2014 Symposium Posters

Project Members
Gilchan Park, Lauren Stuart, Julia M. Taylor, Victor Raskin

The purpose of the first research is to report on an experiment into text-based phishing detection. The developed algorithm uses previously published work on the, so-called PhishNet-NLP, a content based phishing detection system. In particular, this research aims to analyze the keywords that lead used to do some actions in email texts. The algorithm produced the considerable results in filtering out malicious emails (TPR); however, the rate of text falsely identified as phishing (FPR) needed to be addressed. To solve the FPR problem, tradeoff between TPR and FPR was performed to reduce the FPR while minimizing the decrease in the phishing detection accuracy. The second research’s aim is to compare the results of computer and human ability to detect phishing attempts. Two series of experiments were conducted, one for machine and the other one for humans, using the same dataset, and both were asked to categorize the emails into phishing or legitimate. The results prove that machine and human subjects differ in classification of phishing emails. This comparison suggests that humans intelligence to detect some types of phishing emails that machine could not recognize needs to be semantically computerized so as to ameliorate the machine’s phishing detection ability.