Svetlana V. Drachova-Strang
CPSC 681
Feb 23, 2006
Biometrics is use of person’s unique physiological and behavioral, or physical characteristics to determine or verify identity in order to grant access to a protected location or device.
Biometric systems use:
• Physiological Characteristics: fingerprint, retina, iris,
DNA, hand geometry, voice and face recognition
• Behavioral Characteristics: signature, keystroke behavior
• Convenient – always with you, it’s what you are
• Unique for each person – genotype + phenotype
• Stable for lifespan – throughout aging and deceases
• Untransferable to another – can not give, borrow, or steal
• Hard or difficult to fake – liveness is unfakeable
• Quantifiable – counted, numbered, expressed by mathematical means
• Fingerprint identification
• Iris Scan
• Retina Scan
• Hand geometry
• Face Recognition
• Voice Recognition
• DNA ?
• Signature Verification
• Keystroke Behavior
• Others
• What is a “latent fingerprint” ?
• Fingerprints are unique – how about identical twins ?
• Fingerprints are persistent – how long do they last ?
• Fingerprints are well understood – they’ve been around
• Fingerprinting technology is inexpensive …
• What’s in a fingerprint – features (ridges, valleys) and patterns (arches, loops, whorls). loop arch whorl
Prehistoric:
• In Babylon – clay tablets with business transactions
• 14 th century Persia – fingerprint impressions on government papers, uniqueness is observed
• In Ancient China – thumb prints on clay seals
Europe
• 1686 – Marcello Malpighi, professor of anatomy, U of Bologna wrote a treatise on ridges, loops and spirals.
• 1823 – John Evangelist Purkinje, a professor of anatomy, University of Breslau, published his thesis discussing 9 fingerprint patterns.
• 1858 – Sir William Herschel, Chief Magistrate of a district in Jungipoor,
India, used fingerprints on native contracts. Later used palm prints and right thumb. Noticed their uniqueness and permanence.
• 1880 – Dr. Henry Faulds, British Surgeon-Superintendent published an article where he discussed fingerprints as a means of identification, and the use of printers ink as a method for obtaining such fingerprints.
There were others …
America
• 1882 – Gilbert Thompson of the U.S. Geological Survey in
New Mexico, used his fingerprints on a document to prevent forgery. First known use of fingerprints in the United States.
• 1902 – First systematic use of fingerprints in the US by NY Civil Service
Commission for testing. Henry P. DeForrest pioneers US Fingerprinting.
• 1903 – NY State Prison used fingerprints on prisoners.
• 1904 – First repository: National Bureau of Criminal Identification.
• 1905, 1907, 1908 – US Army, Navy and Marine Corps begin using fingerprints.
• 2005 – FBI’s IAFIS has > 49 million records.
• Minutia based – assigns XY coordinates to ridge branches and endings and stores them in file.
• Correlation based – examines overall pattern of ridges and valleys.
… several other techniques…
Many image enhancing algorithms.
Both described methods have drawbacks, neither keeps images, only data. Recreation of the fingerprint image from data is impossible.
• Not 100% accurate.
• Reported FAR/FRR rates 3-5%.
• Quality of the print is critical, many factors involved.
• Fakeable
Yes, numerous ways and recipes on the web: latex fakes, molds, images, cadaver finger, etc…
• Liveness Tests are needed to ensure identity: blood pressure, skin luminosity, resistance, temperature.
Biopoint reader USB reader USB reader
PCMCIA reader IBM laptops
• Criminal Identification
• Homeland Security
• Personal/Child Identification
• ATMs
• Laptops and desktop computers, handhelds, cell phones
• Jobsite clocking in/out devices
• Car locks
• House locks
• Airport security
• Safes, vaults, etc.
Companies:
Numerous:
Identicator, Biocompare, VeriFinger,
… many many others …
• Use geometric shape of the hand to determine user’s identity.
• Camera gets 2 orthogonal 2-D images of the palm and side of the hand.
• Get 90 measurements: finger width, height and length, distance between joints, knuckle shapes.
