基于SVM的金融类钓鱼网页检测方法

针对金融服务领域面临的严峻信息安全挑战,以及现有钓鱼网页检测方法的不足,提出一种基于支持向量机(support vector machine,SVM)的金融类钓鱼网页检测方法.采用网页渲染去除常见的页面特征伪装,提取统一资源定位符(uniform resource locator,URL)信息特征、页面文本特征、页面表单特征以及页面logo图像特征,构建特征向量训练SVM分类器模型,实现对金融类钓鱼网页的识别.在特征提取过程中,利用适合中文的多模式匹配算法AC_SC(AC suitable for chinese)提高文本匹配效率,并采用加速鲁棒特征(speeded-up robust feature,SURF)算法实现logo图像的特征提取与匹配.多方法实验结果对比表明,该方法针对性更强,能达到99.1％的检测准确率、低于0.86％的误报率.

Method of detecting the financial phishing webpage based on SVM

Aiming at the serious information security challenges in the financial service field, and the shortcomings of existing phishing webpage detection methods, a financial phishing webpage detection method based on support vector machine (SVM) was proposed. The method uses webpage rendering to remove common feature camouflage of page, then sets up feature vector according to extract several features including uniform resource locator (URL), text messages, form of page, and logo image. Next, it trains the SVM classifier model by feature vector, and the method realizes the recognition of financial phishing webpage. In the process of features extraction, the method uses multiple pattern matching algorithm AC_SC (AC suitable for chinese) to improve efficiency of text matching, and finishes logo image features extraction and matching by speeded-up robust feature (SURF) algorithm. It shows that the proposed method reveals better in pertinence according to experiment of several methods, and it can achieve 99.1% detection precision and not higher than 0.86% false positive rate.