University of Colorado Boulder Background
Current visual authentication systems are plagued with errors. Many have problems with non-white males, while others can be fooled by siblings, family members, and even a 3D rendering. Systems that rely on visual imaging are also fooled by changes in hair color, facial hair, glasses, and makeup. Some systems that use visual images also require storage of a user’s photo, together with identifying information, such as the user’s name, address, date of birth, etc. This can present a security threat in cases of a data breach.
Researchers at CU Denver have developed an authentication system that reads a person’s unique thermal and 3D contour facial signature. By combining thermal and 3D depth imaging, a biometric facial authentication signature is created that uniquely identifies people and is blind to race, gender, or other sources of bias. Moreover, this system can authenticate based on partial signatures, and is therefore not fooled by changes in hair color, eye color, addition or removal of facial hair, glasses, makeup, etc. Instead, the system uses invariant features based on an individual’s biology, including a person’s unique vascular system, skin contour composition, and capillary pattern.
More accurate facial authentication
Authentication based on partial facial signature
Enhanced privacy for subjects due to encoded imaging
Personal device authentication (e.g. cellphones, computers)
Facial recognition software