【摘要】 恶意URL现如今对网络安全影响巨大,能否高效的检测恶意URL成为一个亟待解决的问题。针对传统基于文本特征的检测方法没有考虑到URL中词的位置和上下文信息的缺点,提出了一种基于上下文信息的恶意URL检测方法,首先利用预处理方法解决了URL中存在大量的随机字符组成单词的问题,使用特殊符号作为分隔符对URL分词,对得到的分词结果使用Word2vec生成词向量空间,然后训练卷积神经网络提取文本特征并分类。实验结果表明,该方法在大量真实数据上能够达到97.30%的准确率、90.15%的召回率和92.33%的F1值。更多还原

【Abstract】 Malicious URLs have a great impact on network security nowadays. Whether malicious URLs can be detected efficiently has become an urgent problem to be solved. Aiming at the disadvantage of traditional text feature-based detection methods that do not take into account the location and context information in URLs, the author propose a malicious URL detection method based on context information. Firstly, a large number words that consist of random characters in URLs are solved by preprocessing method. Special symbols are used as separators to segment URLs. Word2 vec is used to generate word vector space, and then Convolutional neural network is trained to extract text features and classify them. The experimental results show that the method can achieve the accuracy of97.30%, the recall rate of 90.15% and the F1 Score of 92.33% using a big real data set.更多还原