Early access programme

The shield for critical networks.

AEGIS is an Agnostic SIEM, self-hosted and built for telecom operators, banks and government. It collects from hundreds of network element types, detects threats in real time with open Sigma rules, and explains every alert, without your data ever leaving your network.

  • Self-hosted, air-gap ready
  • Open standards: Sigma, OCSF, SQL
  • Privacy by design

From signal to detection

Every source in orbit. Every threat mapped.

Sources orbit AEGIS and stream their events into it. Each event is normalized and checked against detections mapped to MITRE ATT&CK and to GSMA MoTIF, the mobile threat framework of the GSMA Mobile Cybersecurity Knowledge Base.

  • Source in orbit
  • Event ingested
  • MITRE ATT&CK tactic
  • GSMA MoTIF attack surface
  • Detection

Illustration. Detection packs and framework mapping are on the roadmap.

Why security teams look for a new SIEM

Cost grows faster than data

Typical sizing stores hot data uncompressed and keeps cold data on expensive compute nodes.

Onboarding is slow and blind

Hundreds of device and network element types, each set up by hand, and silent sources go unnoticed.

Critical traffic is a blind spot

Signalling, DNS and other domain-specific data are rarely covered by standard SIEM content.

Alert floods

Analysts get thousands of ungrouped alerts with no suppression and little context.

Locked-in content

Rules and parsers in proprietary formats cannot be versioned, tested or taken with you.

Seven reasons AEGIS is different

Network-native

An inventory of every network element type, zone-aware collectors and file collection that never takes the same file twice.

On the roadmap

Open and portable

Sigma rules, the OCSF schema, standard SQL and plain Kafka topics. Everything exports.

On the roadmap

Lower storage cost

Columnar compression and tiering from fast NVMe to object storage, with a sizing model you can rerun.

On the roadmap

Fast

Streaming detection on the event bus and analytical queries that answer in seconds.

On the roadmap

Explainable AI on your premises

A self-hosted model with read-only access. Every answer cites the queries and events behind it.

On the roadmap

Detection as engineering

Rules live in Git, are tested in CI, versioned, and measured for noise and coverage.

On the roadmap

Privacy built in

Subscriber and customer identifiers are tokenized by default; every re-identification is audited.

On the roadmap

How AEGIS works

One pipeline from collection to response, built on proven open-source components and running entirely on your infrastructure.

  1. Collect

    Syslog relays, SFTP collectors, API pullers, webhooks and edge agents, deployed per network zone.

  2. Stream

    Everything lands on Apache Kafka. Nothing is acknowledged before it is safely stored.

  3. Normalize

    Parsers map vendor logs to OCSF, protect personal data, enrich with GeoIP and threat intelligence, and quarantine what they cannot parse.

  4. Detect

    Streaming Sigma rules and correlation in real time; scheduled SQL rules over history.

  5. Store and search

    ClickHouse columnar storage, hot on NVMe and cold on object storage. Search with Explore, SQL or the AI assistant.

  6. Respond

    Grouped alerts, AI triage, investigations, playbooks with human approval, and notifications to your tools.

Built on Apache Kafka, Vector, Apache Camel, ClickHouse, PostgreSQL, Keycloak, OpenBao, Kubernetes, Prometheus, Grafana and OpenTelemetry.

Designed for speed and scale

Design targets. They will be measured and published as the platform matures.

from event to alert
< 60 sTarget
from ingest to searchable
< 30 sTarget
for a 24-hour aggregate query
< 5 sTarget
events lost without a counter
0Target
network element types per deployment
200+Target
events per second at full scale
500kTarget

Roughly half the storage

For 23.5 TB of logs per month, kept 3 months hot and 9 months cold, a typical sizing needs about 449 TB. The AEGIS sizing model needs about 205 TB for the same data.

Modelled comparison, based on 6:1 columnar compression and erasure-coded object storage for the cold tier. To be validated with production samples.

Typical sizing≈ 449 TBAEGIS model≈ 205 TB

Detection engineering, done properly

Write rules in standard Sigma, test them in CI, see how noisy they would have been before you enable them, and roll back in one step.

