Verification system based on long-range iris and Graph Siamese Neural Networks

The main advantage of using biometric information over traditional methods is that instead of requiring information that the user should know or possess (password, codes, PIN, etc.), they use characteristics that univocally and biologically define the users (fingerprints, iris, face, etc.). In particular, these characteristics are universal (all users can be measured), singular (each user has its own measures), permanent in time and context, and can be quantitatively measured [33]. Soft biometrics can be divided into two groups: physical and behavioural biometrics. Techniques of the first category use physical characteristics like face, iris, and fingerprint for their tasks [4], whereas techniques of the second one, use information extracted from user behaviours such as signature, voice, and keyboard typing [39]. Among the physical biometrics, face [15] and fingerprint [2] methodology have been the most explored, and have already been used in many real-world applications such as airport scanners, banking, military access control, smartphones or forensics [7, 36]. However, in the last decade, the use of iris has begun to attract interest in applications such as gender classification [27], iris liveness detection [8], border control [45] and citizen confirmation [22]. In fact, iris biometric represents a secure biometric with low forgery and error rates due to its highly certain features [43]. Furthermore, this biometric information is usually combined with Artificial Intelligence (AI) and Machine Learning techniques (ML) in order to implement user identification and verification systems.