challenges in introducing new identity technologies
TRANSCRIPT
Moderator – Rick Lazarick (CSC)
Panelists Tom Buss (Integrated Biometrics)
Tim Cleland (Retina Biomterix)
Reza Derakhshani (EyeVerify)
Welcome to the participating workshop attendees
Workshop Purpose (Lazarick)
Overview (Cleland)
Case Studies Mature (Buss)
Emerging (Derakhshani)
Promising (Cleland)
Audience Participation Adjourn at 10:25
Obstacles or challenges to new identity technology
• Technology proof of concept, performance
• Familiarity/Acceptance
• Robustness
• Patents/IP
• Perception/Education
• Accepted Standards
• Cost or Return on Investment
• Funding
• Privacy
• Many established modalities
• Universally implemented and accepted into the world marketplace
• Good science and statistically robust
• Drawbacks?
• De novo
• Multimodal
• Must be valid and reliable
• Must be usable and practical
• Other considerations
Two Broad Categories:
• Short term/practical considerations
• Long term/philosophical considerations
• Private sources • Friends and family /referrals- 85% of start-ups
obtain seed funding this way
• Crowd funding
• Angel Investors: gust.com
• Technology Incubators • University – Texas Venture Lab UT Austin
• Venture Capital • After seed funding/start up costs
• Usually during early sales/manufacturing
• A “good idea” will not generate investment
• Process is very expensive! • Patent search
• Patent lawyers-proper background
• USPTO-registered
• 95 % of first time applications are rejected—expect to rewrite!
• Must be new, nonobvious and useful (United States Patent and Trademark Office): • Process
• Machine
• Articles of manufacture
• Composition of matter (new drugs)
• Improvement of any of the above
• Inventions which are not useful or offensive to the public morality • Laws of nature
• Physical phenomenon
• Abstract ideas
• Works of art
• Must be novel
• Must be nonobvious • Thought to be the greatest hurdle
• Invention cannot be patented if it is achieved by combining known methods, techniques improvements or solutions to produce predictable results
• Adequately described
• Claimed by the inventor in clear and definite terms
• “Off-the-self” as much as possible
• Outsource only when necessary
• Human subject testing requires IRB approval • Institutional
• Central
• Depending on modality, may require FDA clearance:
• A 510(k) is a premarket submission made to FDA to
demonstrate that the device to be marketed is at least as safe and effective, that is, substantially equivalent, to a legally marketed device.
• Consultants can be helpful
• Get the word out! • Papers in scientific journals
• Web
• Press
• Clearly communicate intended application • Advantages/disadvantages
• Advisors/Mentors- NDA
• Partnerships with other biometric companies
Outline 1. What is LES fingerprint sensor technology? 2. What was motivation for it (why do it?)
a) What is the need b) Is there a customer to project ROI c) Why we succeeded
3. Current status, results 4. Technical challenges
a) LES film development b) Image processing algorithm development c) FBI certification to appendix F standard
i. Educate and explain physics of LES ii. Develop new measurement tools and convince MITRE iii. Verify via pixel by pixel comparison to certified technology
5. Market Acceptance Challenges a) Natural suspicion & resistance to NEW unproven technology b) Overcome belief that only “Optical” technology can achieve
highest image quality c) Educate and demonstrate to end users to allow LES in RFP’s
Case
Light Emitting Sensor (LES):
“electro-luminescent technology”
LES FILM
Circuit Board
CIS CAMERA
Lens Set
LES (Light Emitting Sensor) Technology is a smaller, lighter, lower cost, more forgiving alternative to Prism based Optical Technology which still achieves FBI appendix F image quality in a Form factor advantageous to Mobile biometric platform applications
Luminescence Layer
(5-15) Microns
Protective
Layer
Transparent Layer
(400 Nanometer ITO)
Volts AC
Fingerprint Ridge Valley
400-450 Microns
AC
Replace the Prism with LES film
Replace the Optics (lens) With TFT camera And ASIC pcb
Camera sensor PCB USB output
Camera sensor PCB USB output
USB output
Impact of LES on Size of FAP 45 devices
Watson
Sherlock
Dermalog Smallest FAP45 Optical device
Motivation for the creation of LES technology: 1. US DOD end users demanded alternative to mobile Certified FAP45
“optical” scanner technology to address their known weaknesses 2. Rapid move to “Mobile” applications required smaller, lighter, better,
lower power fingerprint scanner technology 3. We found an integrator/Solution Provider in need of the technology to
help fund the initial investment which reduced the financial risk.