  • Streaming detection, correlation and scheduled SQL rules
  • Every enabled rule has positive and negative tests
  • Retro-hunt a new rule over past data before it fires
  • Coverage mapped to MITRE ATT&CK and GSMA MoTIF, with ATT&CK Navigator export
  • Suppression and grouping instead of alert floods
  • Every alert links to the exact rule version that fired it

AI that shows its work

Ask questions in plain language. The assistant answers by running read-only, row-capped queries, and shows every query and result it used.

  • Self-hosted, OpenAI-compatible models (vLLM, Ollama, LiteLLM)
  • No outbound model traffic unless an admin configures it
  • Personal identifiers never reach the model in clear text
  • AI proposes; people approve. Every AI turn is audited

Principles we do not compromise on

No silent data loss

Every dropped, filtered or unparseable event is counted and visible. Bad data goes to quarantine, never to the bin.

Each file collected exactly once

No duplicate collection, even when a collector fails over mid-file.

Humans approve actions

AI and automation may suggest. Containment and destructive actions always need a named human approval.

Everything as code

Detections, parsers, pipelines and dashboards live in Git and deploy through CI and GitOps.

Security and privacy by default

Your premises

Runs on your own Kubernetes, across data centres and fully air-gapped networks. No data leaves your infrastructure.

Defence in depth

Access is enforced in the API and in the database. Secrets live in a vault and are never displayed or logged.

Hardened supply chain

Non-root, read-only containers; signed images and an SBOM for every release.

Personal data protection

Designed to support Indonesia's Personal Data Protection Law (UU PDP): tokenization, access logging and retention limits.

Tamper-evident audit

Every change, login, query, export and AI turn is audited, with a hash-chained, write-once copy.

Accessible

WCAG 2.1 AA, usable on a phone, with Dark and Light themes.

Works with the tools you already run

Identity

  • Keycloak
  • OIDC and SAML
  • LDAP
  • SCIM 2.0

Notifications and ticketing

  • Email
  • Slack
  • Microsoft Teams
  • PagerDuty
  • Opsgenie
  • Jira
  • ServiceNow
  • Webhooks

Threat intelligence

  • MISP
  • OpenCTI
  • abuse.ch
  • Spamhaus DROP

Response and cases

  • Shuffle
  • Tracecat
  • DFIR-IRIS

Sources

  • Syslog
  • SFTP
  • OpenTelemetry
  • Okta
  • Microsoft 365
  • Google Workspace
  • GitHub
  • Edge agent

Standards

  • Sigma
  • OCSF
  • MITRE ATT&CK
  • OpenAPI 3.1

The road to general availability

We build AEGIS in phases, each one proven before the next. Early access partners help us set priorities.

  1. In progress

    Foundation

    Delivery platform, single sign-on, secrets vault, signed images and security gates.

  2. On the roadmap

    End-to-end ingest and search

    First syslog and SFTP sources end to end, Explore, SQL and audit, plus sizing validation with real data.

  3. On the roadmap

    Detection core

    Streaming Sigma and correlation, scheduled rules, alerts, notifications and detection as code.

  4. On the roadmap

    Analyst workflow

    Investigations, threat intelligence, cases, playbooks with approvals, AI assistant and dashboards.

  5. On the roadmap

    Network scale

    Network element inventory, bulk onboarding, tokenization, telco detection packs, high availability and reports.

  6. On the roadmap

    General availability

    Parallel run with your current SIEM, parity reports and cutover.

Frequently asked questions

Is AEGIS a cloud service?

No. AEGIS runs on your own infrastructure, including fully air-gapped networks.

Is AEGIS open source?

AEGIS is a commercial product built on open-source components and open standards (Sigma, OCSF, SQL, Kafka), so your data and detection rules stay portable.

How is AEGIS priced?

By daily ingest volume (GB per day, with an events-per-second equivalent). Contact us for a quote. There is no free edition.

Can we keep our existing detection rules?

Sigma rules import directly, and AEGIS exports its detections as standard Sigma. Other formats can be converted.

How do you protect subscriber and customer data?

Identifiers such as phone numbers and IMSI are tokenized at ingest by default. Clear values are only available to authorized roles, with a stated reason, and every request is audited.

Does the AI send our data to the cloud?

No. The model is self-hosted by default and reads data only through a read-only, row-capped interface.

How does migration from our current SIEM work?

Through a parallel run with your current SIEM and per-source parity reports before cutover.

When can we start?

Join the early access programme and we will contact you about a pilot.

Shape AEGIS with us

We are looking for early access partners in telecom, banking and government.

Join early access