Re: Type of encryption is used by Skype.

On Thu, 2 Oct 2003, Kevin McCoy wrote:
> One of  SFs encryption algorithms is Blowfish - a variable length, *448 bit*
> encryption algorithm. By my calculation, Blowfish supports 6.2 x 10 to the
> 57th power more possible keys than does AES.

Beware of a potential oversight.

Just today I was playing with some AES-related code. The standardized
cipher key sizes for Rijndael (the winning AES algorithm) are 128, 192,
and 256 bit, but they can be extended by multiples of 32 bits.

Blowfish implementation in SpeakFreely uses only 128 bits of keyspace.
Another weakness of its scheme is the key generation, which is hashed
passphrase (unless the session key is pre-negotiated and generated
correctly from true random source), which makes the key vulnerable to
bruteforcing. Eg, if you use "-BFfred", the effective key space is only 24
bits (4 characters within 6-bit character space), even if hashed into 128
bits.
IDEA (not the algorithm): We could take least-significant bits from the
microphone input as a strong randomness source.

The announced 448 bits for Blowfish is the MAXIMUM key size possible with
the algorithm. It's very different from the ACTUAL key size used.

> I am a bit suspicious of government sponsored encryption standards.

Not really necessary. The governments have enemies as well. China is a
potential US adversary, with lots of brain and computing power; Russian
cryptanalysts aren't bad too, and they are for hire now.

Even governments have to be secure. If you observe the methods they use to
protect themselves from the most powerful adversaries, together with some
common sense, you may gain decent level of security.

If still in doubts, layer different ciphers with different keys.

Contemporary symmetrical ciphers are difficult to attack, with reasonable
key size (128 is enough, 256 is better, more is overkill). I'd focus
rather to other vulnerabilities, sadly often neglected: key management,
from generation to storage to exchange.

Current cryptosystems are mostly vulnerable not in the algorithms
themselves, but in various implementation details. Skype is not trustful,
as they don't allow to audit the implementation of the algorithms - they
could quite easily leak the keys or use compromised RNG for the key
exchange stage (typical way of compromising cryptosystems - many hardware
units were shipped with tampered-with RNGs on NSA request, it's very
difficult to detect if done well). If you are a target, your machine can
be trojaned and the key gathered from it (keyword: Magic Lantern). Focus
the paranoia in this general direction.

                      * * *

To unsubscribe from this mailing list, send E-mail containing
the word "unsubscribe" in the message body (*not* as the
Subject) to [email protected]