RollYourOwnCrypto [anti-pattern] v0.1.0

Implementing a cipher, hash, MAC, or 'lightweight encryption scheme' from scratch instead of using a vetted library. Includes XOR-with-a-key 'encryption', rolling your own JWT-style token format, hashing-then-truncating for an integrity tag, or using crypto primitives in modes you didn't see in a paper.

Label

Inventing or reimplementing cryptographic primitives

Why Bad

Cryptography fails silently. A scheme can look fine, encrypt and decrypt correctly, and still be trivially broken — padding oracles, length-extension, nonce reuse, timing leaks, weak randomness, or simply choosing ECB. The history of applied crypto is the history of clever people getting it wrong; you will too.

Instead Do

    Use a high-level, audited library that hides the primitives:

      • libsodium (crypto_box, crypto_secretbox, crypto_pwhash)
      • Google Tink (envelope encryption, deterministic AEAD)
      • age / rage (file encryption)
      • ASP.NET Core Data Protection, Java Tink, etc.

    If you must use a primitive directly, use the constructor your
    standard library provides and copy a tested example verbatim. Do
    not invent. Do not 'optimise' the example. Do not skip the MAC.
  

RollYourOwnCrypto [anti-pattern] v0.1.0

Implementing a cipher, hash, MAC, or 'lightweight encryption scheme' from scratch instead of using a vetted library. Includes XOR-with-a-key 'encryption', rolling your own JWT-style token format, hashing-then-truncating for an integrity tag, or using crypto primitives in modes you didn't see in a paper.

Label

Inventing or reimplementing cryptographic primitives

Why Bad

Cryptography fails silently. A scheme can look fine, encrypt and decrypt correctly, and still be trivially broken — padding oracles, length-extension, nonce reuse, timing leaks, weak randomness, or simply choosing ECB. The history of applied crypto is the history of clever people getting it wrong; you will too.

Instead Do

    Use a high-level, audited library that hides the primitives:

      • libsodium (crypto_box, crypto_secretbox, crypto_pwhash)
      • Google Tink (envelope encryption, deterministic AEAD)
      • age / rage (file encryption)
      • ASP.NET Core Data Protection, Java Tink, etc.

    If you must use a primitive directly, use the constructor your
    standard library provides and copy a tested example verbatim. Do
    not invent. Do not 'optimise' the example. Do not skip the MAC.
  

Label

Inventing or reimplementing cryptographic primitives

Why Bad

Cryptography fails silently. A scheme can look fine, encrypt and decrypt correctly, and still be trivially broken — padding oracles, length-extension, nonce reuse, timing leaks, weak randomness, or simply choosing ECB. The history of applied crypto is the history of clever people getting it wrong; you will too.

Instead Do

    Use a high-level, audited library that hides the primitives:

      • libsodium (crypto_box, crypto_secretbox, crypto_pwhash)
      • Google Tink (envelope encryption, deterministic AEAD)
      • age / rage (file encryption)
      • ASP.NET Core Data Protection, Java Tink, etc.

    If you must use a primitive directly, use the constructor your
    standard library provides and copy a tested example verbatim. Do
    not invent. Do not 'optimise' the example. Do not skip the MAC.