crypto-lab-kdf-arena

What It Is

KDF Arena is a live, in-browser benchmarking tool that compares four key derivation functions side-by-side: HKDF-SHA256, PBKDF2-SHA256, scrypt, and Argon2id. It measures wall-clock derivation time and estimates memory usage for each function using identical input (password + random 16-byte salt). HKDF is included for educational contrast — it is an extract-and-expand KDF for already-strong key material, not a password-hashing function. The three password KDFs (PBKDF2, scrypt, Argon2id) are compared at their recommended default parameters to illustrate the trade-off between iteration hardness and memory hardness.

When to Use It

• Choosing a password KDF for a new system — run the benchmark on target hardware to see real timing costs before committing to PBKDF2, scrypt, or Argon2id.
• Tuning cost parameters — adjust iterations (PBKDF2), N/r/p (scrypt), or time/memory/parallelism (Argon2id) and observe the impact on derivation time and memory.
• Teaching the difference between HKDF and password KDFs — the sub-millisecond HKDF result makes it visually obvious that HKDF is not designed to resist brute-force attacks on passwords.
• Comparing browser WASM performance — Argon2id runs via argon2-browser (compiled to WASM), so results reflect real browser overhead.
• Do NOT use these results as a server-side benchmark — browser single-threaded WASM performance does not represent native C/Rust implementations on a server.

Live Demo

systemslibrarian.github.io/crypto-lab-kdf-arena

Enter any password string and click Run Benchmark to derive 32-byte keys with all four KDFs using a shared random salt. The results panel shows wall-clock time in milliseconds, estimated memory usage, and a hex preview of each derived key. A horizontal bar chart provides a visual timing comparison.

What Can Go Wrong

• PBKDF2 iteration count too low — using fewer than 600,000 SHA-256 iterations (NIST SP 800-132 2023 guidance) makes offline brute-force feasible on modern GPUs.
• scrypt N parameter too small — if 128 * N * r fits comfortably in GPU memory, scrypt loses its memory-hardness advantage over PBKDF2.
• Argon2id memory set below 64 MB — reducing the memory parameter shrinks the cost asymmetry between defender and attacker; OWASP recommends at least 64 MB for interactive logins.
• Using HKDF to hash passwords — HKDF has no cost parameter and completes in microseconds, offering zero brute-force resistance.
• Salt reuse across users — all four KDFs require a unique random salt per credential; reusing a salt enables precomputation (rainbow-table) attacks.

Real-World Usage

• Argon2id — default password hash in the libsodium crypto_pwhash API, adopted by 1Password, Bitwarden, and the PHC (Password Hashing Competition) winner.
• scrypt — used by Tarsnap for key derivation and by Litecoin's proof-of-work algorithm; recommended in RFC 7914.
• PBKDF2-SHA256 — required by WPA2 for Wi-Fi key derivation, used in LUKS disk encryption, and specified in NIST SP 800-132.
• HKDF-SHA256 — used by TLS 1.3 (RFC 8446) for deriving traffic keys from the handshake secret, and by the Signal Protocol for ratchet key derivation.

How to Run Locally

git clone https://github.com/systemslibrarian/crypto-lab-kdf-arena
cd crypto-lab-kdf-arena
npm install
npm run dev

Related Demos

• crypto-lab-kdf-chain — sibling KDF demo with a decision tree, derivation chains, and an attacker-cost model.
• crypto-lab-bcrypt-forge — bcrypt cost-factor tuning, the other major password-hashing family.
• crypto-lab-phantom-vault — PBKDF2-SHA-256 key stretching with HMAC-DRBG in a vault context.
• crypto-lab-shadow-vault — Argon2id plus ChaCha20-Poly1305 for file encryption.
• crypto-lab-mac-race — HMAC and other MACs, the PRF underneath these KDFs.

Accessibility & Testing

KDF Arena is built to a WCAG 2.1 AA standard and verified by an automated audit harness (audit/run.mjs) that drives the real page in Chromium:

• axe-core (WCAG 2.0/2.1 A + AA) on desktop and mobile viewports, in both the initial and post-benchmark states — 0 violations.
• Lighthouse (mobile form factor) — 100 accessibility, 100 best-practices, 100 SEO, 100 performance.
• A screen-reader accessibility-tree check asserting the heading outline, landmark/region names, labelled controls, the skip link, and the named timing meters.

Highlights: logical heading order, a skip link, a real <form> (Enter runs the benchmark), an aria-live results region with aria-busy, 44px touch targets, a 16px input floor to prevent iOS zoom, system-preference theme detection, prefers-reduced-motion support, and AA-compliant contrast in both themes.

npm install
npx playwright install chromium   # one-time, for the audit
npm run audit                     # build + axe + Lighthouse + SR-tree checks

---

One of 120+ browser demos in the Crypto Lab suite.

"So whether you eat or drink or whatever you do, do it all for the glory of God." — 1 Corinthians 10:31