mark@homelab / labPERSONAL RESEARCH
❯ tree ./lab -L 1
./lab
infrastructure/compute, inference & cloud
virtual-machines/security tools & test systems
projects/build logs & practical guides
agentic-soc/local AI meets security telemetry
❯Build. Break. Investigate. Repeat.
[ Infrastructure / Security research / Local AI ]

Inside the Homelab

My environment for exploring security operations, analysing malware, testing detections, and building automation. A self-built Proxmox server provides the core lab, alongside a local AI workstation and Azure services.

Local compute. Cloud integration.

Each part of the environment has a distinct role: hosting lab machines, running models locally, or providing cloud security telemetry.

Virtualisation host

Self-built Proxmox server

  • Intel Xeon E5-2648L
  • 32 GB RAM
  • 2 TB SATA SSD
  • Proxmox virtualisation
Local inference

Windows workstation

  • AMD Ryzen 7 5800X
  • 64 GB RAM
  • Radeon RX 7600 XT / 16 GB VRAM
  • Ollama serving local Qwen models
Cloud services

Azure

Microsoft Sentinel and Log Analytics provide telemetry for the SOC projects. This website is hosted separately in Azure Blob Storage and delivered through Azure Front Door.

SOC project / Component relationships
Proxmox / Ubuntu VMDocker hosts n8n orchestration and the Azure MCP service.
Windows / Ollaman8n calls the local model for summaries and investigation reasoning.
Azure / SentinelThe MCP service provides authenticated access to Log Analytics queries.

n8n coordinates the workflow, Ollama supplies local inference, and Azure MCP connects the investigation to Sentinel telemetry.

The virtual lab

A collection of monitoring, analysis, and target systems supports practical security exercises.

Monitoring & automation

QRadar + Ubuntu Docker VM

QRadar supports log ingestion, parser development, and detection testing. The Ubuntu Docker VM hosts services used in the automation projects, including n8n and Azure MCP.

SIEMContainersWorkflow automation
Malware analysis

FLARE-VM + REMnux

Windows and Linux analysis environments for examining suspicious files and developing practical malware-analysis skills.

Windows analysisLinux analysis
Testing & investigation

Kali Linux

A dedicated VM for security testing and generating controlled activity to investigate in the lab's monitoring tools.

Security testingLab validation
Windows environment

Domain controller, client & IIS

A Windows domain controller and client provide a domain environment for lab exercises. A separate standalone IIS server supports web-server testing and process-telemetry investigations.

Active DirectoryWindows clientIIS

Build logs & practical guides

The configuration work behind the lab, from deploying a SIEM appliance to turning raw firewall logs into useful security events.

Infrastructure / QRadar CE 7.3.3

Installing QRadar on Proxmox

Importing the appliance disk, configuring the VM, and documenting logging and auto-update troubleshooting.

Log engineering / pfSense

Building a custom pfSense DSM

Taking raw filterlog syslog through event identification, QRadar mappings, and normalised network-property extraction.

Security tooling / Local AI

RedAmon on Proxmox

Deploying RedAmon with a separate Ollama inference host, validating initial reconnaissance, and tracing an unresolved AI Agent tool-selection failure. The write-up distinguishes working deployment and recon from unverified agent execution.

DockerOllamaNeo4j

For forensic investigations and detection-validation reports, explore the technical field notes.

Building towards an agentic SOC

A series of projects exploring local inference, workflow orchestration, and access to real security telemetry. The first project is a news briefing pipeline; the later projects focus on Sentinel investigations.

Stage 1 / Local AI briefing

From RSS to a SOC-oriented briefing

n8n retrieves security news, passes source material to a local Ollama model, and delivers summaries through Telegram. This project established the local inference and automation foundations.

RSSOllamaTelegram
Stage 2 / Tool integration

Connecting Azure MCP

An authenticated MCP service connects the n8n agent to Azure, enabling real Log Analytics queries against Sentinel telemetry.

Azure MCPDockerKQL
Stage 3 / Investigation proof of concept

Investigating Sentinel incidents

A locally hosted agent queries Cowrie honeypot telemetry through MCP and delivers an investigation report to Discord. The write-up records the demonstrated workflow and its evidence-validation limitations.

Microsoft SentinelQwenDiscord
Stage 4 / Incident looping & history

Stateful, multi-incident processing

The workflow investigates incidents sequentially, skips those already recorded in a persistent processing history, and aggregates Discord message responses before writing one completion record per incident. Execution remains on demand; automatic scheduling is deliberately deferred.

Loop Over ItemsData TablesDiscord aggregation

The website moved. The lab kept growing.

Raspberry PiProxmoxAzure

This website began on a Raspberry Pi, later moved onto the Proxmox server, and is now hosted in Azure. The homelab remains the environment behind the experiments and projects documented here.