I don't stop at running security tools — I check whether their results are right. My own analyzers and rules are tested before I trust them, and I record what failed as carefully as what worked. Career switcher into security, with a hardware background (KAIST IDEC, 831 h of semiconductor design courses).
| Area | What I actually did |
|---|---|
| Software supply chain | API-BOM (CycloneDX) + SBOM (Syft → Dependency-Track) linked endpoint → code path → package → CVE. 309 of 403 CVEs (77%) unreachable from any endpoint — measured after fixing 5 bugs in my own analyzer |
| Detection engineering | Wazuh rules generated from an approved OpenAPI allowlist, verified with wazuh-logtest. Found and fixed a deployed BOLA rule that worked backwards |
| Infrastructure security | Terraform checked without a cloud account (terraform test + mocks, Trivy, Checkov). Replaced an AdministratorAccess role with split least-privilege roles and an IAM action checker |
| Next: software → kernel correlation | Linking API endpoint → process → libraries actually loaded at runtime (strace · eBPF) → SBOM package → CVE → kernel. Today the kernel-layer graph is a static sketch |
| Study → apply | openssf-study — each OpenSSF concept applied to my own project and verified before / after (first: least privilege) |
| Project | Highlight |
|---|---|
| Multi-Cloud DR — AWS → Azure | Site-to-Site VPN · DMS · Route 53 / CloudFront failover (team). Re-verified locally: RTO 18.7 s · RPO 0 · terraform test 8/8 |
| API Security · Supply Chain | API-BOM / SBOM / VEX / reachability + automated Wazuh rules (team final project). Tested on apps not used in development: VAmPI failed (off-spec probe recall 60%) → fixed in v2.1 (93%) → 4th unseen app DVRA, config only: 15 / 15 probes caught by rules, real BOLA flagged high by ML (ML release gate narrowly missed — recorded as is). Grafana dashboard built as code |
| API Defense Lab | A practice range for API defenders: a light core (vulnerable DVRA + nginx auto-block + Sigma shadow-API rules + ML BOLA detection + answer key + scorer, ~220 MB RAM) plus optional modules by job — Wazuh for SIEM rule authoring, ES + Grafana for live monitoring. Fresh-clone runs: probe blocked in 2–3 s, BOLA in ~6.5 s, 0 / 20 normal users blocked after fixing a first-run false positive. Sigma and Wazuh verdicts match line by line (15 / 15). Looking for testers on Linux / macOS |
| SOC Detection Notes | Wazuh home SOC lab — MITRE rules (T1547, T1046) and SIEM incident write-ups |
| ZenSOC — Wazuh → Slack | Grouping cut duplicate alerts 32 → 3; the entropy hypothesis came out the opposite way — recorded as such |
| Enterprise Network Lab | Multi-site network in Packet Tracer: VLSM, OSPF multi-area with dual-link cost failover (primary link down → backup in one lost ping), Router-on-a-Stick vs SVI, inter-VLAN ACL (ICMP 3/13), static + OSPF split routing — 7 problems and fixes logged |
- wazuh/wazuh#35707 — Sysmon EventID 13 reaches
archives.jsonbut no alert fires (4.9.0); traced the rule-chain difference between Security- and Sysmon-channel events - wazuh/wazuh#36029 — custom
if_sidrules forwindows_eventchannelstop firing in 4.9.0 (if_groupworks); behavior change since 4.7.5, andwazuh-logtestdiffering from liveanalysisd
- Supply chain — CycloneDX · VEX · Syft · Grype · Dependency-Track · Trivy · Checkov · GitHub Actions
- Detection — Wazuh · Event-Driven Ansible · Elasticsearch / Kibana
- Infrastructure — AWS · Terraform · Kubernetes · Linux · Python
- KAIST IDEC semiconductor design training — 67 courses · 831 h (2024.04 – 2026.03): digital design (Verilog, RTL-to-GDSII), computer architecture, AI accelerators, embedded networks (CAN · FlexRay)
- Embedded systems coursework — in progress
RHCSA · AWS Advanced Networking – Specialty · AWS Data Engineer – Associate · AWS AI Practitioner · 정보처리기사 · CKA (in progress)

