gwashburn

Also, to open the database the TwofishCipher plugin with my patch to not handle padding must be used or a padding exception will be raised.

EDIT: My analysis was wrong for the KDBX 4.x format, and this does not fix the issue for that format. By chance the 4.x database I tested on did not trigger the issue. There is an EOS for the HmacBlockStream used by 4.x, but, unlike 3.x where the HashedBlockStream contains the XML, the HmacBlockStream contains the CipherStream, which contains the XML. So indeed when decrypting with no padding, the padding will get appended to the XML data. The only way that I can see around this is to have the parser...

Do not read past ending KeePassFile tag in xml stream

Until this gets fixed upstream, I wrote in another thread how to resolve this , provide a patch, and an updated TwofishCipher plugin that fixes this.

I'm now running into this issue. There's actually no need for the compatibility issues. The current compatibility issues arise because v1.3 pads the cipher stream with NULL bytes, but v1.4 expects it to have PKCS7-style padding. TwofishCipher has always expected to read the same padding style as the one it writes. So currently a kdbx written with the v1.3 will have a padding of NULLs and if read using the v1.4 plugin an exception will be thrown because v1.4 expects a PKCS7-style padding. However,...

KdbxFile:CloseStreams swallows exceptions thrown by streams

Looking into this further, what's happening is that the encryption plugin is throwing an exception based on the incorrect padding and the exception is being swallowed in KdbxFile:CloseStreams, which is why the actual exception bubbled up to the UI is the xml parse error. Since we now know that the problem is originating from an exception in the TwofishCipher plugin due to incorrect padding, then I agree there's no change in KeePass that will allow KeePass to read the database. So this feature request...

Hi Dominik, first, thanks for all the work on KeePass2. I understand that padding is handled by the encryption stream, ie the Twofish plugin. What I'm not asking is for KeePass2 to do anything with the padding. The real issue here is that KeePass seems to not handle garbage at the end of the HashedBlockStream, which in this case are 1-15 NULL bytes. What I'm saying is that for KDBX 3.1 , there is no reason for trailing garbage in the HashedBlockStream to affect anything, and yet it currently does....