The CUSUM anomaly detection algorithm was created to automatically search a vast database of network diagnostic test results from M-Lab for series of unusually high or low measurements. It uses a CUSUM control chart with parameters like expected mean and standard deviation to detect anomalies. The CAD algorithm was implemented in R using a sliding window technique to apply CUSUM charts to subsets of time series data and detect potential anomalies, which are then output along with a graph of the annotated time series. It was tested on M-Lab data showing known anomalies with low false positive and negative rates. Future work aims to optimize parameters and make the method more scalable.