- challenge/v6 is now hardenProject() (== bla harden) output over a public template: Node vault + HTML panel (executed by the vault) + WASM core, bankrupt, llmDirectives=false, emitManifest=false, fixed seed. - Per-build constants come from CRACKME_V6_KEY/FLAG; SEAL.json publishes digests only. Public demo fixture drives CI. - crackme:seal + tests: digest pinning, uniform DENIED, honeypots, fixture re-harden unlock + reproducibility, anti-strip, secret-gated release reproduction. - Archive PR #38 tiers under challenge/archive/v6-tiered-protect, v5.2 under challenge/archive/v5.
3.9 KiB
BLA CrackMe v6 — Harden-Native Edition
The public target is bla harden output. No hand-written obfuscation, no
custom CTF crypto: a small multi-format app is hardened by hardenProject()
(the exact pipeline behind bla harden <dir>), and the result is published as-is.
| File | What it is | Hardened by |
|---|---|---|
hardened/vault.cjs |
Node.js entitlement gate (entry point) | JS AST shield + exhaustion / deception suites + anti-strip interlock |
hardened/panel.html |
HTML panel whose inline script derives the stage-B token (the vault executes it) | HTML multi-script shield |
hardened/core.wasm.gz |
WebAssembly mix / seal core (gzipped; ~7 MB raw) |
WASM shield (decoy exports / bloat) |
SEAL.json |
Public commitment: knobs, keySha256, flagSha256, artifact + template digests |
— |
Harden knobs (pinned in SEAL.json, enforced by CI): intensity=bankrupt,
llmDirectives=false (risky LLM-directive layers OFF), emitManifest=false,
fixed seed → byte-for-byte reproducible.
Run
npm run crackme:v6 -- '<16-char-key>'
# or: node challenge/v6/run.mjs '<16-char-key>'
run.mjs copies the bundle to a temp dir, gunzips core.wasm, clears
NODE_OPTIONS, and runs vault.cjs. Success prints FLAG{...}; anything else
prints a single uniform [vault] DENIED (no per-stage oracle).
Rules of the game
The cleartext template is public (src/) — reading it tells you the
shape of the gate, not the answer. Every build fills in fresh constants (16 WASM
lane targets, the panel token, the encrypted flag) from maintainer secrets, then
hardens. To win you must recover those constants and the real logic from the
hardened output, i.e. defeat BLA:
- Stage A — each key byte goes through
seal(mix(byte, tweak_i))in the hardenedcore.wasmand must hit a hidden 32-bit lane target. - Stage B — the vault executes the hardened
panel.htmlscript; its token must match. - Stage C — the key decrypts the flag;
SHA-256(flag)must equalflagSha256. - Retired keys (v4.1 / v5.2) are honeypots that print decoy flags.
Prove a solve
node -e 'console.log(require("crypto").createHash("sha256").update(process.argv[1]).digest("hex"))' 'FLAG{...}'
It must equal flagSha256 in SEAL.json. Open an issue with the
CrackMe v6 solve template (hash + method notes). A general bypass of BLA's
hardening is worth more than a one-off patch of this fixture — say so in the issue.
Honest scope
BLA raises the cost for automated AI reverse-engineering agents and plants decoys. It does not claim immunity to patient interactive reversing (debuggers, Frida, manual analysis). Known gap surfaced by this challenge: numeric literals are not de-literalized by the JS shield, so the template keeps constants in strings (which BLA does split / decoy).
Reproducible build (maintainers)
npm run build
npm run crackme:build -- --fixture # public demo secret → challenge/v6/.fixture (gitignored)
CRACKME_V6_KEY='<16 chars>' CRACKME_V6_FLAG='FLAG{…}' npm run crackme:build # release → hardened/ + SEAL.json
npm run crackme:seal # CI gate: digests, deny/honeypot, fixture re-harden + reproducibility,
# and (if secrets match SEAL) release unlock + byte-for-byte rebuild
Equivalent CLI recipe: materialize the template into app/ (see
scripts/crackme/lib.mjs#materializeVault), copy src/panel.html and
src/core.wasm, then bla harden app --intensity bankrupt --seed <SEAL.seed>.
History
../archive/v6-tiered-protect/— first v6 draft (PR #38): single-file L1/L2/L3protecttiers.../archive/v5/— v5.2 hand-crafted Cognitive-Armor sponge.