Question

I have several strings that were encrypted using OpenSSL. For instance:

$ echo "original string" | openssl aes-256-cbc -p -a -pass pass:secret
salt=B898FE40EC8155FD
key=4899E518743EB0584B0811AE559ED8AD9F0B5FA31B0B998FEB8453B8E3A7B36C
iv =EFA6105F30F6C462B3D135725A6E1618
U2FsdGVkX1+4mP5A7IFV/VcgRs4ci/yupMErHjf5bkT5XrcowXK7z3VyyV1l2jvy

I would like to decrypt these things using Python. I’m attempting to use PyCrypto. Here’s an exmaple script using the above data:

from base64 import b64decode, b64encode
from hashlib import md5
from Crypto.Cipher import AES

secret = 'secret'
encoded = 'U2FsdGVkX1+4mP5A7IFV/VcgRs4ci/yupMErHjf5bkT5XrcowXK7z3VyyV1l2jvy'
encrypted = b64decode(encoded)
salt = encrypted[8:16]
data = encrypted[16:]
key = md5(secret + salt).hexdigest()
iv = md5(key + secret + salt).hexdigest()[0:16] # which 16 bytes?
dec = AES.new(key, AES.MODE_CBC, iv)
clear = dec.decrypt(data)

try:
    salt_hex = ''.join(["%X" % ord(c) for c in salt])
    print 'salt:     %s' % salt_hex
    print 'expected: %s' % 'B898FE40EC8155FD'
    print 'key:      %s' % key.upper()
    print 'expected: %s' % '4899E518743EB0584B0811AE559ED8AD9F0B5FA31B0B998FEB8453B8E3A7B36C'
    print 'iv:       %s' % iv
    print 'expected: %s' % 'EFA6105F30F6C462B3D135725A6E1618'
    print 'result: %s' % clear
except UnicodeDecodeError:
    print 'decryption failed'

Here’s the output:

salt:     B898FE40EC8155FD
expected: B898FE40EC8155FD
key:      4899E518743EB0584B0811AE559ED8AD
expected: 4899E518743EB0584B0811AE559ED8AD9F0B5FA31B0B998FEB8453B8E3A7B36C
iv:       17988376b72f4a81
expected: EFA6105F30F6C462B3D135725A6E1618
decryption failed

You can see that the salt matches, and the key matches the first half of what OpenSSL shows, so I seem to be on the right track, but there are two main questions:

1. Why are the values for key and iv from OpenSSL twice as long as PyCrypto (and presumably AES256) allows?
2. How do I generate the correct values? The technique I’m using was taken from a blog, but if the IV is always supposed to match the block size (16 bytes), MD5 will never work. And even if I could figure out where the other half of the key comes from, PyCrypto would refuse it for being too long.

I realize I’ll need to remove the padding as well, but I left that out for brevity.

Answer 1

You have three problems:

1. You use AES256 (32 byte key) in OpenSSL and AES128 (16 byte key) in your python code.
2. The IV computation is wrong. Each step in the OpenSSL’s key derivation function uses the the MD5 digest computed last.
3. You mix up binary and hexadecimal representation. Keep any conversion to hexadecimal as the last step, before visualization.

The following code should be correct:

from base64 import b64decode, b64encode
from binascii import hexlify
from Crypto.Cipher import AES
from Crypto.Hash import MD5

secret = 'secret'
encoded = 'U2FsdGVkX1+4mP5A7IFV/VcgRs4ci/yupMErHjf5bkT5XrcowXK7z3VyyV1l2jvy'
encrypted = b64decode(encoded)
salt = encrypted[8:16]
data = encrypted[16:]

# We need 32 bytes for the AES key, and 16 bytes for the IV
def openssl_kdf(req):
    prev = ''
    while req>0:
        prev = MD5.new(prev+secret+salt).digest()
        req -= 16
        yield prev
mat = ''.join([ x for x in openssl_kdf(32+16) ])
key = mat[0:32]
iv  = mat[32:48]

dec = AES.new(key, AES.MODE_CBC, iv)
clear = dec.decrypt(data)

try:
    salt_hex = ''.join(["%X" % ord(c) for c in salt])
    print 'salt:     %s' % salt_hex
    print 'expected: %s' % 'B898FE40EC8155FD'
    print 'key:      %s' % hexlify(key).upper()
    print 'expected: %s' % '4899E518743EB0584B0811AE559ED8AD9F0B5FA31B0B998FEB8453B8E3A7B36C'
    print 'iv:       %s' % hexlify(iv).upper()
    print 'expected: %s' % 'EFA6105F30F6C462B3D135725A6E1618'
    print 'result:   %s' % clear
except UnicodeDecodeError:
    print 'decryption failed'