Abnormal data detection method and apparatus

Abstract

Embodiments of the invention provide an abnormal data detection method and apparatus. The method comprises: converting a graph used for describing to-be-detected data into adjacent matrixes A, wherein L is used to represent adjacent matrixes without sub-graphs of abnormal data in the adjacent matrixes A, and R is used to represent adjacent matrixes with the sub-graphs of the abnormal data in the adjacent matrixes A; detecting whether the to-be-detected data contain noise data or not; establishing different convex optimization expressions according to different detection results, wherein in the established convex optimization expressions, the low-rank property of L is ensured by utilizing a nuclear norm of L, the sparsity of R is ensured by utilizing a first-order norm of R, the nuclear norm of L refers to the sum of singular values of L, and the first-order norm of R refers to the sum of absolute values of all elements in R; and solving the established convex optimization expressions to obtain L and R. According to the embodiments of the invention, the complexity of detection can be lowered and the accuracy of detection can be improved.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of electronic data processing, and in particular, to a method and device for detecting abnormal data. Background technique [0002] The goal of anomaly detection is to find objects that are different from most other objects. Applications of anomaly detection are, for example: fraud detection, where a person who steals a credit card may have different purchasing behavior than a credit card holder, where a credit card company tries to find the purchase patterns of a thief by looking for it, or Detect thieves by noticing changes in behavior that differ from common ones. Similar methods can also be used for other types of fraud detection; intrusion detection, common attacks on computer systems and network systems, some attacks are obvious, such as those aimed at Attacks that disable or take control of computers and networks, but other attacks, such as those aimed at covertly gathering infor...

AI Technical Summary

Problems solved by technology

[0009] However, the constraints of the non-negative matrix make it more difficult to solve the problem, and it is more difficult to realize, and the problem solved by this method is a non-convex optimization problem, and only a partial solution can be obtained. The final result is consistent with the assumption before the solution. Strong correlation, different assumptions can get different solutions, and the accuracy of identifying abnormal data is poor

Images