Edge computing has emerged as a promising tool to resolve the issues like high bandwidth and privacy requirements for Internet of Things (IOT) devices that were not resolved by traditional cloud computing. The data generated at edge nodes are often private and sensitive. Edge computing emphasizes doing computation nearer to the data source, which also helps transfer lesser data over the cloud to minimize the traffic. It is very important that proper security measures are taken to protect the edge nodes and the data from any security breach. This paper first discusses various aspects of cloud computing, its threats and limitations, followed by a discussion of edge computing, its architecture, threats, and security challenges. The paper further reviews the literature on anomaly detection, i.e., one of the security challenges in edge computing for the year 2021. The extensive review results in developing a taxonomy of methods used to do anomaly detection in edge computing. The review later discussed how researchers have tried to implement their model that does anomaly detection, keeping in mind the limitation of edge nodes such as computation ability, storage, and high bandwidth requirement. The review also summarizes a few usecases in which researchers have used anomaly detection. The review also points out that deep learning methods have been favored by the majority of the researchers and suggest future research direction. The taxonomy presented in this paper can be served as a summarized framework to anchor research work in the area of anomaly detection. Practitioners can also use the taxonomy as a ready-to-go reference tool for the researchers could utilize this working in the area of anomaly detection. [ABSTRACT FROM AUTHOR]