基于数据挖掘的异常交易检测方法
发布时间：2018-08-31 11:12
【摘要】：提出一种基于数据挖掘的异常交易检测方法,可以在业务层面和操作层面对交易中的异常进行检测。当一个用户提交一笔新的消费交易时,采用贝叶斯信念网络算法判断当前交易属于正常交易的后验概率,作为在业务层面的可信因子;然后提取该用户在当前交易之前的若干个操作,与当前交易一起构成一个固定长度的操作序列,并通过BLAST-SSAHA算法将其与该用户正常操作序列和已知异常操作序列进行比对,得出在操作层面的可信因子。综合考虑业务层面的可信因子和操作层面的可信因子,最终决定当前交易是否为异常交易。
[Abstract]:An anomaly detection method based on data mining is proposed, which can detect anomalies in transactions at the operational and business levels. When a user submits a new consumer transaction, the Bayesian belief network algorithm is used to judge the posteriori probability of the current transaction as a trust factor at the business level. Then, several operations of the user before the current transaction are extracted, together with the current transaction, a fixed length sequence of operations is formed, and the sequence of normal operations and known abnormal operations of the user is compared by the BLAST-SSAHA algorithm. A confidence factor at the operational level is obtained. Considering the trust factor of business level and the credibility factor of operation level, whether the current transaction is abnormal or not is finally decided.
【作者单位】： 中国银联股份有限公司电子支付研究院;复旦大学计算机科学技术学院网络与信息安全研究所;
【分类号】：TP393.08;TP311.13
本文编号：2214805
[Abstract]:An anomaly detection method based on data mining is proposed, which can detect anomalies in transactions at the operational and business levels. When a user submits a new consumer transaction, the Bayesian belief network algorithm is used to judge the posteriori probability of the current transaction as a trust factor at the business level. Then, several operations of the user before the current transaction are extracted, together with the current transaction, a fixed length sequence of operations is formed, and the sequence of normal operations and known abnormal operations of the user is compared by the BLAST-SSAHA algorithm. A confidence factor at the operational level is obtained. Considering the trust factor of business level and the credibility factor of operation level, whether the current transaction is abnormal or not is finally decided.
【作者单位】： 中国银联股份有限公司电子支付研究院;复旦大学计算机科学技术学院网络与信息安全研究所;
【分类号】：TP393.08;TP311.13
【相似文献】
相关会议论文 前1条
1 杨本良;;压缩候选的贝叶斯信念网络构造算法[A];广西计算机学会2005年学术年会论文集[C];2005年
相关硕士学位论文 前2条
1 黄丹梅;多分类器系统在蛋白质功能预测方面的应用[D];吉林大学;2010年
2 马达;基于贝叶斯的判别理论及其算法实现[D];中国地质大学（北京）;2011年
，本文编号：2214805
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2214805.html
