Artificial intelligence face recognition for authentication

Abstract

Urgency of the research. Technical progress leads to a tremendous increase in number of cybercrimes. Almost every person around the globe has a range of digital accounts containing sensitive private information which is in fact protected by simple password. Therefore, security systems have a great and important role to guard privacy. It is necessary to have a solid systemwhich can distinguish between people and act differently based on their permissions.In difference between face recognition authentication and other identification solutions such as passwords, email verification or fingerprint identification - biometric facial recognition uses unique mathematical and dynamic patterns that make such system one of the safest and most effective. Target setting. Face recognition authentication is about to be one of the most stable. There is a range of methods that are available for detecting and processing faces using different levels of complexities. Summing up - face recognition for authentication purposes can emphasizes security. Convolutional neural networks (CNN) outperform any possible humans’ recognition rate. However, such systems should be continuously manually improved. Another problem with such systems is that they require accurate data to be trained before they are actually being deployed. It is essential for such system to be fast enough to recognize people and that the training should be accomplished without much difficulty and also be fast. Actual scientific research and issues analysis. Face recognition algorithms have been reviewed in a range of scientific papers such as Haar Cascades, Kalman Filter and applied in various spheres. Among research papers, there is a range of security systems that use face recognition technology. Facial recognition approach for security access and authentication presented by Jeffrey S. Coffin uses custom VLSI Hardware and Eigenspaces method. Systems provided by Shankar Kartik uses Eigenfaces method for face identification as well which in fact gives weak results with moderate accuracy. Uninvestigated parts of general matters defining. The swiftness of the particular face recognition systems heavily depends on the changes in conditions of light, expression, camera density, and on partial blocking of the face. Several scientific works have already proposed range of approaches for face recognition under unpleasant conditions, but not much of them actually work. The research objective. This article aims to describe Face Recognition authentication system experimental architecture inside informational system accessible via web interface. The Face Recognition authentication system consists of a camera node, a cloud server and input-output device for interacting with users by means of web interface. With the advancement in web and cloud, this article represents development of the authentication system based on Face Recognition System. Using Google next-generation system, TensorFlow with a deep learning framework on board. TensorFlow is flexible, portable and open source project. The statement of basic material. As it is known - the human brains vision seems to be very easy functioning. It does not take any difficulty to tell apart a dog and a cat, read a word or recognize a human face. But in difference from human - these tasks are really difficult problems for solving with a computer. Recognition process only seems easy because human brain is really good at perception and as a result in understanding images. During last years, machine learning has made great progress in solving these difficult problems. In particular, model called - deep convolutional neural network can result in reasonable performance on solving difficult visual recognition tasks which are matching or exceeding human performance in some aspects. Conclusions. This paper introduced a new method of obtaining data for training security systems from social media and human interaction for future use in authentication process in various informational systems. There are several advantages of proposed system which can be described. First of all, one should mention that using of TensorFlow can be adaptive, powerful, and flexible. Moreover, training time is acceptable in comparison with other frameworks and much more faster if one uses distributed TensorFlow.