Technological advances have improved security measures in businesses, offices, shops and other facilities. Today, many people prefer to use modern security equipment and devices to secure their buildings. Security cameras, fingerprint readers and facial recognition systems, to name but a few! The use of biometric facial recognition has increased dramatically in modern times. It is the most popular biometric technology. It makes the building safer.
The application uses a complex algorithm for identity verification. It takes into account the unique features of the face that do not age over time. There are eighty such nodes on the face. Few of these nodes are taken into account by a facial recognition system. The facial data of registered persons is stored in the system’s database as a facial model. The next time the person looks at the camera, the details of the face are compared with the models in the database. The registered person will be recognized.
No security is infallible. Biometric facial recognition is no exception. It does not guarantee 100% accuracy. There has been some controversy over its excessive use. Many complain that it does not offer 100% protection. It is prone to errors and can make mistakes in the recognition of registered persons. This is true to some extent.
Let us now discuss about some metrics used to evaluate a face recognition algorithm. Two errors can be illustrated below:
False Acceptance Rate (FAR): This is considered a fatal error in the facial recognition system. It gives unauthorized access to applications that can be disastrous. The database here contains the templates of the registered persons. These are the only people who need to be recognized. However, the system can also incorrectly identify people who are not registered. This gives them access to confidential data or premises, thus endangering the safety of the individual.
False Rejection Rate (FRR): This case is intended to occur more often than the False Acceptance Rate. Here, a person can be registered in the facial recognition system. However, for certain reasons, it still cannot be recognized by the application. The reasons may be numerous; for example, the lighting conditions may not be adequate or the registered person may not have positioned him/herself correctly in front of the camera. It’s not that bad, but it can be very frustrating for the user.
FAR and FRR reduce the accuracy of biometric facial recognition technology. The patterns upon which relies the algorithm have a big influence on the overall accuracy. If the algorithm needs too much facial detail to match, the risk of False Rejection Rate is higher. If it needs too few facial details to match, there is a higher risk of False Acceptance Rate.
Modern devices are designed with the possibility of all these errors in mind. Most suppliers promise accuracy of over 90% in their use.
Even modern applications should not be confused with a picture. The facial recognition system only identifies living faces. It is not easy to mislead or abuse. Although it is not completely waterproof, it is more accurate than any other means of identity verification. In highly sensitive areas, primitive means of security cannot be relied upon.
