ABSTRACT
In this paper, an initiative passive continuous authentication(CA) system based on soft biometricsis presented. Human facial features are used as hard biometricinformation for the authentication process is employed as the soft biometric information. Thepassive CA system keeps verifying, without interrupting the userfrom concentrating on his work. It also provides the capacityfor the machine to recognize who is in front of the terminal,reduces the potential security leaks, and denies access to theinvader with the stolen account and password. In this system, theface recognition core is implemented not only by the Eigenfacemethod, but also assisted by the interactive artificial bee colonyoptimization algorithm. The proposed method is implemented by bee colony algorithm. Face Recognition and Spoofing Detection System Adapted To Visually Impaired People Original
The conventional authentication system only requests the user to provide the authorized account and password to log into the system once they start to use a computer or a terminal. However, under this authentication framework, the machine can only recognize the user’s identity from the login information. It lacks the information to know who is using it. The common drawback of the one-time authentication system, which people use in the daily life, is that when the user leaves the seat for a short break, i.e., to get some documents or have a drink, anyone can sneak up to the computer and pretend to be the authorized user to access their data, or do anything under a fake identity. Later on, nobody will know who used the computer. This kind of security defect is not acceptable in some applications with the sensitive data, for example, the banking financial record or the customer personal information, the military industry, and the business confidentiality.
To avoid this disadvantage under the conventional authentication system, the user can only log off from the terminal or lock up the screen manually before leaving, and log in again when coming back to continue the work. This causes an inconvenience to the user, especially when the user is busily coming and going, and doing other things. Sometimes, the user may skip the log-off process just to keep away from the annoyance caused by repeating the log-off and re-login processes. Hence, the leak of the information security appears. Nevertheless, these situations will not happen in a passive continuous authentication (CA) system. Responding to the need of heightened security applications, different CA systems have been proposed in the past five years. The system presented in this paper is able to authenticate and memorize both the user’s hard and soft biometric information, e.g., face and clothes’ color, and continuously authenticating whether the person using the terminal is as the same valid user as the one login at the beginning. The principal strong points of the initiative passive CA system can be summarized in four ways.
In this paper, an effective CA system by combining the techniques in swarm intelligence, face recognition, image or video processing, and pattern recognition is presented. It aims at automatically overcoming the disadvantage mentioned above by the biometric features without interrupting the user from his work.
Swarm intelligence is a branch under evolutionary computing. Algorithms in swarm intelligence utilize puny intelligence from the creatures in Mother Nature to solve engineering problems. Artificial bee colony (ABC) optimization is a good example of the algorithms created by the inspiration from Mother Nature. It also leads to a series of newly developed algorithms in ABC and many useful applications on solving engineering problems. In the proposed CA system, interactive artificial bee colony (IABC) algorithm is employed to assist the face recognition module for raising the hard biometric recognition rate. The processes of IABC are executed offline to train a weighting mask for adjusting the value of the input feature. The goal for IABC to achieve is finding the proper mask to adjust the input features. The trained weighting mask is capable of extending the difference between different registered users and narrowing the difference between the registered images from the same user.