encfs_aes_getpass.py generates short passwords sometimes with a new lines about python-trezor

[aussiehash]

I've modified lines
85
trezor_entropy = client.get_entropy(32).encode('hex')
86
urandom_entropy = os.urandom(32).encode('hex')
110
print passw.encode('hex')

Although perhaps only the 3rd edit is required.

Please confirm action on your device. 28bee36480aaf1dec7137fbdc0b0e5d0cfde58295f1f47cdc63dfa8af695fbbb

Edit : base64 encoding is probably a better choice
print passw.encode('base64')

[ddworken]

This is not actually a vulnerability. Rather than looking at the output, look at the code to determine if this is secure.

trezor_entropy = client.get_entropy(32)
urandom_entropy = os.urandom(32)
passw = hashlib.sha256(trezor_entropy + urandom_entropy).digest()

So the fact that passw is often short, does not affect the security of this. By examining the code one can see that this is made up of combining two sources of entropy, both 32 bytes of randomness, and hashing them. The length of the output is irrelevant. There is plenty of randomness in this case.

The perceived vulnerability is due to the output of passw not being stored as as base64 encoded string. It is not actually made up of question marks, those are due to your computer not understanding how to interpret the data as a string.

Encoding passw in base64 or hex does not improve the security of anything. Instead it improves the perceived security of it if you look at the output. But for anyone trying to brute force it, this will not change anything. It would just forced them to generate random values that are base64 encoded instead.

[prusnak]

Right. Autogenerated password is binary.