Omniculus — Vision

Domain: omniculus.com Tagline (working): See the whole picture. Decide with a human.

What Omniculus is

Omniculus is a defensive security analytics research platform. It ingests security telemetry, fuses it into a knowledge graph, detects dangerous activity, and supports a human analyst's decisions. The name reads as omni (all) + oculus (eye) — a wide field of view over your own environment, in service of defending it.

What Omniculus is not

This boundary is the product, not a disclaimer. It is deliberately chosen and should survive every future design decision:

  • Not a mass-surveillance system. It runs on data you are authorized to use: your own lab/network telemetry, public threat-intel feeds, and synthetic data. It is not for monitoring people you have no authority over.
  • Not an autonomous actuator. It recommends and explains; a human decides and acts. There is intentionally no response-execution component in the codebase. Consequential alerts route to a person.
  • Not a production weapon. It is developed and validated against synthetic data and lab replicas / digital twins — never pointed at infrastructure the operator does not own and is not authorized to test.

Why it exists

To study, in a safe lab setting, two hard questions:

  1. How do you make a detection system more vigilant over time without making it a black box — i.e. learning that stays explainable and auditable?
  2. How little can you collect and still answer the security question — can strong defensive analytics be built with data minimization rather than maximal collection?

Operating principles

Principle What it means in the code
Human-in-the-loop investigate/ ranks and routes; no module executes a response.
Synthetic-first ingest/synthetic.py is the default data source; real connectors are opt-in and authorization-gated.
Explainable detections Every Alert carries the rule that fired and its evidence events.
Auditable learning feedback/ logs every threshold change with the precision that justified it.
Data minimization A design goal and research question, not an afterthought.

Audience

Security researchers, blue teams, and students building intuition about detection engineering, data fusion, and accountable decision support.