Advantages:
• Need to keep small amount of data, no large images to save.
• Simple algorithm.
• Devices are rugged.
Disadvantages:
• Hands are not highly distinctive.
• Devices are large and expensive.
• Error rates 2-10%
• Difficult to fake a real hand
• Embedded liveness test
Uses:
• Controlled accesses in the banks, and other places
• Airport security
• Supplemental Security
Companies:
• Dominated by Recognition Systems, Inc.
NOVAtime Technology, Inc.
Kantech, Inc.
• Retinal scan uses blood vessel patterns for identification
• Retinas are unique – even in twins and both person’s eyes
• Exceptionally accurate – came from ophthalmology, less than .5% error rate.
• Non-intrusive – shines infrared into your eye, no scary lasers !!!
• Persistent – deceases and injury do not change it.
Advantages:
• No, no way to replicate retina, cadaver retina deteriorates too fast.
Disadvantages:
• Very Expensive hardware
Used:
• Access to highly secure areas, military installations, power plants.
Companies:
• EyeDentify dominates the market.
• Emergent company Retinal Technologies.
Scanning airline passengers
Access Scanner at Workplace
• Iris scan analyzes random variations in the visible features: speckles, furrows, rings.
• Irises are unique – even in twins and both person’s eyes
• Exceptionally accurate – came from ophthalmology, less than 1 in a million error rate.
• Persistent – deceases and injury do not change it.
Advantages:
• No, no way to replicate iris, systems detect pupil dilation.
Disadvantages:
• Very Expensive hardware
Used:
• Access to highly secure areas, military installations, power plants, airports.
Companies:
• EyeTicket, IrisGuard, IrisPass,
Iridian Technologies, LG Electronics.
Iridian Technology, Inc.
Panasonic iris scanner
Iridian Technology, Inc.
Measure and analyze overall structure and shape, distance between eyes, nose, mouth, jaw lines.
May include:
• Facial geometry
• Facial thermogram
• Skin pattern recognition
• Smile recognition
Most systems use straight 2D
Images – “mug shots”
Advantages:
• Inexpensive equipment – can use a small digital camera connected to a PC
• Fast
Disadvantages:
• Liveness required to detect face molds
• Should be use as part of multimodal system, combined with facial thermogram for example.
• Error rate is high: 4 in 100 faces
Face Recognition Face Thermogram
Skin Pattern Scan
Dual pin/face scan system Face recognition system
NEC 3D Face Recognition system
DNS is different from other biometrics:
• Part of genotype, unique, except in identical twins.
• Does not employ feature extraction, needs to compare actual samples.
• Intrusive, requires a physical sample such as blood, skin cells, hair, etc.
• Can’t be done in real time yet, though processing is getting shorter.
• The odds that two people have same DNA profile is 1 in a billion.
• Process not fully automated.
How it works:
• Every person’s typing is unique, determined by neuro-physiological and behavioral aspects.
• Templates are built beforehand by having user type dynamic or static text sequences and analyzing them.
• User’s typing is compared with the stored template, either once at login, or over the duration of the session, of at certain intervals.
• Access is granted if current typing behavior matches template according to preestablished probabilities.
Measurements used in building a template:
• length of time each key is held down
• length of time between keystrokes
• key pressure
• typing speed
• keystroke sequences involved in capitalization.
• tendencies to switch between a numeric keypad and keyboard numbers
Advantages:
• Easy to implement and use: no HW required, relies on SW
• Low cost or free
• Non-intrusive
Disadvantages:
• Not much data is currently available about its accuracy
• Has to be a supplemental authentication technique
• Influenced by too many factors: hand injuries, fatigue, arthritis, temperature, body position, type of keyboard used.
• Convenient – always with you, it’s what you are
• Unique for each person – genotype + phenotype
• Stable for lifespan – throughout aging and deceases
• Untransferable to another – can not give, borrow, or steal
• Hard or difficult to fake – liveness is unfakeable
• Quantifiable – counted, numbered, expressed by mathematical means