TWC: Small: Unsupervised and Statistical Natural Language Processing Techniques for Automatic Phishing and Opinion Spam Detection

在网络钓鱼中，攻击者试图窃取敏感信息，例如，银行/信用卡账号、登录信息等，来自互联网用户。美国社会和经济对互联网和网络的依赖程度越来越高，而网络钓鱼问题也越来越严重。一种流行的网络钓鱼方法是创建一个模仿好网站的网站，然后通过电子邮件吸引用户，这是迄今为止最流行的媒介，以吸引不知情的用户到网络钓鱼网站。由于这种做法和网络钓鱼造成的损害，设计有效的分类器，电子邮件和网站是很重要的。 在这个项目中，新技术，灵感来自自然语言处理方法，正在设计用于网络钓鱼电子邮件和网站检测。然后，它们在现实的数据集上得到严格的实施和验证。它们也被应用于自动检测意见垃圾邮件。拟议的研究预计：（i）在推动自然语言处理技术的信封是有用的，（ii）产生这些技术在网络安全的新应用。在过去，PI在让妇女和少数群体，包括代表性不足的少数群体参与其研究方面非常成功，这一努力将在本项目中继续下去。休斯顿大学已被公认为西班牙裔服务机构，PI将继续他过去的成功努力，让代表性不足的少数民族，包括非洲裔美国人和西班牙裔美国人参与这项研究。这项研究将进入课堂，并通过出版物和网上软件广泛传播。

In phishing, an attacker tries to steal sensitive information, e.g., bank/credit card account numbers, login information, etc., from Internet users. The US society and economy are increasingly dependent on the Internet and the web, which is plagued by phishing. One popular phishing method is to create a site that mimics a good site and then attract users to it via email, which is by far the most popular medium to entice unsuspecting users to the phishing site. Because of this modus operandi and the damages caused by phishing, it is important to design efficient and effective classifiers for emails and web sites. In this project, new techniques, inspired from natural language processing methods, are being designed for phishing email and web site detection. They are then implemented and validated rigorously on realistic data sets. They are also applied to automatic detection of opinion spam. Proposed research is expected to: (i) be useful in pushing the envelope of natural language processing techniques, and (ii) yield new applications of these techniques in cyber security. In the past, the PI has been very successful in involving women and minorities including underrepresented minorities in his research and this effort will be continued in this project. University of Houston has been recognized as a Hispanic-serving institution and the PI will continue his past successful efforts to involve underrepresented minorities including African Americans and Hispanics in this research. This research will be moved into the classroom and broadly disseminated through publications and software on the web.