Use stronger encryption instead of Blowfish

[tsileo]

Use stronger encryption instead of Blowfish

[tsileo]

I totally agree with you, I'm working on a bakthat rewrite (not backward compatible, with a lot of new features) and I'm already planning to drop beefish.

Months ago, I wanted to give a try to Elliptic Curve Crypto (and with https://github.com/bwesterb/py-seccure), but according to http://safecurves.cr.yp.to/, I shouldn't use any NIST curves so, I think I will drop this idea too.

I will look further into AES-256/AES-128, I will keep you updated.

Do you have a lot of experiences with pycrypto / cryptography ?

Thanks!

[mxsasha]

I have some knowledge of cryptography, but focused more around the environment, i.e. how to use it securely, than details of specific ciphers. I know very little about elliptic curve crypto.

For AES-256, you could have a look at https://github.com/rnapier/RNCryptor/tree/master/python - that seems to be the most complete library out there. It also takes care of PBKDF2 and HMACs, and packages the data, hmac, salts and IV all into a nice data format, so it takes care of the complete package. As a plus, there are implementations in a few different languages too. It is not on PyPI currently though. You could also adopt it into bakthat itself. Note that there is a (very difficult to exploit) timing attack still open: RNCryptor/RNCryptor#88

[tsileo]

Thanks for the details, and for the AES-256 reference implementation, but the RNCryptor as is can't handle file that doesn't fit in ram, It should read by chunks and deals with padding (like this snippet), maybe I can try to do it myself if you can validate what I'm doing.

[mxsasha]

Well, I can't make any promises, but I imagine we can at least make it safer than with blowfish without PBKDF2/bcrypt. What would be even nicer is if we could integrate the support you need into RNCryptor itself, so that others can reuse it in the future.

[iluminite]

@tsileo, I'm no crypto guru, but it sounds like you might like https://github.com/jedisct1/libsodium, there are two python libraries, https://github.com/stef/pysodium and https://github.com/dstufft/pynacl. have you seen these?