Abstract: Anomaly/outlier detection is an important area of machine learning which finds its application in intrusion-detection, fraud-detection, etc. In recent times, the focus of data analytics has shifted to big data analytics, i.e., analytics on large-scale data and fast-moving data streams. The traditional data processing tools and algorithms are not able to handle big data, so, there is a need of algorithms to be implemented in a parallel model like MapReduce to solve this problem. In this study, the researchers implement frequency-based algorithm on Spark MapReduce as a scalable and accurate solution for anomaly detection on large-scale as well as streaming datasets.