1. Work 01
  2. Craft 02
  3. Research 03
  4. About 04
  5. Writing 05
  6. Contact 06
Email LinkedIn GitHub CV
All work

10 / 12 · AI · Security research

SentinelAI

An architecture for intelligent threat detection and attack-chain analysis for small and medium enterprises.

Page from the SentinelAI paper
Category
AI · Cybersecurity research
Role
First author of four
Context
Architecture paper, VIT
Status
Evaluation planned, not completed

What it does

  1. Five-layer security pipeline.
  2. OCSF-aligned telemetry.
  3. Isolation Forest combined with rule-based detection.
  4. Entity-based event correlation into candidate attack chains.
  5. MITRE ATT&CK classification and D3FEND countermeasure mapping.
  6. Language model restricted to a validated explanation layer.
  7. Evaluation on CICIDS2017 and an Atomic Red Team lab is planned.

The main design decision is a boundary. No language model touches detection, correlation or classification. The model only explains an incident record that the earlier layers have already built.