# Security Hardening Assessment — Hoffmann Infrastructure

**Date:** 2026-04-21  
**Requested by:** Matt (The Director)  
**Assessors:** Socrates (Backend/Infrastructure), Wadsworth (General Systems/Coordination)

---

## Executive Summary

This document assesses the current security posture of the Hoffmann home infrastructure and outlines hardening requirements across three layers: **Host/OS**, **Network/Access**, and **Application/Data**. Given the sovereign nature of this deployment (self-hosted AI systems, personal data processing), the recommended posture balances **zero-trust principles** with **operational sustainability**.

---

## Current Architecture Overview

| Component | Host | Network | Exposure |
|-----------|------|---------|----------|
| OpenClaw Gateway | Beelink (titanium-butler) | Tailscale + localhost:18789 | Internal only |
| Family Assistant | Beelink (titanium-butler) | Tailscale, localhost Radicale | Internal + Cloudflare Worker (email) |
| Local AI Inference | Gaming PC (3080 Ti) | Tailscale 100.104.147.116 | Internal only |
| ChromaDB/Radicale | Beelink | localhost:5232, Tailscale | Private network only |
| Cloudflare Worker | Cloudflare Edge | Public internet | Email webhook only |

**Key characteristics:**
- Tailscale mesh network for all inter-host communication
- No public-facing services (except Cloudflare Worker for email routing)
- Token-based authentication for OpenClaw
- htpasswd (bcrypt) for Radicale CalDAV
- App passwords preferred over OAuth refresh tokens

---

## Layer 1: Host/OS Hardening

### 1.1 Current State Assessment

**Beelink (titanium-butler):**
- Ubuntu 24.04 LTS (good — LTS with 5-year support)
- Systemd services: radicale.service, hoffdesk-webhook.service (active)
- No evidence of automatic security updates configured
- Disk encryption status: **Unknown** — requires verification
- Backup: Daily tar to local + Gaming PC via Tailscale SSH (sovereign backup)

**Gaming PC (inference node):**
- Windows (assumed, given 3080 Ti gaming setup)
- Tailscale for network access
- Ollama server on localhost:11434
- Status: **Needs assessment** — this is a gap

### 1.2 Required Hardening Actions

#### Critical (Do First)

| Priority | Action | Component | Effort | Risk |
|----------|--------|-----------|--------|------|
| 1 | Enable automatic security updates | Beelink | Low | Minimal |
| 2 | Verify disk encryption (LUKS) | Beelink | Low | None |
| 3 | Secure Ollama endpoint (bind to localhost only) | Gaming PC | Low | Minimal |
| 4 | Create service account for Family Assistant | Beelink | Low | Low |
| 5 | File permissions audit (600 for secrets) | All hosts | Low | None |

#### Important (Do Soon)

| Priority | Action | Component | Effort | Risk |
|----------|--------|-----------|--------|------|
| 6 | Fail2ban for SSH (if SSH enabled) | Beelink | Low | Low |
| 7 | UFW firewall rules (deny incoming default, allow Tailscale) | Beelink | Low | Low |
| 8 | Audit systemd service sandboxing (PrivateTmp, NoNewPrivileges) | Beelink | Low | Low |
| 9 | Log rotation for command-logger.jsonl | Beelink | Low | None |
| 10 | Windows security baseline (if applicable) | Gaming PC | Medium | Low |

#### Deferred (Post-v1.0)

- SELinux/AppArmor profiles for custom services
- Kernel live patching (canonical-livepatch)
- HIDS (AIDE or similar)

---

## Layer 2: Network/Access Hardening

### 2.1 Current State Assessment

**Strengths:**
- Tailscale provides mTLS mesh networking
- No exposed SSH ports to public internet
- Cloudflare Worker acts as email→webhook proxy (no direct email server exposure)
- Gateway bound to 127.0.0.1:18789 (localhost-only)

**Gaps:**
- Radicale CalDAV exposed on 0.0.0.0:5232 (necessary for mobile CalDAV sync?)
- No network segmentation between services
- Unknown if Tailscale ACLs restrict inter-device access
- No DDoS protection (not critical for home use)

### 2.2 Required Hardening Actions

#### Critical

| Priority | Action | Effort | Risk |
|----------|--------|--------|------|
| 1 | Restrict Radicale to Tailscale IPs only (or localhost + reverse proxy) | Low | Medium* |
| 2 | Audit Tailscale ACLs — verify principle of least privilege | Low | Low |
| 3 | Verify Cloudflare Worker token rotation | Low | None |

*Risk: Could break mobile CalDAV sync if not done carefully.

