Survey of Anomaly Detection Methods
This survey defines the problem of anomaly detection and provides an overview of existing methods. The methods are categorized into two general classes: generative and discriminative. A generative approach involves building a model that represents the joint distribution of the input features and the output labels of system behavior (e.g., normal or anomalous) then applies the model to formulate a decision rule for detecting anomalies. On the other hand, a discriminative approach aims directly to find the decision rule, with the smallest error rate, that distinguishes between normal and anomalous behavior. For each approach, we will give an overview of popular techniques and provide references to state-of-the-art applications.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 900157
- Report Number(s):
- UCRL-TR-225264; TRN: US200709%%559
- Country of Publication:
- United States
- Language:
- English
Similar Records
Process Anomaly Detection for Sparsely Labeled Events in Nuclear Power Plants
General-purpose Unsupervised Cyber Anomaly Detection via Non-negative Tensor Factorization