Significant Advantages of non-optical LES technology as it applies to Certified Mobile Fingerprint scanning applications: 1. LES is un-affected by sunlight 2. LES does not require constant cleaning of latents on the platen surface 3. LES does not require a silicon membrane or messy moisturizers to
capture dry fingers, the biggest challenge in mobile fingerprint capture today.
4. LES is significantly lighter, thinner and lower in power consumption than optical equivalents.
Why do It?
What is current state of Acceptance?
WE have seen wide spread acceptance in certified mobile biometric applications. Watson Mini is currently being bid in a number of significant opportunities worldwide. Recent Large scale acceptance in Brazil of our two finger desktop scanner for a Brazilian Voter registration program consisting of 14,000 + Biokits being provided which include a Watson Mini for fingerprint enrollment and identification purposes.
Cross Match Seek Avenger, uses Sherlock aimed at DOD and Special Op’s; Multi-Modal
Northrup Grumman Bio-Sled: Adaptive boot for Sherlock with Samsung Android Phone; Market is Law Enforcement, DOD, Border Patrol Multi-Modal
LES Design Adopters
Coppernic FAP 30 Columbo – Multi-Modal, android based handheld
NeoScan 45™ Mobile Fingerprint Collection Device , This wireless device uses Sherlock and at 9 oz. is the smallest FAP45 fingerprint collection device in the industry. Target market is Public safety, Border Patrol, Intelligence Agencies
Amrel XP7: Ruggedized with Sherlock; Military, Law Enforcement, First Responder Markets; Android Based
Corvus Unity: Ruggedized Multi-modal device aimed at Military markets and border Patrol; Uses Sherlock ; Windows Based
Credence Trident 1: uses Watson-Mini; in use in Indonesia by Police; Multimodal device; Android based
Customer Design Wins
Booz Allen – Vampire DOD – Sherlock based product Target, Criminal, DOD, Border patrol
Main challenges we had to overcome Technical challenges
1. LES (Light emitting sensor) electroluminescent Film development (Particle chemistry, size, consistency, film construction, manufacturing techniques, environmental and durability performance)
2. Image processing algorithm development a) Calibration mask (compensates for Lens and Film uniformity) b) Dynamic gain/voltage adjust for moist and dry finger capture
3. FBI certification to appendix F standard
a) Educate and explain physics of LES to MITRE b) Develop new measurement tools and techniques and convince
MITRE of their equivalence to optical methodologies c) Verify via pixel by pixel comparison to certified optical technology d) MITRE first approached 7/2011, “F” Cert granted to Watson
7/2012 and Sherlock 1 year later
Market acceptance 1. Natural suspicion & resistance to NEW unproven technology
(people hesitate to try what they are unfamiliar with. Education, hands on demonstrations and establishing early adopters was necessary to bring the market around)
2. Overcome belief that only “Optical” technology can achieve highest image quality. (again, Education, hands on demonstrations and establishing early adoptors was necessary to bring the market around)
3. Educate and demonstrate to end users and government agencies to allow LES in RFP’s (RFP’s were written requiring Optical fingerprint sensors ONLY.)
Main challenges we had to overcome (continued)
New Challenges Ahead for LES
Continue to expand the product line to larger FAP Devices, ie. Ten print, Palm Print, higher resolution, etc…
Kojak, 10 Print Appendix F scanner
Complex and unique patterns, but hard to get to unless...
Hand and finger vein Retinal vasculature
???
