In this blog, we will take a deep dive into anomaly detection using two ready-to-use Python plugins for real-world IoT use cases. We will be leveraging the Python Processing Engine within InfluxDB 3 Core or Enterprise.

InfluxDB is a time-series database platform for collecting, storing, and analyzing time-stamped data, offering tools for monitoring, observability, and IoT applications. From this resource users can learn about time-series data modeling, querying, and visualization to monitor and analyze their systems' performance and behavior in real-time.

InfluxData

Real-time anomaly detection can be performed directly within InfluxDB 3 using Python plugins and the Processing Engine. Two complementary approaches are available: the MAD (Median Absolute Deviation) plugin for detecting immediate spikes and outliers, and the ADTK (Anomaly Detection Toolkit) plugin for identifying sustained instability and variance shifts. Both plugins run in-database without separate infrastructure, support configurable thresholds and windows, and integrate with notification systems like Slack. The MAD approach uses statistical methods for instant detection on every write, while ADTK employs scheduled checks to analyze historical windows for pattern changes. Together, they provide comprehensive monitoring for IoT sensors and industrial systems.

Performing Real-Time Anomaly Detection with InfluxDB 3: An In-Depth Guide

Understanding the anomaly detection landscape

How to use MAD and ADTK plugins in InfluxDB 3

ADTK plugin for detecting sustained instability