一种多特征分类识别算法融合的网络钓鱼识别技术

针对页面特征提取实时性差的问题进行了研究，提出将特征分类，并行提取、检测、再融合结果的方法。首先提取三个类别的主要特征，包括文本、视觉和网络链接；然后，分别利用了贝叶斯算法、EMD算法以及网络爬虫来进行分类；并且基于后验概率来确定权值的最终选取。最后，把这三个分类结果进行融合。通过对贝叶斯、加权和加权贝叶斯的比较，从正确率、漏报率和误报率对算法进行评估，实验表明采用加权贝叶斯的方法来进行融合计算效果最佳，能够提供较高的准确率和较低的误报和漏报，提高检测的精度和实时性。

A phishing recognition technology based on the fusion of multiple features classification and recognition algorithm

In view of the problem of poor real-time performance of page feature extraction is studied, the method of feature classification, parallel extraction, detection and re fusion results are proposed. First extract the main features of the three categories, including text, visual and Internet connection; Then, using the Bayesian algorithm, EMD algorithm and web crawler to classify; And to determine the weight of the final selection based on the posterior probability. Finally, the fusion of these three classification results. Experiments show that a phishing recognition based on weighted Bayesian algorithm has better performance, through the comparison of Bias , Weighted and Weighted Bias, which evaluates the algorithm according to the correct rate, false negative rate and false alarm rate. The accuracy rate can provide higher to improve the accuracy of detection; while ensuring low false positives and false negatives to improve the real-time of detection.