Zero-effort scanning: natural no-gazing solution with front facing cameras in real world scenarios
Front facing cameras have lower quality sensors and optics, but highly acceptable due to popularity of selfies
The most natural interactions with a mobile slate: looking at it, touching it
Mitigation: computational photography Multi-capture SNR boosting (exposure stepping and averaging) followed by sharpening Vesselness/eyevein-tuned filter banks Cascade classification with assistive ocular micro features
Mobile use case imaging artifacts: lighting, motion blur, glare and specularities, eyelid/eyelash/glass occlusions
Mitigation: no reference quality metrics, robust matching/fusion to partial data
37
Top left: iPhone 5 front facing with regular capture, top right: same with added convex lens, bottom: same ROI with back facing camera.
Template security and revocability, entropy
Steganography
Device-bound isometric transforms
Revocable, private, and high entropy private key generation
Stable despite biometric intra-class variations
Always resolves to a private key, but valid sequence only achieved during successful genuine comparison
Market acceptance of a new technology
Statistically significant large dataset (US University)
Independent tests: iBeta, major payment company
Liveness (vs. spoof) detection Viable (though not scalable) attack vectors
Stunts Not a real threat but will leave a bad impression
Harder to prevent, plus the attacker just needs a single success for publicity
• Unfamiliar • Poor understanding of eye anatomy • Technologically difficult • History of unsuccessful implementations • Very expensive • Others
• Retinal identification first published in 1935 Drs. Simon and Goldstein1
• Robert Hill: EyeDentify, Inc. 1976 • Based in Beaverton, Oregon
• 53 employees in 1986
• 2001 ceased production of the “Icam” • At that time, only company (world wide) to use
retina scanning in a commercial device
1 C. Simon and I. Goldstein, “A new Scientific Method of Identification,” New York State Journal of Medicine, vol. 35, no. 18, pp. 901 – 906, Sept. 1935.
1 New York State Journal of Medicine, Sep 1935, Vol. 35, No. 18, pp 901-6
• Retinal Technologies, LLC • “System for capturing an image of the retina
for identification” 2007 patent.
• Used green and red LEDS for retinal illumination (visible light)
• No commercially available device
• Retina is known to be the most unique biometric
• Identical twins do not share same retinal vascular pattern1
• Vascularity of retina lends itself to liveness detection
1 P. Tower, “The fundus Oculi in Monozigotic Twins: Report of six pairs of identical twins,” Archives of
Ophthalmology, vol. 54, pp. 225-239, 1955.
• Information on the internet contains many inaccuracies (multiply reproduced)
• Confusion between iris and retina
• Cataract, macular degeneration, glaucoma, astigmatism, contact lens wear make imaging impossible
• Unable to overcome technological challenges
• Intrusive: device must contact the eye
• Dangerous and unhealthy
• Previous devices used 1980s technology • Long image acquisition times
• Visible light for retinal illumination
• No simple end hardware interface
• Operator dependent, not autonomous
• Others
EyeDentify ICAM 2001
Part Number ICAM 2001
Size 9.25 in. wide
6 in. high
4 in. deep
Weight 2 lbs.
Power 12 VDC
1.8 amps peak . 900 ma nominal
Verification time 1.5 seconds
False Rejection Rate 12.4 per cent (one-try) 0.4 per cent (three-try)
False Acceptance 0 Percent
User Capacity 3000 Users
Memory Retention Up to 5 years via the standard internal lithium battery
Transaction Storage 3300 transactions
ID Number Length 1 to 10 digits
Baud Rate 9600 K bps
Communications RS-232, RS-485/422, Wiegand 2-In/ 1-Out
• Public perception?
• 160 million eye exams per year (est)
• In your eye doctor’s office: • Scanning Laser Ophthalmoscopy
• Optical Coherence Tomography
• Fundus Photography
• Unfamiliar? • Poor understanding of eye anatomy? • Technologically difficult? • Very expensive? • Valid, reliable and robust? • Liveness detection? • Spoof-proof?
Audience contributions: Identify yourself and affiliation
Convey your experience or ask questions Tie comments/questions to topics in Framework
Be concise, allow for multiple contributors
Moderator will move the discussion along.