mirror of
https://github.com/telekom-security/tpotce.git
synced 2025-11-03 05:52:52 +00:00
30 lines
1.4 KiB
Text
30 lines
1.4 KiB
Text
|
-----------------------------
|
||
|
|
Some random food for thought:
|
||
|
|
-----------------------------
|
||
|
|
|
||
|
|
1) If you run p0f on any reasonably popular server, you will probably see quite
|
||
|
|
a few systems that seem to be leaking memory in TCP headers (e.g. ACK number
|
||
|
|
or second timestamp set on SYN packets, URG pointer without URG flag, etc).
|
||
|
|
You will also see HTTP traffic with non-stripped Proxy-Authorization headers
|
||
|
|
and other hilarious abnormalities.
|
||
|
|
|
||
|
|
Unfortunately, pinpointing the sources of many of these leaks is pretty hard;
|
||
|
|
they often trace to proprietary corporate proxies and firewalls, and unless
|
||
|
|
it's *your* proxy or firewall, you won't be finding out more. If you wish to
|
||
|
|
put some investigative effort into this, there are quite a few bugs waiting
|
||
|
|
to be tracked down, though :-)
|
||
|
|
|
||
|
|
2) After some hesitation, I decided *against* the inclusion of encrypted traffic
|
||
|
|
classification features into p0f. Timing, packet size, and direction
|
||
|
|
information lets you, for example, reliably differentiate between interactive
|
||
|
|
SSH sessions and SFTP uploads or downloads; automated and human password
|
||
|
|
entry attemps; or failed and successful auth.
|
||
|
|
|
||
|
|
The same goes for SSL: you can tell normal HTTPS browsing from file uploads,
|
||
|
|
from attempts to smuggle, say, PPP over SSL. In the end, however, it seems
|
||
|
|
like stretch to cram it into p0f; one day, I might improve my ancient 'fl0p'
|
||
|
|
tool, instead:
|
||
|
|
|
||
|
|
http://lcamtuf.coredump.cx/soft/fl0p-devel.tgz
|
||
|
|
|