Hash Generator

What people actually use hashes for

Hashes are a "one-line answer" tool to a surprisingly large set of unrelated problems. The five most common uses, in roughly the order most engineers encounter them:

1. File integrity. "Did this 4 GB file arrive intact?" Hash the file, compare against a published digest. The classic SHA-256 use case — same algorithm used in shasum, sha256sum, every CDN's checksum manifest, and TLS certificate pinning.
2. Cache keys. "Have I computed the result for these inputs before?" Hash the inputs, use the digest as a Redis/Memcached key. MD5 is fine here — collisions don't matter because the values can be recomputed.
3. Deduplication. "Are these two large blobs identical?" Same as integrity but inverted: compare hashes instead of bytes. Git's content-addressed storage is the canonical example (originally SHA-1, now transitioning to SHA-256).
4. Digital signatures and JWT. Sign a hash of the document, not the document itself, then verify the signature against a fresh hash. The JWT decoder and JWT generator use this pattern; the HMAC generator is the keyed variant.
5. Password storage. But not with this tool. SHA-256 is the wrong algorithm for password storage. See the section below before you do this.

Picking the right algorithm

Five algorithms are available above. A pragmatic decision table:

• SHA-256 (256-bit, 64 hex chars): the default. Use it unless you have a reason not to. Used by TLS, Bitcoin, Git (as of 2.29+), most code-signing systems.
• SHA-512 (512-bit, 128 hex chars): same security level as SHA-256, but uses 64-bit operations internally. Slightly faster on 64-bit CPUs (counterintuitively, given the longer output), slightly slower on 32-bit. Pick when you specifically need 256 bits of collision resistance instead of 128 (i.e., basically never for typical workloads).
• SHA-384 (384-bit, 96 hex chars): SHA-512 truncated. Used by some US government suites (Suite B). Rare outside those contexts.
• SHA-1 (160-bit, 40 hex chars): broken for collision resistance since 2017 (the SHAttered attack produced a real PDF collision). Still safe for HMAC and for non-adversarial integrity checks — but for new use cases, just use SHA-256.
• MD5 (128-bit, 32 hex chars): broken since 2004. Trivially collidable on a laptop. Still acceptable for cache keys, file fingerprinting where adversaries are not involved, and legacy compatibility. Never for anything that touches security.

The mistake to never make: SHA-256 for passwords

This tool can compute SHA-256 of any input, including a password. That does not mean you should store user passwords this way. SHA-256 is designed to be fast — billions of operations per second on a modern GPU. That speed is a feature for integrity checking. It is a catastrophic flaw for password storage. If a SHA-256-hashed password database leaks, an attacker can test the entire 8-character English-words dictionary against every account in minutes.

Use a password-specific hash: bcrypt, scrypt, or Argon2id (the current preferred default per OWASP). These are intentionally slow and memory-hard — they take roughly 100ms per hash on a modern CPU instead of nanoseconds, and on a GPU they are not meaningfully faster. That speed asymmetry is the entire point. The password generator creates strong inputs; the language-level libraries (bcrypt-ruby, argon2-cffi, etc.) do the actual storage.

Collision history, in 30 seconds

• MD5 — collision attack demonstrated 2004 (Wang et al.). Practical chosen-prefix collisions 2009 (Stevens et al.). A rogue intermediate CA certificate was forged using MD5 collisions in 2008. Effectively broken for two decades.
• SHA-1 — theoretical attacks since 2005. Real collision (the SHAttered PDF) produced 2017 by Google + CWI Amsterdam at a cost of ~$110k in cloud compute. Chosen-prefix collisions in 2020 at ~$45k. Officially deprecated by major CAs.
• SHA-256 — no practical attacks. Best known weakness is theoretical: in 2024, researchers found differential characteristics in 31 of 64 rounds. The full algorithm remains secure with no plausible cryptanalytic path forward. Conservative estimate: 10-20 more years of safety, longer if quantum computers do not materialize at scale.
• SHA-3 (not in this tool) — a completely different design (Keccak sponge construction) standardized in 2015 as a hedge against SHA-2 ever being broken. Use it if you need diversity in your cryptographic supply chain; SHA-256 is otherwise still the default for new systems.

The avalanche property, demonstrated

Type two near-identical inputs into the calculator above — hello and Hello, say — and compare the SHA-256 outputs. Despite the inputs differing by a single bit (the case of the first letter), the outputs share essentially nothing. About half of the 256 output bits flip with any single-bit input change. This is called the avalanche property and is the reason hashes work as fingerprints: even if you had a million similar inputs, you could not recognize which two were closely related from their digests alone.