Backend & Systems Engineer · Security Research · Portfolio

I build backend systems in Go and C/C++, and research security at the kernel and binary level. Security is a design input, not a patch.

• Shipped a 10-module warehouse inventory and dispatch system for a live enterprise client
• Co-authored an IEEE paper on parallel blockchain execution (4.3x throughput)
• Merged PRs in freellmapi and Epic Games' PixelStreamingInfrastructure
• Focus: eBPF, TPM attestation, binary analysis, post-quantum migration

Crypto-usage evidence engine for compiled binaries. Ghidra P-code lifting plus a GNN recognizer for classical and post-quantum primitives. Deterministic taint engine emits CWE-mapped facts with provenance paths. CycloneDX 1.6 CBOM output. Trained on a 9,295-binary corpus across GCC/Clang, x86_64/AArch64.

Hardware-anchored attestation for Linux. TPM2 quote bound to TLS session (RFC 9266) closes relay attacks. eBPF-LSM monitors kernel module loads and ptrace. Verified ALLOW/DENY on Secure Boot on/off hosts.

Adaptive SSH honeypot in pure Go. Ephemeral Docker container per session. ML (Logistic Regression, Naive Bayes, anomaly detection) trained on 170MB+ of Cowrie/CIC-IDS/NSL-KDD scores intent live. Mutates the environment to defeat fingerprinting. eBPF execve tracing, egress sinkhole, AWS honeytokens.

Cryptographic inventory and governance platform. Go scanner, CT-log subdomain discovery, Quantum Migration Risk Scores. PNB PSB Hackathon 2026.

Conflict-aware parallel transaction scheduler for DAG blockchains. IEEE paper.

Blockchain alumni mentorship platform. · AI healthcare and emergency app. · Live full-stack platform for a startup.

• freellmapi: admin hardening. Stricter CSP, per-IP rate limiting, password-gated key export (#498)
• PixelStreamingInfrastructure: opt-in streamer token auth on signalling server
• MorphDAG: benchmarking feature

🥇 Smart SNU Hackathon '25 · 📄 IEEE 2025

📄 Paper (IEEE)
Optimized DAG-based blockchain designed for throughput and concurrency improvements.
By introducing per-account read/write frequency classification, OPT-MorphDAG improves transaction scheduling, reducing block latency and increasing throughput up to 2.5× over serial execution and 1.4× over MorphDAG baseline, with no inconsistencies.