MULTIFACTOR AUTHENTICATION SYSTEM BASED ON NEURAL NETWORKS

Abstract

The biometric approach is considered one of the most relevant in identification and authentication systems. The bi ometric method is based on the analysis of unique human characteristics. Face recognition is an important task, be cause it is the first stage of identification, to find out who owns the face and whether it is in the database, you must first locate it. To solve this problem, different approaches are used among them: empirical methods, method based on learn ing, method based on comparison with a template, method based on contour models. When recognizing a face, the sys tem that solves this problem must take into account a number of factors: differences in the faces of different people, changing the angle of the face, the possibility of certain features, changing facial expressions, the presence of obstacles in the image that may partially obscure the subject. Artificial intelligence is both a field for development and a chal lenge. Due to the fact that the development of machine learning and artificial intelligence is often focused on processing large data sets, and machine learning algorithms directly depend on the quality of the information it processes, inter ference and misinformation can disable further operation of the algorithm, which can lead to incorrect conclusions, the correctness of which will be difficult to verify because of the large data sets. The choice of method for solving the prob lem of face detection depends on the specific problem and the conditions in which the algorithm should operate.In this article the possibilities of neural networks for application in the system of multifactor authentication are considered and analyzed. Options for possible implementations using an artificial network, prospects for the development of these networks and the importance in our time are considered. Modern research in this field among the leading countries of the world is analyzed. One of the methods for application is the EIGENFACE face recognition algorithm. Prospects for the use of neural networks, artificial intelligence, review of the features of learning artificial neural network and algo rithm EIGENFACE for use in multifactor authentication and proposed steps to improve this algorithm based on fuzzy set theory. The paper clarified what a neural network, an artificial neuron, the operation of the Eigenface recognition algorithm is, because knowledge of the principle of the algorithm greatly facilitates its application in practice, the learning process is considered for further possible implementation. Additional stages of algorithm improvement with the help of fuzzy set theory are offered, which becomes a powerful tool for building intelligent hardware and software pattern recognition systems. The introduction of a fuzzy filter into the algorithm calculates the fuzzy increment so that the images become less sensitive to local changes in structures, boundaries of objects. The filter will provide a high degree of distinction between noise and structural objects of the image. Segmentation allows you to split images into smaller parts, which greatly improves system recognition.