In this article
Hash the word hello with SHA-256 and you get 2cf24dba5fb0a30e26e83b2ac5b9e29e1b161e5c1fa7425e73043362938b9824. Change the last letter to make hellp and you get fdd7585e08c4e2af…. Not similar. Not nearby. Completely different. That is not a bug. It is the point.
A hash function takes any amount of data and boils it down to a fixed-length fingerprint. Same input always gives the same fingerprint. A different input, even by one character, gives a wildly different one. Once you accept those two rules, a lot of everyday technology makes sense: file checks, version control, password storage and more.
Three promises a good hash makes
- Same input, same output. Hash the same text a million times and you get the same fingerprint a million times, on any computer.
- Fixed size. A single letter or a ten-gigabyte video both give you, for SHA-256, exactly 64 hexadecimal characters.
- One way. You can go from the text to the fingerprint easily. Going back from the fingerprint to the text is, for practical purposes, impossible.
Whatever you feed in, the answer has exactly this shape.
One letter, a completely different fingerprint
Look at three tiny inputs that differ by one character, hashed with SHA-256. Only the beginning of each fingerprint is shown, but you will see that they have nothing in common.
| Input | SHA-256 (first 16 characters) |
|---|---|
hello | 2cf24dba5fb0a30e |
hellp | fdd7585e08c4e2af |
Hello | 185f8db32271fe25 |
Changing e→p, or h→H, produces fingerprints with no resemblance at all. There is no “close” in the output, even when the input is very close.
This is called the avalanche effect. If you compare the full 256 bits of hello against hellp, 131 of them differ. Comparing hello and Hello, 125 differ. That is about half, which is exactly what you would expect from two unrelated random strings.
- Bits that changed131
- Bits that stayed125
About half the bits flip, no matter how small the change. That is what stops anyone reading a pattern in the output.
What hashes are used for
- Checking a download. A site publishes the SHA-256 of a file. You hash your copy. If the fingerprints match, the file was not corrupted or tampered with on the way.
- Finding duplicates. Two files with the same hash are, for all practical purposes, identical, so you can find copies without comparing them byte by byte.
- Version control. Git identifies every commit by a hash of its contents, so history cannot be quietly rewritten without every later fingerprint changing.
- Digital signatures and certificates. Signing a long document is slow, so you hash it and sign the short fingerprint instead.
- Storing passwords. A site should keep a hash of your password, not the password. But read the warning below, because the type of hash matters a lot.
Hashing is not encryption
Encryption is meant to be reversed with a key. A hash is meant to not be reversed at all, and there is no key. You cannot “decrypt” a SHA-256 fingerprint. What people do instead is guess: hash many candidate inputs and see which one matches.
Which algorithm, and which are retired
Longer output means more room for the fingerprint to be unique, but length is not the only thing that decides safety.
- MD5 and SHA-1 have known collision attacks: researchers have built two different inputs that give the same fingerprint. Do not use them where an attacker is involved, such as signatures. They are still fine for catching accidental corruption.
- SHA-256 and SHA-512 are the sensible defaults today for integrity checks and signatures.
- Larger is not always better. SHA-512 is not “safer” than SHA-256 for most uses. It is just longer, and sometimes faster on 64-bit machines.
A small trap: invisible differences
If you hash text and your friend hashes the “same” text but gets a different answer, look for invisible differences: a trailing space, a newline at the end, a different capital letter, or text saved in a different encoding. A hash sees every byte. Anything a human would ignore, the function counts.
Questions people ask
What is a hash?
A fixed-length fingerprint of some data produced by a hash function. The same input always gives the same output, and a tiny change to the input gives a completely different output.
Is hashing the same as encryption?
No. Encryption can be reversed with a key. A hash is one-way and has no key, so you cannot decrypt it. You can only guess inputs and compare their hashes.
Can a SHA-256 hash be reversed?
Not by calculation. People recover short or common inputs by guessing and hashing many candidates, which is why weak passwords are still at risk even when hashed.
Should I use SHA-256 to store passwords?
No. It is too fast, which helps attackers. Use a slow, salted password hash such as bcrypt, scrypt or Argon2 instead.
Why do MD5 and SHA-1 count as broken?
Because researchers have shown practical ways to create two different inputs with the same hash. That is unsafe for signatures and security checks, though fine for catching accidental file corruption.