGE
Run anywhere with a common architecture
Instantly enable new teams
Manage access and resources
Scale cost with usage
THE PLATFORM FOR INDUSTRIALIZED AI
End-to-end machine learning infrastructure for teams at scale
37. © 2019 Cloudera, Inc. All rights reserved. 37
WHAT INDUSTRIALIZED MACHINE LEARNING LOOKS LIKE
Predictive
Services
BI Tools and
SQL Editors
Data Products
DATA, METADATA, SECURITY, GOVERNANCE, WORKLOAD MANAGEMENT
MACHINE
LEARNING
DATA
ENGINEERING
DATA
WAREHOUSE
OPERATIONAL
DATABASE
Sensors/IoT
Devices
DATA FLOW &
STREAMING
38. © 2019 Cloudera, Inc. All rights reserved. 38
MACHINE LEARNING FOR CLOUDERA DATA PLATFORM
Cloud-native enterprise machine learning as-a-service
• DS/ML from research to production
• Purpose built for Seamless Data Engineer and Data
Science workflows
• Seamless scale-out experience
• Rapid provisioning and elastic autoscaling
• Automatic dependency management
• Distributed CPU and GPU model training
• Managed service on CDP
• No (Spark) clusters to manage
• Containerized multi-cloud portability
• Private cloud option with CDP-Private
CML AI Runtime
CDP Kubernetes
Object Store (S3 or ABFS)
40. © 2019 Cloudera, Inc. All rights reserved. 40
SOLVING CYBERSECURITY AT SCALE
An architecture for real time cybersecurity analytics
41. © 2019 Cloudera, Inc. All rights reserved. 41
Flexibility– Gain flexibility with extended protection capabilities.
100% open source big data technology – Open source data platform enables companies
to ingest, manage and process ALL security data at massive scale.
Pre-built data ingestion dataflows– Metron is able to handle a wide variety of security
data sources out of the box.
Pre-built Machine Learning and Data Science workbench – Enable automation of
threat detection with precision, using custom or subscribed models.
KEY DIFFERENTIATORS
Established Data Model – Metron provides a robust and canonical schema that
recognizes a wide variety of security data sources out of the box.
Enterprise
Scale
Data
Science
Protect
Data
Catalog
Tools