COMMAND
ISN generation
SYSTEMS AFFECTED
Ascend MAX
PROBLEM
For those that don't know, the MAX is Ascend's most popular WAN
dialup server that is widely used among mid-sized ISPs. It is not
my idea to put stuff about SN on these pages, but what Marc Slemko
found is really out of mind.
Marc decided to take a look at the ISN (sequence number)
generation on the MAX. And results were surprise:
time isn difference
========= ========== ==========
17.979213 1024896036
19.928152 1025024036 128000
21.848087 1025152036 128000
23.648915 1025280036 128000
25.221257 1025408036 128000
26.750434 1025536036 128000
28.230257 1025664036 128000
29.692895 1025792036 128000
31.172050 1025920036 128000
32.534981 1026048036 128000
33.955213 1026176036 128000
35.424979 1026304036 128000
36.856233 1026432036 128000
38.295180 1026560036 128000
That is pathetic. In certain limited situations, this renders
packet filtering useless WRT connections to the MAX itself. While
it presents no risk to traffic passing through the MAX, it does
potentially expose connections to (and actually, also connections
being made from) the MAX.
Note that this violates section 4.2.2.9 of RFC-1122 in addition to
simply being a stupid thing to do from a security standpoint.
There must be half a dozen freely available examples of how to do
this in a TCP stack. The method suggested in RFC-793 is
inadequate for obvious reasons, but even it is 100 times better
than this.
SOLUTION
Write to Ascend and ask them for health. In meantime, use other
dialup server.