#### Important

| Priority | Action | Effort | Risk |
|----------|--------|--------|------|
| 4 | Implement Tailscale device approval for new nodes | Low | Low |
| 5 | Review Tailscale exit nodes/ subnets for data exfiltration paths | Low | Low |
| 6 | DNS-over-HTTPS (DoH) on all hosts | Low | None |

---

## Layer 3: Application/Data Hardening

### 3.1 OpenClaw/Gateway Security

**Current:**
- Token authentication enabled
- RedactSensitive: "tools" configured
- Dreaming, command-logger hooks active
- Active Memory disabled (performance issue, not security)

**Required:**

| Priority | Action | Effort |
|----------|--------|--------|
| 1 | Run `openclaw security audit --deep` | Low |
| 2 | Review and tighten OpenClaw file permissions (`--fix`) | Low |
| 3 | Verify gateway.bind is strictly localhost | None |
| 4 | Audit fallback model chain for unexpected API key exposure | Low |
| 5 | Review agent session isolation (prevent cross-contamination) | Low |

### 3.2 Family Assistant Security

**Current:**
- Secrets in `~/.openclaw/.env` and `scripts/.env` (chmod 600 assumed)
- Google Calendar API replaced with Radicale (reduced OAuth scope)
- Service account JSON in `~/.openclaw/secrets/` (needs verification)

**Required:**

| Priority | Action | Effort |
|----------|--------|--------|
| 1 | Verify all .env files are chmod 600 and gitignored | Low |
| 2 | Audit ChromaDB persistence for PII exposure | Low |
| 3 | Implement request rate limiting on webhook endpoint | Medium |
| 4 | Add circuit breaker for Radicale auth failures (3 consecutive → pause + alert) | Medium |
| 5 | Review LLM prompt injection vectors (email→calendar parsing) | Low |

### 3.3 Data Protection

| Priority | Action | Effort |
|----------|--------|--------|
| 1 | Verify backup encryption (at rest and in transit) | Low |
| 2 | Implement backup integrity verification (checksums/test restores) | Medium |
| 3 | Document data retention policy (ChromaDB, email cache, logs) | Low |
| 4 | PII scrubbing audit (confirm no plaintext PII in logs) | Low |

---

## Risk Tolerance Profile Recommendation

Based on the sovereign home-lab context, I recommend **"Home/Workstation Balanced with Zero-Trust Overlay"**:

- **Firewall:** Default deny incoming, explicit allow for required services
- **Updates:** Automatic security updates enabled
- **Access:** Tailscale-only for management, no public-facing admin interfaces
- **Backups:** Encrypted, tested, with retention policies
- **Monitoring:** Security audit alerts → Matt's DM (already configured)

---

## Implementation Roadmap

### Phase 1: Critical (This Week)
1. Run `openclaw security audit --deep` on Beelink
2. Enable automatic security updates on Beelink
3. Verify disk encryption status
4. Secure Ollama endpoint on Gaming PC
5. Verify all .env files are chmod 600

### Phase 2: Important (Next 2 Weeks)
6. Restrict Radicale to Tailscale-only access
7. Audit Tailscale ACLs
8. Configure UFW on Beelink
9. Implement log rotation
10. Create dedicated service account for Family Assistant

### Phase 3: Hardening (Ongoing)
11. Backup integrity verification
12. systemd service sandboxing
13. Application-level rate limiting
14. Circuit breaker for auth failures

---

## Cross-Agent Coordination Notes

**Socrates (Backend/Infrastructure):**
- Primary owner of host/OS hardening (Phase 1-2)
- Will execute `openclaw security audit --deep` and report findings
- Responsible for systemd service hardening and UFW configuration

**Wadsworth (General Systems/Coordination):**
- Primary owner of process coordination and scheduling
- Should verify backup integrity and retention policies
- Responsible for cross-agent security event routing

**Daedalus (Frontend):**
- Should verify no secrets in frontend build artifacts
- Review any client-side storage for PII

---

## Open Questions for Wadsworth

1. What's the current backup verification process? Do we test restores?
2. Are Tailscale ACLs currently restricting inter-device access, or is it default allow?
3. What's the status of command-logger.jsonl rotation?
4. Should we schedule periodic `openclaw security audit` via cron?

---

## Appendices

### A. Commands Reference

```bash
# OpenClaw security audit
openclaw security audit --deep
openclaw security audit --fix

# Ubuntu automatic updates
sudo apt install unattended-upgrades
sudo dpkg-reconfigure unattended-upgrades

# Check disk encryption
sudo cryptsetup status <device>

# UFW basic rules
sudo ufw default deny incoming
sudo ufw default allow outgoing
sudo ufw allow from 100.64.0.0/10  # Tailscale
sudo ufw enable

# File permissions
chmod 600 ~/.openclaw/.env ~/.openclaw/scripts/.env
chmod 600 ~/.openclaw/secrets/*.json

# Log rotation (add to /etc/logrotate.d/)
/home/hoffmann_admin/.openclaw/command-logger.jsonl {
    daily
    rotate 7
    compress
    delaycompress
    missingok
    notifempty
}
```

### B. Verification Checklist

- [ ] `openclaw security audit --deep` passes with no critical findings
- [ ] Automatic security updates enabled on Beelink
- [ ] Disk encryption verified (LUKS active)
- [ ] All .env files chmod 600
- [ ] Radicale bound to Tailscale IPs only
- [ ] UFW active with appropriate rules
- [ ] Backup integrity verified (test restore completed)
- [ ] Log rotation configured
- [ ] Ollama on Gaming PC bound to localhost
- [ ] Tailscale ACLs audited

---

**Document status:** Draft — awaiting Wadsworth assessment and Matt's priority confirmation.

**Socrates Assessment:** Backend infrastructure hardening is straightforward and low-risk. The bigger concern is operational sustainability — ensuring backups work, updates don't break services, and alerting is meaningful rather than noisy. The sovereign nature of this deployment means we trade cloud-managed security for direct control — that's the right trade for this context, but it requires discipline on monitoring and maintenance.