Independent security researcher — electronic travel document authentication (eMRTD / ePassport, ICAO 9303 & PKI) and self-hosted AI security infrastructure. Founder, VIATREE.

An iPhone app that runs the same cryptographic passport check used at airport immigration counters — implemented directly against the ICAO 9303 standard rather than wrapped around a commercial SDK.

ICAO Doc 9303 Parts 10/11 · BAC / PACE over CoreNFC · Passive Authentication (RSA · ECDSA · SHA-256) · ICAO PKD Master List trust anchors · iOS 16+ / Swift

Fully on-device — MRZ OCR, the chip session and signature validation never leave the phone. No accounts, no server calls, no images uploaded.

→ App Store · epassport-web.vercel.app · site source

Python tooling for the ICAO 9303 PKI stack — BAC, PACE, Active Authentication, Chip Authentication, Terminal Authentication, Passive Authentication.

Extracted and parsed 581 CSCA certificates across 112 countries from the German BSI Master List. Analysed how AA_Failed / CA_Failed surface as independent error conditions across commercial SDKs (Regula Document Reader) and open-source validators (OST Kinegram eMRTD Validator).

Also used to disprove a third-party claim of having defeated AA and CA verification: the observed Passive Authentication failure was evidence against the claim, not for it.

Two AI models attack an isolated lab; the surviving defenses evolve; every result has to be provable rather than self-reported.

• Attack-A and Attack-B must be different model revisions, with independent sessions, run directories, lab clones, budgets and event streams
• Scripts, fixtures, static JSON and same-model aliases are barred from registering as contestants
• Every real match emits a run_id, provider request evidence, tool events, an environment state diff and an event root hash
• Winning one match earns candidate champion only — a formal champion clears 5 hidden scenarios x 3 seeds
• Raw traces, match evidence, training provenance, frozen assets and generational lineage are never auto-pruned

Windows control plane to WSL2 GPU worker, Electron desktop client, macOS/iOS runner planned.

Device-security research turned into a working console: how far a stock, non-jailbroken iPhone can be legitimately reached from a Windows PC across arbitrary networks — no jailbreak, no exploit, no cable, developer-signed on owned hardware.

Existing iOS mirroring tools land in one of two buckets: view-only, or touch via Bluetooth pairing — which puts the computer in the same room as the phone. PRISM does neither. Screen and input travel over a single end-to-end encrypted link, so the PC and the phone never have to share a network.

iOS 26.5.2 · live screen + touch injection · text input · hardware buttons · multi-terminal handover (~1s switchover)

Measured: 108 minutes continuous with the phone off Wi-Fi on carrier data only — 205 probes, 2 failures, both inside the Wi-Fi→cellular handover. Not yet verified: tower handover with the phone physically off-site.

The platform I'm building to turn security research into something usable — packaging tooling, findings and demo material into verifiable, shareable outputs, with a distribution and delivery pipeline of its own. Same discipline as everything else here: nothing ships as a claim unless it's measured.

In active development.

→ viatreebot.com

• No fabricated metrics. If a number isn't measured, it doesn't ship.
• Open standards over vendor SDKs. If it passes here, it passes at a real border.
• On-device by default. Cloud only as an explicit, declared fallback.

📧 303@viatreebot.com · 🌐 viatreebot.com · 💼 linkedin.com/in/viatree