Improved Spoof Detection for Facial Recognition
As biometrics become more prevalent in our technology, the more important security and detection of illegitimate access attempts becomes. This technology improves the spoof detection capability of facial recognition technology, providing not only identification and rejection of a spoof face, but provide its rationale for rejection as well.
Description of Technology
This MSU-developed technology is capable of distinguishing a spoof face from a live face when presented with a print, replay, or mask attack. Using a novel neural network architecture and highly specialized training database, this software is not only capable of recognizing a presentation attack, but also provides intelligent feedback to the user. This “explainable artificial intelligence” feature is possible due to classification of attacks and the software’s ability to recognize spoof patterns.
- Classifies spoof vs. real face
- Reasoning for rejection stated
- Learns spoof patterns
- Trained with diverse database
- Allows for better performance across various races, genders, etc.
- Facial recognition software
- Biometric lock
- Surveillance systems
Licensing Rights Available
Non-exclusive rights available
Inventors: Xiaoming Liu,Yaojie Liu, Amin Jourabloo
Tech ID: TEC2018-0077
For Information, Contact:
Michigan State University