About anomaly detection

InfoScale provides an AI-powered anomaly detection system that employs advanced machine learning algorithms to provide comprehensive threat identification and system monitoring capabilities. The system continuously monitors file system statistics and entropy patterns to establish baseline profiles, and then applies machine learning models to detect deviations from established trends. The system analyzes data points over time to detect established patterns of behavior and identifies outliers as anomalies.

The following file system attributes are monitored:

Any deviations in these attributes are considered as an anomaly and reported in the InfoScale Operations Manager UI as trend graphs and real-time alerts are generated.

Entropy is calculated by sampling a percentage of data for unusual patterns. By default 10% of the data is sampled for deviation in data patterns. Changes in entropy values indicate potential malicious activities. These changes are flagged as anomalies and displayed in the UI as entropy graphs.

AI-based anomaly detection is available with InfoScale version 9.0.2 or later.

At a high level, AI-based anomaly detection includes the following stages:

See Enabling anomaly detection.

See Training the ML engine.

See Viewing graphs and faults for anomalies.