부산대학교 도서관

Mobile devices are vulnerable to different security and privacy threats and attacks from attackers. In order to protect user information from attacks by opponents, it is increasingly important to perform continuous identity authentication on intelligent users' devices. Different from other static authentication technologies, continuous authentication technology improves the security and reliability of the system through regular authentication of intelligent devices. By fusing the sensor and the front camera, we propose an efficient multibiometric-based continuous authentication scheme. On the one hand, use the built-in sensors of smart phones such as accelerometers, gyroscopes and magnetometers to continuously monitor the user's behavior patterns. On the other hand, the smartphone's front-facing camera is used to capture faces to continue continuous authentication when the sensor data does not change. We extract features from the collected data, and then use two-class SVM to train the classifier to complete user authentication. The experimental results show that the proposed method can achieve the authentication accuracy of 98.3%.