3.1. Face recognition

Face Recognition systemsBy Alexandr Notchenko

September 15, 2013

/whoami

Alexandr Notchenko

Neuroscience grad student at ITTPDegree in Technical counter-intelligenceInterests: Artificial Intelligence, Neuroscience

Work at . . .

Face detectionHaar cascade visualisation

(Face tracking)Real-time Grid reconstruction

Face recognition:Thousands of them.

Some events:

1960s at Panoramic Research, Inc.(semi-automatic)

Facebook – photo face tagging (Started 2011, May 2012 – aquire Face.com (100mil),Febuary 2013 – turning of in Europe)

Google – 2009 added FR to Picasa, search for faces

OptimEyes – more than 6k screens (UK, Germany, Poland, Switzerland, UAE, Oman, Kenya, Angola and South Africa.)

March 2012, “Hitachi Kokusai Electric” – 36mil

August 2011 - London Riots Facial Recognition (10k)

FBI's NGI - In a July 18, 2012 “on scope, on schedule, on cost and 60 percent deployed.” (by Lockheed Martin, for $1 billion ) (ready 2014 , with Iris, face, DNA, Fingerprint, Voice recognition)

April 2013 – Boston bombing fail

lScale:

lFace Recognition Vendor Test l(FRVT) 2006

lIris Challenge Evaluation (ICE) 2006

lSponsored by the National Institute of lStandards and Technology (NIST)

l2006, 100x better than 1995

Types of Attack:

• Photo-Attack (only 2D systems) Intruder displays photograph of correct face (Solved by Replay)ON THE VULNERABILITY OF FACE RECOGNITION SYSTEMS TO SPOOFING MASK ATTACKS(Neslihan Kose, Jean-Luc Dugelay)

• Spoofing mask attack (2D – 3D systems)Intruder creates a 3D mask with face featuresof victim.

Countermeasures for spoof mask attack

Countermeasures

privacy visor Mask

CV Dazzle

Face Recognition systems

The end.Any questions?