Newer
Older
aegis-public / docs / tutorials / 04-ai-driven-protection.md

Tutorial Series: Exposing a Service with Aegis

# Tutorial Description
1 Local HTTPS with a whoami service Configure the gateway manually through the UI
2 Configure the Gateway with Owl AI Let Owl AI do the configuration from a single prompt
3 Understanding the Dashboard Read live traffic data and analyse request patterns with Owl
4 AI-Driven Protection — Disable & Re-enable a Service ← you are here Use Owl to take a service offline under attack and bring it back

Part 4 — AI-Driven Protection: Disable & Re-enable a Service

Prerequisite: Complete Part 1 or Part 2 so you have whoami.local running and accessible.

When a DDoS or targeted attack hits one of your services, the fastest response is to pull it offline at the gateway — no firewall rules, no server changes, just one command to Owl. The service goes dark in under a second. Once the attack subsides, Owl brings it back just as fast.

This tutorial demonstrates the full cycle: disable a cluster, verify the service is unreachable, then re-enable it.

Owl disabling and re-enabling the whoami cluster


Step 1 — Ask Owl to disable the cluster

Open the Owl chat panel (🦉 bottom-right) and type:

Disable the whoami cluster

Owl will call gateway_toggle_resource to disable the cluster. Aegis immediately rebuilds the xDS snapshot without it — Envoy stops routing to whoami within ~1 second.


Step 2 — Verify the service is unreachable

Open https://whoami.local in your browser. You should see a 503 Service Unavailable — Envoy has no healthy upstream for the filter chain.

This is exactly what happens during a DDoS response: the service is shielded at the gateway layer while the backend stays untouched.


Step 3 — Ask Owl to re-enable the cluster

Once you're ready to bring it back:

Re-enable the whoami cluster

Owl toggles the cluster back on. Aegis pushes the updated snapshot to Envoy and https://whoami.local is accessible again within a second.


Why this matters

Scenario How Owl helps
DDoS targeting one service Disable that cluster — attackers hit a 503, backend is protected
Vulnerability disclosed overnight Owl can pull affected services offline while you patch
Maintenance window Disable cleanly from chat, no SSH or YAML edits
Gradual re-enable after incident Re-enable, watch dashboard, disable again if attack resumes

The key advantage: this operates at the xDS layer. Disabling a cluster removes it from Envoy's routing table entirely — no lingering connections, no half-open sockets. The backend process keeps running; only the gateway path is cut.


AI Patrol automation

Rather than waiting for you to notice an attack, Aegis can detect and respond automatically. When the AI patrol sweep identifies a high-threat pattern — flood of requests, known scanner signatures, spike from a single ASN — it can be configured to:

  1. Block individual IPs automatically (based on threat score thresholds)
  2. Alert you via Telegram/Discord/Slack with a summary
  3. Let you instruct Owl to disable the cluster via a single reply

Go to Settings → AI to configure patrol sweep thresholds and notification channels.


Part 3 — Understanding the Dashboard