[ pycrypto-Bugs-1967068 ] Blowfish differs from others

Bugs item #1967068, was opened at 2008-05-19 06:03
Message generated for change (Comment added) made by nobody
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=120937&aid=1967068&group_id=20937

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
Summary: Blowfish differs from others

Initial Comment:
Blowfish in pycrypto produces different results that many other implementations, e.g. Java's JCE, Bruce Schneier's reference implementation, crypto etc.

Reproduce by doing the following encryption in different languages:
binascii.hexlify(Blowfish.new("1234567890123456").encrypt("12345678"))

PyCrypto outputs:
61d2570dc6e09632

Java, Schneier's C implementation, openssl output:
e00723bbb58234aa

In both cases it was same little-endian machine.
[email protected]

----------------------------------------------------------------------

Comment By: Nobody/Anonymous (nobody)
Date: 2008-05-20 03:52

Message:
Logged In: NO 

I wish PyCrypto can pass Bruce Shneier's Blowfish test suite!
Get it from his website
http://www.schneier.com/code/vectors.txt

E.g. the following
print binascii.hexlify(Blowfish.new(binascii.unhexlify(
    'FFFFFFFFFFFFFFFF'
    ), Blowfish.MODE_ECB).encrypt(binascii.unhexlify(
    'FFFFFFFFFFFFFFFF'
    )))
must output 51866FD5B85ECB8A, not 63c6e3f875f553b3 like 2.0.1 does. 

That will make the library interoperable with other standard
implementations. In current state (2.0.1) blowfish is mostly useless, you
can't write a Python client to network server which uses Blowfish for
traffic encryption etc.

----------------------------------------------------------------------

Comment By: Nobody/Anonymous (nobody)
Date: 2008-05-20 00:27

Message:
Logged In: NO 

Looks like problem is in private function used

PyCrypto's:
#define bf_round(l,r,n) \
       l ^= P[n]; \
       r ^= ( (sub(S[0],l>>22 & 0x3fc) + sub(S[1],l>>14 & 0x3fc)) \
          ^ sub(S[2],l>>6 & 0x3fc) ) +S[3][l & 0xff] 

Schneier's is different:
#define S(x,i) (SBoxes[i][x.w.byte##i])
#define bf_F(x) (((S(x,0) + S(x,1)) ^ S(x,2)) + S(x,3))
#define ROUND(a,b,n) (a.dword ^= bf_F(b) ^ PArray[n])

Java's is the same different:
    private int func(int x)
    {
        return (((S0[(x >>> 24)] + S1[(x >>> 16) & 0xff]) ^ S2[(x >>> 8) &
0xff]) + S3[x & 0xff]);
    }


----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=120937&aid=1967068&group_id=20937

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft 
Defy all challenges. Microsoft(R) Visual Studio 2008. 
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/