← ProjectsLightweight incident-response sensor

SENTINEL

A small C-based Linux telemetry daemon for rapid incident response and threat hunting, emitting portable JSON events without requiring a heavyweight endpoint platform.

C + minimal dependenciesJSON-first telemetryDetect → preserve roadmap

Problem

Response teams often need immediate host visibility before a full EDR or SIEM integration can be deployed. SENTINEL aims to provide useful process, file, network, and authentication context with minimal dependencies and very little operational ceremony.

Constraints

  • Remain lightweight enough for rapid deployment on Linux hosts, containers, and lab systems.
  • Emit structured events that can flow into existing logging stacks instead of inventing a proprietary backend.
  • Collect useful host context while keeping the implementation understandable and auditable.
  • Preserve a path toward richer detection and evidence collection without turning the sensor into a remote-control agent.

Architecture

  1. 01

    A /proc scanner discovers process activity while inotify watches sensitive files and network inspection tracks listener changes.

  2. 02

    Collectors normalize observations into line-oriented JSON events on standard output.

  3. 03

    Deployment can remain native or containerized, with downstream transport delegated to tools such as Fluent Bit, syslog, Loki, Splunk, or Wazuh.

  4. 04

    Future phases add event-driven kernel interfaces, filtering, rule evaluation, enrichment, and evidence-preservation actions.

Failures / lessons

  • Small sensors are easier to deploy during an incident when they do not demand a new management plane first.
  • A neutral event stream lets collection stay decoupled from whatever SIEM happens to be available.
  • Detection logic and collection mechanics should evolve independently so richer rules do not make the sensor harder to trust.
maydit://command↑↓ history · TAB complete · ESC close

MAY'D IT command interface v0.2.1

Type help for commands. Try play fragment.