以视觉相似为基础的Phishing检测方法

目前检测网络欺诈均采用人工的方式。该文提出了一种基于EMD(earth move distance)算法的视觉相似度欺诈网页检测算法,算法将网页在像素级别的相似度比较结果作为依据,判断某网页是否为欺诈网页。在2个实际系统中应用和实践了该算法,通过对9280个网站进行检测,检测出欺诈的正确率为90.90%,结果显示该算法具有较高的欺诈检测精度以及较低的时间性能开销。

Phishing detection method based on visual similarities

Current anti-phishing efforts rely on human intervention for phishing detection.A method based on the EMD(earth move distance) algorithm was developed to improve the efficiency of automached phishing detection.The visual similarity between two web pages on the pixel level was calculated by the EMD algorithm,and then to judge whether a web page is a phishing.The algorithm has been deployed on two systems.Analyses of 9 280 web pages accurately identified 90.90% of the phishing pages.Tests show that the EMD anti-phishing algorithm provides excellent detection precision and good real time performance.