thatDot Novelty Detector is a context-aware anomaly detection engine built on the Quine Streaming Graph platform. It is designed to process high-cardinality categorical data—such as IP addresses, user IDs, and file paths—natively without lossy numerical encoding. Operating in an unsupervised manner, the system constructs a dynamic probabilistic graph model in real-time from streaming data. It learns behavioral fingerprints and profiles Indicators of Behavior (IoBs) to contextualize system and user actions, enabling it to distinguish between malicious anomalies and benign unique events like zero-day threats or fraud.
The system delivers millisecond-latency novelty scores and adapts automatically to new data without batch retraining cycles. For integration into enterprise data pipelines, it exposes a REST API. The output includes ranked results and explainable AI context detailing why an event was flagged. Alerts are delivered in a human-readable, subject-predicate-object format containing node objects, properties, and relationships, which can be visualized directly as a graph.
