Active Learning with Convergent Stop Rule to Collect User’s Behavior for Training Model Base on Non-intrusive Smartphone Authentication

• Main Authors: Pei-Wen Pan, 潘珮玟
• Other Authors: Deron Liang
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/3m663e

碩士 === 國立中央大學 === 資訊工程學系 === 105 === In order to protect the data within the smartphone, intrusive and non-intrusive user authentication mechanisms were developed. Traditional authentication mechanisms like number lock and pattern lock are intrusive user authentication mechanism. Non-intrusive user authentication mechanism doesn’t require any user interface, but collect user’s behavior in the background and authenticate it. Several non-intrusive authentication mechanisms were proposed, but them still have problem needed too much data for training. Actually user to provide the training samples can be very time-consuming. This study proposes a method to collect real-time detection with the use of active learning support vector machine choose training samples to identify the effect of the acceptable range in a small amount of training samples construction of non-invasive identification mechanism. For this study, we propose new stopping rule and model analysis with proposed active learning method. We analyze line of SVM validation accuracy, add a new stopping rule with convergence, and replace support vector with the closest sample points as standard for model analysis. Finally, this study presents an optimal method to collect real-time detection (active learning) compared with old version, results are total reduce half of the training time with a same recognition results of old version.