引用本文:易树平,李嘉佳,易茜.基于行为流图的可信交互检测方法[J].控制与决策,2020,35(11):2715-2722
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览次   下载 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于行为流图的可信交互检测方法
易树平,李嘉佳,易茜
(重庆大学机械工程学院,重庆400044)
摘要:
为保障人-网站交互的可靠性和可信性,以探寻交互行为模式的独特性为出发点,采用行为流图描述用户与网站的交互活动,通过分析可信交互行为模式提取与用户生理及心理特性相关的交互行为特征,提出一种以可信行为特征作为度量的可信交互检测方法,并基于某网站真实日志数据验证所提可信行为特征的功效.将用户一次会话作为记录单元,描绘出用户与交互环境、工具、会话行为和所在页面4个维度相结合的行为流图;然后,依据数据分析,提取可信行为特征参数并使用SMOTE算法平衡数据集;最后,利用决策树和随机森林算法完成用于检测交互可信性的模型训练与测试.通过实验对实际数据进行检测,所提出方法在决策树模型中对用户不可信行为的错误接受率为0.44%,随机森林算法中则低至0.31%.研究结果表明,可信行为特征的组合具有用户可辨别性和独特性,证明了人-网站交互行为模式具有个体特性,与他人存在差异性,可用于检测交互行为发起者与账户真实所有者间身份的一致性.
关键词:  人-网站交互  可信交互  行为流图  可信行为  机器学习  决策树  随机森林
DOI:10.13195/j.kzyjc.2018.1618
分类号:TP181
基金项目:国家自然科学基金项目(71671020);重庆市技术创新与应用发展专项重点项目(cstc2019jscx-mbdxX 0049).
Trustworthy interaction detection method based on user behavior flow diagram
YI Shu-ping,LI Jia-jia,YI Qian
(College of Mechanical Engineering,Chongqing University,Chongqing 400044,China)
Abstract:
A trustworthy interaction detection method based on user behavior flows is proposed to ensure the dependability and trustworthiness of human-web interaction. Firstly, the behavior flow diagram is used to capture the all relevant factors of user behavior from web log, in which the uniqueness is taken as the starting point of this. The behavior units are recorded as “one session”. The behavior flow diagram describes the interactions in four dimensions, namely interactive environment, interactive tool, session behavior, and the current page. Then, the behavior features related to individual psychology and physiology are extracted as trustworthiness measures on the basis of data analysis. After balancing the data set through synthetic minority over-sampling technique(SMOTE), the training and testing trustworthy interaction detection model are completed by the aid of the decision tree and random forest algorithm. Finally, an instance is given to illustrate that the false accept rate(FAR) of the untrustworthy behavior of the proposed method in decision tree model is 0.44%, while it is as low as 0.31% in random forest. The results indicate that the combination of trustworthy behavior features has differentiation and uniqueness among users, which proves that the behavior patterns of human-web interaction have personality and distinguishable otherness with someone else. It can be used to detect identity consistency between the dominator of interactive behavior and the real owner of an account.
Key words:  human-web interaction  trustworthy interaction  behavior flow diagram  trustworthy behavior  machine learning  decision tree  random forest

用微信扫一扫

用微信扫一扫