STARS: a Self Teaching Alarm Response System

The STARS component of SentryPort Sentinel™ is a self adaptive or self-teaching IDSS which works with sequences of activations. It accepts the full stream of activations from other parts of the SentryPort Sentinel™ system and maintains a filtered list of activations. This filtered list comprises:

Those activations which fall outside normal patterns that have occurred in the past, and
Optional summary descriptions of common event sequences, as supplied by a human operator at some previous stage.

STARS works in unison with IRIS, and continually provides it with additional cases corresponding to the knowledge STARS acquires.

STARS is unique in that it

Learns without human intervention,
Is extremely efficient at detecting normal patterns in the activation stream,
Can learn at the same time that it processes the activation stream, and
Progressively "forgets" about normal patterns that are no longer relevant. However, it rapidly reinstates such "forgotten" patterns if they resume occurring on a regular basis in the future.

STARS is based on the proprietary Sequence Graph algorithm.