luke.yt
Security fundamentals, built not memorized.
Secure Infrastructure · SIEM · PKI · AI security research
$ ls -lt latest/
Latest
Can Sushi Fool AI?
A made-up sushi restaurant hides secret instructions in its web pages. Dozens of AI models were asked to summarise it. See what the hidden text says, and which models fell for it.
Read →
Prompt Injection Benchmark
The full results: a leaderboard of 25 models, per-attack verdicts, the attack surface of the page, and a detector test. Local, Claude, and OpenRouter models on the same 21 private attacks.
View the data →$ ls -la projects/
Featured Work
mTLS Infrastructure
Built a full public key infrastructure from scratch with mutual TLS enforced by HAProxy. A virtual OPNsense router handles rules and routing. All running on a remote proxmox server.
Read →
Wazuh SIEM
Deployed Wazuh, via Docker on a hardened Ubuntu server. Tuned alerts, correlated logs, and produced a full threat intelligence report on a real critical alert. Tracked attacker IPs, geolocated traffic, and mapped behaviour to MITRE ATT&CK.
Read →
$ cd explore/
What's Here
Staying Safe Online
Not here for the homelab deep-dives? This section is written for everyone - practical steps to protect yourself online, no technical background needed.
Get started →Prompt Injection Lab
A collection of pages with hidden prompt injection techniques to demonstrate how AI tools can be subtly manipulated. Seven attack categories with varying levels of complexity.
Open lab →
Research Report
Local LLMs vs Frontier Models - a preliminary look at Prompt Injection resistance. Eight local models and two frontier models tested across the full PI Checker suite.
View the test suite →
Security & Homelab Blog
A full build log - every decision, mistake, and fix, from the initial Proxmox setup through to a live mTLS-secured infrastructure with an active SIEM. 14 entries across 7 categories.
$ whoami
CompTIA Security+ · AA Cybersecurity · CFC Computer Science · Swiss Defcon Chapter Member
$ tail -n 13 skills.md