808bits

log

  1. 007 2026-08-19 The ECDSA tax: a field guide to threshold signing Every MPC wallet pays a price set by one line of algebra. A tour of Lindell17, GG18/GG20, CGGMP, DKLs and FROST - who implements them, who audited what, which ones broke - with benchmarks from one machine.
  2. 006 2026-08-15 The key that never exists: a threshold signing ceremony in your browser interactive Three keyholders run a real 2-of-3 distributed key generation and FROST signing session, live on this page: every protocol message shown byte for byte, plus a sabotage lab. Zig compiled to WebAssembly, no server, no dealer.
  3. 005 2026-08-05 FIPS 140-3 is not a security guarantee, and auditors know it FIPS-validated modules shipped ROCA, EUCLEAK, Dual_EC_DRBG, and a YubiKey flaw caused by the FIPS self-tests themselves. What the certificate actually attests, what FIPS mode costs in performance, why it breaks Bitcoin tooling like BIP32, and where auditors actually look.
  4. 004 2026-08-02 You can't audit entropy after the fact: operator lessons from the Coldcard seed failure A five-year firmware bug left Coldcard seeds with 40-72 bits of entropy instead of 128, and ~$88M moved before most users heard about it. What survived, what didn't, and what key-generation evidence should have recorded.
  5. 003 2026-06-12 The Fireblocks recovery tools are free. Here's what they prove, and what they don't A technical look at Fireblocks' open-source key recovery tooling: what verify mode actually checks, what a successful run proves, and the recoverability questions it leaves open.
  6. 002 2026-06-05 DORA exit strategies for digital-asset custody: what a reviewer actually wants to see DORA requires documented, tested exit strategies for critical ICT providers. For crypto custody platforms, most firms have a paragraph where evidence should be. Here's what credible looks like.
  7. 001 2026-05-29 How to run a custody key-recovery drill without exposing a single private key A practical structure for rehearsing digital-asset custody recovery: scope, air-gapped setup, verify-only execution, deviation capture, and evidence that survives an audit.