그래프 기반의 피싱 URL 탐지

초록

Cyber crime through phishing is on the rise. To prevent the damage caused by phishing attacks, many studies to detect phishing URLs such as content-based and string-based have been conducted. A content-based method is a method of downloading and analyzing web page content, which poses a security risk. The string-based method analyzes URL string patterns and uses them for phishing URL detection. In this paper, we construct a graph with URL strings based on the phishing URL trends identified from previous studies. And, we detect phishing URLs through the graph-based inference algorithms such as Random Walk with Restart, Belief Propagation, and graph embedding methods such as DeepWalk, Node2vec. As a result of comparing our graph-based phishing URL detection method with the conventional phishing URL detection methods, our method shows better prediction performance in all accuracy measures.

키워드