These notes document the work of Sein Linn, a cybersecurity engineer based in Singapore. The practice spans offensive security research, including three published CVEs and a seventh-place finish at the Magnet Virtual Summit 2026 in digital forensics, and applied AI work spanning browser-only language clients and Burmese natural language benchmarks. A consistent thread runs through both: building the tools the work itself requires, and publishing them openly.
Eight entries from the public repository, drawn from the work itself rather than from a résumé.
A zero-dependency Python proof-of-concept for Copy Fail, a Linux kernel local privilege escalation through AF_ALG and splice page-cache corruption. Runs on Python 3.7+, nothing to compile.
security research · python · lpe poc
An AI chat client that runs entirely in the browser. The user brings an OpenRouter key; nothing is logged or proxied. Four modes: chat, visible-reasoning, web search, and vision.
next.js · typescript · openrouter
A shared benchmark for Burmese natural language understanding: four tasks, fixed splits, reproducible baselines, so results across papers and projects are comparable.
nlp · benchmark · low-resource
Payloads and working notes from completing the PortSwigger Web Security Academy: thirty topics covering SQL injection, XXE, request smuggling, template injection, and more.
web security · payloads · owasp
Bulk OpenAI API key validation behind a terminal-styled interface: paste or upload keys, watch them validate in real time, export the results.
flask · api tooling · concept stage
A Next.js retrieval-augmented generation pipeline: documents go in, answers come out with the retrieved chunks attached, so the search can be checked.
rag · vector search · embeddings
A Firefox build of Video DownloadHelper with the premium features enabled. The downloader engine is upstream code; only the paywall checks changed.
firefox extension · repack
A mirror of a0rtega/pafish, the tool malware uses to detect virtual machines and analysis environments, kept open so analysts can study anti-analysis techniques.
malware analysis · sandbox detection