Application of Image and Data in Security

• Main Authors: Yan-Rui Lin, 林彥瑞
• Other Authors: Yung-Kuan Chan
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5396040%22.&searchmode=basic

碩士 === 國立中興大學 === 資訊管理學系所 === 107 === Because the rapid development of information technology, data usage is not restricted as it used to be. Through rational integration and application of data, it will become valuable information. In this paper, we applied artificial intelligence techniques to two different types of data, mobile applications and image, which are related to user safety. With the popularity of mobile devices, malware has gradually shifted from computer systems to mobile phone systems. Because the Android system has the highest market share, and due to its open platform, the number of users who have been victimized by suspicious mobile applications has increased year by year. In order to solve this problem, this paper proposes a static analysis system of AMSD, which use Permission Based and Code Based analysis methods to intercept android file features. Using Recognizability and F-statistics features selection methods to pick up important features, and the unknown Android applications are detected by machine learning. Compared with the anti-virus software of mobile phones on the market, AMSD system shows good detection effect. In construction sites, hardhats protect workers from falling objects. However, head injuries caused by not wearing hardhats are still frequent. In order to confirm whether people in the environment actually wear hardhats, this paper proposed a hardhat wearing detection system to strengthen the management and supervision of the site, hoping to reduce the occurrence of head injuries caused by not wearing hardhats. This system is based on YOLO-V3, will input images for analysis and detect whether the people in the images are wearing hardhats. The experimental results show that the accuracy can reach 90% on average.