Maths background. Building privacy tools.
Four projects, one thread: measuring what an adversary can learn — about your documents, your identity, your transactions — and reducing it.
PDF Changer — 21 browser-only PDF tools. Files never leave the tab. CSP-enforced connect-src 'none', triple-monitor integrity chain, cryptographic audit trail. 29k lines, 299 tests.
threadr — Open source reconnaissance for security teams. Dempster-Shafer evidence fusion, spectral clustering via Fiedler vector, Lévy-stable request timing. 9.8k lines, 282 tests.
ε-tx — Cryptocurrency privacy analysis. Information-theoretic bounds on what blockchain observers learn. 8 attack surfaces, 4 adversary models, Boltzmann entropy, inverse-OSPEAD ring construction. 8.7k lines, 340 tests.
degauss — Identity attack surface reduction. Scans 35+ platforms for your accounts, verifies which are yours, filters false positives, gives direct delete/privatise links. Tor-routed scanning, census-backed exposure scoring, attack feasibility assessment, legal request generation with regulatory escalation. 11.6k lines, 303 tests.
58k lines of TypeScript. 1,224 tests. Everything is MIT.