作者: Vedrana Krivokuća , Waleed Abdulla
关键词: Construct (python library) 、 Biometrics 、 Matching (graph theory) 、 Computer science 、 Pattern size 、 Authentication 、 Fingerprint (computing) 、 SAFER 、 Minutiae 、 Artificial intelligence 、 Pattern recognition
摘要: Instead of using the entire minutiae template to generate a protected fingerprint template, recently non-invertible cancellable construct based on 3-5 pattern was proposed as safer alternative. This paper investigates recognition accuracy attainable by this new construct. It is found that five samples person's reference and allowing for maximum three authentication attempts provides genuine user with best chance being successfully authenticated. An evaluation FAR FRR in scenario demonstrates can be tuned suit performance security requirements different applications adjusting size matching thresholds. The then modified improve its ability discriminate between users impostors. Compared other protection schemes, favourable.