基于Navie Bayes算法与k-means聚类算法的财务数据库异常检测

Anomaly Detection of Financial Database Based on Navie Bayes Algorithm and k-means Clustering Algorithm

为了有效消除用户异常行为对企业财务数据库所带来的安全隐患,以往的数据库异常检测技术(如Navie Bayes算法)通常采用查询反馈,并建立用户行为特征(用户行为轮廓)的方法查找安全隐患,而该方法构建训练集耗时较多,效果不显著。因此,提出一种基于Navie Bayes算法与k-means聚类算法相结合的财务数据库异常检测技术。通过调取财务数据库运行日志中的用户查询内容及相应结果,采用k-means聚类算法进行用户分组,采用Navie Bayes算法构建异常检测模型。应用测试结果表明,与传统的用户行为轮廓算法相比,所提算法的训练效率更高,准确率大幅提高,综合F_(1)值有所提升,提高了财务数据的安全性。

In order to effectively eliminate the security risks caused by user abnormal behavior to enterprise financial database,previous database anomaly detection technologies(such as Navie Bayes algorithm)usually use query feedback and establish user behavior characteristics(user behavior contour)to find security risks.However,this method takes more time to build training set and has no significant effect.Therefore,a financial database anomaly detection technology based on the combination of Navie Bayes algorithm and k-means clustering algorithm is proposed.By retrieving the user query contents and corresponding results in the operation log of financial database,k-means clustering algorithm is used to group users,and Navie Bayes algorithm is used to construct anomaly detection model.The test results show that compared with the traditional user behavior contour algorithm,the proposed algorithm has high training efficiency,greatly improved accuracy,improved comprehensive F_(1) value,and greatly improved the security of financial data.