碩士 / 南臺科技大學 / 資訊管理系 / 107 / The Internet is constantly improvement and innovation, has already become an important tool in the life. According to the statistics of the Internet Data Research Institute, everyone average spends over six hours surfing the Internet every day, which means there is one quarter of the time on Internet. The whole world shows the trend of networking gradually, but also bring network security issues. Phishing has a major impact in the social. According to the statistics of Anti-Phishing Working Group (APWG), the global was attack from phishing over 255,065 times. The users to be affected when they browsing the web, because they are afraid to fall in phishing. Therefore, the study based on phishing feature vectors to find and determine which most likely to be phishing feature vectors.
In the study, we search for phishing feature vectors from “UCI Machine Learning Repository” as research object. Through the machine learning with classification algorithm to the research analysis, they are: Decision Tree, Random Forest, Neural Net, BPNN and SVM, using RapidMiner to predict the construction of the model, figure out every feature vectors’ weight and accuracy. The results of the study found that “Adding Prefix or Suffix Separated by (-) to the Domain.”, “Sub Domain and Multi Sub Domains.”, “HTTPS”, “URL of Anchor.” and “Number of Links Pointing to Page.” are the most recognizable. “HTTPS” has the largest weight and identification. Therefore, the study will teach the rules of five feature vectors, giving the users as a reference for phishing and early warning.
Identifer | oai:union.ndltd.org:TW/107STUT0396003 |
Date | January 2019 |
Creators | HUNG, MU-LAN, 洪慕藍 |
Contributors | JENG, YU-LIN, 鄭鈺霖 |
Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
Language | zh-TW |
Detected Language | English |
Type | 學位論文 ; thesis |
Format | 83 |
Page generated in 0.0021 seconds