UnAvoids

UNAVOIDS presents a novel approach for identifying and visualizing anomalies in security data by integrating advanced detection algorithms with exploratory visualization techniques. Central to its innovation is the use of a two-dimensional Neighborhood Cumulative Distribution Function (NCDF) space, a unique transformation that effectively captures and distinguishes observations based on their neighborhood characteristics, rather than merely projecting data onto a plane. This method, inspired by the cumulative distribution function of a random variable, allows for the classification of data points as either normal or anomalous based on their local surroundings. In this space, each data point is depicted as an individual curve, making it straightforward to distinguish anomalies from regular patterns. Crucially, the NCDF space retains the full integrity of the original data, making it a highly precise tool for anomaly detection. As an entirely unsupervised model that assigns a normalized score to each observation to measure its level of anomaly, UNAVOIDS has shown exceptional accuracy in detecting anomalies in both simulated and real-world datasets, surpassing other existing methods.