© 2012 NACHA — The Electronic Payments Association. All rights reserved. No part of this material may be used without the prior written permission of NACHA. This material is not intended to provide any warranties, legal advice, or professional assistance of any kind.
Making voice biometrics work in real world implementa5ons
AKA, Beyond the bullet point
Mark Lazar CEO, Victrio
2 Proprietary © 2013 Victrio, Inc.
Victrio Intro
• Emerging leader in passive voice biometrics
• Team expertise in voice recognition and call center – Ex-Nuance, BeVocal, Tuvox
• Industry acclaim – Gartner Cool Vendor 2013 for Security:
Identity and Access Management
– RSA Most Innovative Company 2013 - Finalist
3 Proprietary © 2013 Victrio, Inc.
Deep Experience in Voice Biometric ImplementaAons for Financial Firms
• 3 years of implementation experience
• Clients include 3 of top 5 financial firms
• Screening 100M calls in 2013
• Largest global fraudster database in world
4 Proprietary © 2013 Victrio, Inc.
History of Industry’s Voice Biometric Efforts
1990’s
Schwab
Early 2000’s
Mid 2000’s
Late 2000’s
HSN
HarOord Insurance
MarrioP
US Bank
AT&T
Visa
Bank of America
First Horizon
Bell Canada
Bank of America
ABN/AMRO
Santander
American Express
5 Proprietary © 2013 Victrio, Inc.
1990’s
Schwab
Early 2000’s
Mid 2000’s
Late 2000’s
HSN
HarOord Insurance
MarrioP
US Bank
Visa
AT&T
Bank of America
First Horizon
TERMINATED
TERMINATED
UNKNOWN
INTERNAL ONLY
INTERNAL ONLY
TERMINATED
INTERNAL ONLY
TERMINATED
TERMINATED
TERMINATING Bell Canada
Bank of America
ABN/AMRO
Santander
American Express
Gap Between Theory and PracAce
6 Proprietary © 2013 Victrio, Inc.
Today: Even More ImperaAve to rescue call center
Customers are increasingly frustrated…
Web Attacks vs. Phone Attacks
While professional fraudsters aren’t stopped
65%: Frustrated
Source: Opus Research 2012 Survey Source: RSA Client Study
50%: Too time- consuming
7 Proprietary © 2013 Victrio, Inc.
Passive Voice Biometrics: Promising Theory
• No customer interruption
• Passive enrollment
• Passive authentication
• Alerts agent in real time – Shorter call and AHT
– Happier customer
– Stronger fraud detection
8 Proprietary © 2013 Victrio, Inc.
Victrio ID Confidence Cloud
Voice Biometric Analysis
Account Data
Call & Behavioral
Data
Voiceprint Match
ID Confidence Reported to
Agent
High
Fraud
?
Low
Caller Voice
Call Center / IVR
Mobile
Passive Customer Voiceprints
Shared Fraudster
Voiceprints
Signal Analysis
Ideal System Overview
9 Proprietary © 2013 Victrio, Inc.
Life in the Trenches: Fraud DetecAon
• Call center infrastructure is never ready on Day 1
– Call recording system poorly configured • Association with accounts not always available • Relevant metadata not available
– Clients have to assert ownership of data
• Need new people processes to support solution • Poor flagging and capture of fraud interactions • Poor understanding of call center “big data”
10 Proprietary © 2013 Victrio, Inc.
Life in the Trenches: Fraud DetecAon
• Business Model?
– Licensed product?
