Use Context To Reduce False Positives
thatDot Anomaly Detector dramatically reduces false positives by using contextual information—learned automatically from the data.
Typical anomaly detection is a very mathematical process, requiring exclusively numeric data-which can provide anomaly scores for each item in a batch, but only when processed together. thatDot Anomaly Detector uses the structure created from all observations to determine whether the observation is anomalous given the relevant context, in real-time.
The image shows the scores of real-world data streamed into thatDot Anomaly Detector. The color scale on the right shows how unique each observation is, the yellow representing entirely unique observations that have never been seen before. Instead of just labeling all unseen observations as anomalous, thatDot Anomaly Detector uses a rich graphical model of the data to integrate context into the anomaly score. Yellow dots with low scores (near the bottom of the plot), are examples of unique data that are not anomalous. “New” is not always “novel”. Only thatDot Anomaly Detector can automatically use context to learn the difference.
Sample Use Cases
Monitor SaaS and IaaS usage for unusual configuration changes or resource access patterns.
Identify network route inefficiencies by discovering anomalous traffic behavior.
Analyze transactions for suspicious activity, or to recommend common alternative products—all in real-time.
Log Volume Reduction
Actively filter logs and events for novelty to reduce log storage costs and log analysis licensing.
Need To Transform And Join Data First?
Many of the most valuable observations are the result of transforming, combining and interpreting data from many different sources. thatDot Connect streamlines the process of real-time data cleaning and transformation, including the multiplexing of log, event and metrics data into any desired data format. And, with all thatDot solutions, you can use non-numeric data in it’s native format, no conversions to numeric values needed.