v3 vs v4 naming issue and kdbx v4 support

[Evidlo]

It seems like there is some confusion in the codebase over v3 vs v4 databases. From what I can tell, libkeepass supports KeePass 1.x databases (.kdb) and a subset of KeePass 2.x datbases (.kdbx v3.x only).

The newer kdbx v4 databases have a number of significant changes in supported algorithms and structure of binary data. Here is a summary of the important changes in the spec:

• support for argon2 key derivation
  kdbx v3.x supports AES-KDF only. v4 supports both and defaults to argon2

• key derivation information has been moved

  • transform seed and transform rounds header items have been deprecated
  • a new dynamically sized header item with id 11 has been defined

• stream information for salsa20/chacha20 has been moved

  • a new dynamically sized binary "inner header" has been defined, immediately preceeding the XML data
  • are inner stream key and inner stream id deprecated now? spec didn't say yes

• payload block decryption now uses HMAC-SHA256 instead of just SHA256

• header hash has been moved

  • <HeaderHash> in the decrypted XML has been deprecated
  • a 32 byte SHA256 hash of the header immediately follows the header
  • a 32 byte HMAC-SHA256 hash immediately follows the previous SHA256 hash.
    this hash is calculated like so:

    hmac_key = hashlib.sha512(b'\xff' * 8 + hashlib.sha512(master_seed + transformed_key + b'x01').digest()).digest()
    hmac_hash = hmac.new(hmac_key, header_bytes, hashlib.sha256).digest()

• new CustomData field on Entries and Groups

I've been working on some code that demonstrates decrypting kdbx v3 and kdbx v4 in a step-by-step fashion.

• kdbx v3 decryption
• kdbx v4 decryption

[Evidlo]

pykeepass now supports KDBX4