Next: PAK-R
Up: No Title
Previous: Naive Password-Authenticated Diffie-Hellman
We start with a simplified version of the PAK protocol.
The idea is to do an authenticated Diffie-Hellman key exchange,
but entangle a hash of the password with the first Diffie-Hellman
value sent by Alice.
Figure 4:
Simplified PAK protocol
 |
Details:
- p is a 1024-bit prime, and
all exponentiations and multiplications are performed mod p.
- g is a generator of a subgroup of
of size q
where q is a 160-bit prime that divides p-1.
(This means most exponents only need to be 160-bits
instead of 1024 bits.)
- H1() is a hash function that outputs 1024+160 bits, and
thus
will be almost uniformly random in
. - H2a(), H2b(), and H3 are hash functions that
output 160 bits each.
The intuitive reason this protocol
works is that Bob can only guess one password
at a time, since Bob sends his authentication value (k) first,
while Alice can only guess one password at a time since she has
entangled her value X with the password, and cannot compute
a valid Z for any other password (by the hardness of the
Diffie-Hellman problem).
However, there are some questions you may ask, like:
- 1.
- What if Alice sends a bogus X value, like zero?
- 2.
- What if the output of
is not in the subgroup generated
by g? Doesn't that lead to subgroup confinement attacks?
To solve these problems, we must enhance the protocol, as shown
in the next section.
Next: PAK-R
Up: No Title
Previous: Naive Password-Authenticated Diffie-Hellman
Phil MacKenzie
9/29/2000