A phone architecturally incapable of betraying you

(github.com)

2 points | by Godofall 8 hours ago ago

1 comments

  • Godofall 8 hours ago

    I've been building a sovereign mobile computing platform for the past year. The core thesis: most "secure" phones are policy-secure, not architecturally secure. A court order, a national security letter, or a compromised update server can still get your data. We're trying to make that architecturally impossible.

    The stack:

    seL4 microkernel (formally verified, capability-based) RISC-V ISA (open, auditable, no vendor backdoors) Post-quantum crypto: ML-KEM-1024, ML-DSA-87, CRYSTALS-Dilithium BeskarVault HSM: 32 key slots, physical tamper destruction Continuous Guardian: 50ms integrity checks, lives in ROM BeskarLink: Signal Protocol + post-quantum augmentation Shield Ledger: immutable Merkle-tree audit log Zero cloud dependency — fully offline capable Honest status: Phase 1, running on VisionFive 2 (JH7110). Not a phone you can buy. Architecture validation platform. 11/11 tests passing. CI/CD configured. Dual licensed (open source core + commercial tiers).

    What we're looking for: cryptographers, seL4 developers, RISC-V engineers, formal verification people. Not hiring — contributing. This exists because the problem is real.

    Happy to answer technical questions about the seL4 capability model, the post-quantum primitives, or why we chose RISC-V over ARM for this threat model.