Security software using neural networks

摘要

This paper describes the structure and the theoretical principles of the security software that the authors have developed. This software is an industrial application based on neural networks theory. Its aim is to recognize somebody's face and thus to add one more protection level to Windows NT and Windows 95 security access. They implemented the face learning phase by using projection onto an eigenvectors matrix and the backpropagation algorithm. They stored, in a database, the identification of the faces which have been learned and added a security protection when opening the personal session of the operating system Windows NT and created a new level of protection for Windows 95. They tested their algorithms on images of other types than faces and the results allow the use of the software in industrial control.