Iris recognition technologies are used in many applications nowadays because of the always-growing demand for identity authentication. This paper presents a strong deep learning-based system for exact iris localisation and recognition. The proposed framework consists of three steps: region segmentation, feature extraction, and recognition. The iris images were first obtained from three benchmark datasets. The multimedia university (MMU)-iris dataset, the IITD-iris dataset, and the UB-iris dataset. The interesting iris areas are then split using DIDO method in the second dimension. Thirdly, three techniques - the Harris detector, ResNet-18, and speeded up robust features (SURF) - are aggregated to extract features from the segmented iris sections. Feature extraction helps to emphasise the discriminative characteristics of Iris images clearly, therefore enabling classification models to differentiate between different patterns. These resulting discriminative features are subsequently put into the upgraded GRU model to detect matching and non-matching iris patterns. The empirical analysis revealed using the IITD-iris, MMU-iris, and UB-iris datasets that the enhanced GRU model acquired maximum recognition accuracy of 99.67%, 99.43%, and 98.78%. These achieved results surpass those of comparative models, including GRU, RNN, LSTM.

Security and user privacy are areas of concern in digital transactions over a network. Integration of biometric pattern matching into an identity management system (IMS) enhances security of transactions and improves ease of use. Privacy of users in a biometric based system is improved by using keys generated directly from feature sets instead of conventional stored templates. This paper proposes a framework for integrating biometric key based authentication into an IMS. The generated keys need to be consistent. Consistency of keys is attained if the same key is generated for every instance of biometric input. Generation of long and consistent keys directly from feature sets poses a challenge due to intra and inter user variations inherent to biometric data. A novel methodology for generating consistent keys using principles of vector quantization and cluster based learning is introduced. 15-bit keys have been extracted from handwritten signature datasets. The results are promising and can be extended to multi-modal biometric feature sets.

A system is proposed that considers minimal features using subpattern analysis which leads to less response time in a real time scenario. Using training samples, with a high degree of certainty, the minimum variance quadtree components [MVQC] of a signature for a person are listed to be applied on a testing sample. Initially the experiment was conducted on wavelet decomposed information for a signature. The non-MVQCs and core components were analyzed. To characterize the local details Gaussian-Hermite moment was applied. Later Hu moments were applied on the selected subsections. The summation values of the subsections are provided as feature to radial basis function [RBF] and feed forward neural network classifiers. Results indicate that the RBF classifier yielded 7% false rejection rate and feed forward neural network classification technique produced 9% false rejection rate. Promising results were achieved, by experimenting on the list of most prominent minimum variance components which are core components using RBF.

This paper proposes an online signature authentication based on polynomial modeling. The derived feature acceleration is used. The order of the polynomial is carefully selected for each subject using the acceleration vector. The dynamic warping algorithm and mean squared error measure aid in standard template selection process for a subject. The best fit order is directly dependent on the cursive information of the signature sample. We provide a practical analysis of residual term, number of zeros inside the unit circle and distance measure on pole values for the purpose of classification. Storing order and standard template for a subject, reduces the runtime and memory requirements for a dynamic system. The best performance was achieved based on the number of zeros inside unit circle with false rejection rate as 5% and false acceptance rate as 7%. The aim of the work is to provide simple robust authentication suitable for hand held devices. Even compared to state-of art experiments, the proposed system is aimed to provide authentication using a single derived feature.

Authentication systems which are covenant with a measurable behavioral and physiological traits are essential for an online system. In this paper, two types of biometric sample authentication from different databases on a common algorithm using Continuous Dynamic Programming [CDP] are discussed. Using a common algorithm, a method for user-dependent threshold decisions can be achieved for both biometrics in a uniform fashion. The integration of static iris information and dynamic signature information are done at decision level. Inferences are drawn using voting techniques. The derived kinematic feature, acceleration, is used in this paper.