– Managed service? • Evolve system to anticipate evolving fraudster behavior
• Track changing impact of factors in multi-factor scoring
• Example: Growing usage of using $8 burner phones weakens
device detection; adjust weighting
11 Proprietary © 2013 Victrio, Inc.
Field Results: High Repeat Fraud Rate
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
BROKERAGE 71% 89%
BANK ACCOUNT FRAUD 42% 76%
CREDIT CARD FRAUD 72% 95%
12 Proprietary © 2013 Victrio, Inc.
0 20 40 60 80 100 120 140 160 180 200
Duke Kobayashi Jeffrey Johns
Tyler Miresh Charles Davis Robert Albert
Bill Koflin Josh Dumas Greg Barker Sally Davino Michael Lee
Shauna Koshik Alexandra Laurel
Patrick Donally Elizabeth Wright
Jerry Lucent Carla Gonzalez
Francis Montgomery Michael Henley Shauni Mccoy Mark Martigas
Brian Reed
Recurring PaPern: DramaAc Fraudster AcAvity ConcentraAon
TOP 20 FRAUDSTERS
~ Half of fraud activity perpetrated by top 10 Fraudsters
Calling > 30 times per month
C A L L S
13 Proprietary © 2013 Victrio, Inc.
Average Number of Attacks From “Sharks” Drops (Study of Top 10)
17
20 21
10.5 12
5.5 6.5
1.5 2 1 1
0
5
10
15
20
25
Jul-11 Aug-11 Sep-11 Oct-11 Nov-11 Dec-11 Jan-12 Feb-12 Mar-12 Apr-12 May-12
Average Fraud Events Per “Shark" Per Month
Fraud Events
Trial and error period
Churn/abandonment period (fraudster frustration)
19 calls on average
Previously successful
fraudsters give up after Victrio is in
place
14 Proprietary © 2013 Victrio, Inc.
Case Study: Dramatic Impact of Voice Biometric Fraud Detection
• 3:1 false positive ratio exceeds industry standards
• 82% reduction in baseline fraud since Victrio launch
• Strong endorsement from fraud operations
15 Proprietary © 2013 Victrio, Inc.
Life in the Trenches: AuthenAcaAon
• Affirm the need and business case – What are the authentication metrics today? – Translate differential into hard costs – How to value soft benefits?
• Better customer experience • Top of wallet • Churn and new customer acquisition cost
• Resolve privacy considerations – Engage legal and security early – Design optimal “consent” strategy
16 Proprietary © 2013 Victrio, Inc.
Life in the Trenches: AuthenAcaAon
• “Dual screening” maximizes accuracy and speed to answer – 3X accuracy with combined blacklist & whitelist vs. whitelist alone
• Multi-factor approach – Voice biometric, call and behavioral, account data, signal analysis
• Deliberate “graylist” process – When do you trust a newly enrolled voiceprint?
• Prepare organization with fraud detection first – Low customer exposure while you address the above
17 Proprietary © 2013 Victrio, Inc.
Field Results: AuthenAcaAon
• Over 28M calls analyzed
• Range of AHT reduction metrics
• Leverage historical call recordings to kickstart process
• Full customer base coverage over time
18 Proprietary © 2013 Victrio, Inc.
AuthenAcaAon Accuracy: Integrated Approach Drives Performance
• Whitelist: 3X accuracy with combined blacklist and whitelist versus whitelist alone
• Whitelist: Additional 25% improvement in accuracy when combined with metadata
• Blacklist: 2X lift in accuracy for blacklist when combining with metadata
19 Proprietary © 2013 Victrio, Inc.
Business Impact and Customer Experience
?
2,000 calls
Legitimate Callers 9,998,000 calls
Scre
en 2
: Fr
auds
ter V
oice
prin
t Scr
een
Scre
en 1
: C
usto
mer
Voi
cepr
int M
atch
97% Legitimate Callers
3% Legitimate Callers
~15 -25 Seconds Faster Authentication: $60M savings
Current Authentication
Caught Fraud 1,992 calls
Uncaught Fraud 8 calls
Greater Fraud Detection: $40M savings
Fraudsters
Live Agent Calls 10M per Month
20 Proprietary © 2013 Victrio, Inc.
Further materials and informaAon
• Web access to demos
• White paper
• Industry data
Mark Lazar, Victrio CEO