基于项目流行度和新颖度分类特征的托攻击检测算法

发布时间：2018-04-04 16:01

本文选题：托攻击　切入点：项目流行度　出处：《工程科学与技术》2017年01期

【摘要】：针对有监督检测方法在检测托攻击时准确率不高的问题,提出一种基于项目流行度和新颖度分类特征的托攻击检测算法。首先,根据真实概貌和攻击概貌在选择评分项目方式上不同,从流行度和新颖度角度,提出有效区分正常用户和攻击用户的特征;然后,基于这些特征提出一种集成检测框架,通过Boosting提升技术产生多个差异较大的基分类器,并且通过融合带有权重的基分类器的预测值得到最终的检测结果。实验结果表明,基于项目流行度和新颖度分类特征的托攻击检测算法能够提高攻击检测的准确率和召回率。
[Abstract]:In order to solve the problem that the accuracy of supervised detection method is not high when detecting support attacks, an algorithm based on item popularity and novelty classification features is proposed.First of all, according to the differences in the selection of scoring items between the real profile and the attack profile, the features of distinguishing normal users from attacking users are proposed from the perspective of popularity and novelty, and then an integrated detection framework based on these features is proposed.The Boosting lifting technique is used to generate many different base classifiers, and the final detection results are obtained by merging the prediction values of the basis classifiers with weights.The experimental results show that the algorithm based on item popularity and novelty can improve the accuracy and recall of attack detection.
【作者单位】：燕山大学信息科学与工程学院;河北省计算机虚拟技术与系统集成重点实验室;
【基金】：国家自然科学基金资助项目(61379116) 河北省自然科学基金资助项目(F2015203046) 河北省高等学校科学技术研究重点资助项目(ZH2012028)
【分类号】：TP309

【参考文献】