On Thu Sep 21 2006 at 15:57, Douglas Nordwall wrote:

> While i know this to be true (although I was not aware that
> safe_checks slowed things down, I may have missed mention of it, but
> I don't remember seeing it)

nessus_tcp_scanner computes two parameters:
min_cnx = 8 * MAX(max_checks, 5);
if (safe_checks)  max_cnx = 24 * MAX(max_checks, 5);
else              max_cnx = 80 * MAX(max_checks, 5);
Those basic values may be lowered if the machine is loaded or short of
file descriptors.
min_cnx is used to initialize open_sock_max which is the number of
open parallel connections: open_sock_max = min_cnx / (pass + 1);
open_sock_max is then adjusted during the scan; it never goes higher
than max_cnx.

The number of probes per second is *never* adjusted. But you can
estimate it from open_sock_max and the RTT  (roughly < open_sock_max / rtt)

For example, if you set max_checs to 5 and safe_checks:
min_cnx=40; max_cnx = 120;
During the first pass, nessus_tcp_scanner will first try to open 40
sockets at once. If the target answers well, the number of parallel
connections will increase but never go above 120.
If there is a second pass, nessus_tcp_scanner will first open 20
sockets at once
...
On the beginning of the tenth pass (definitely a problem with the
target), it will open 4 sockets at once. 

> I needed very slow scans (on the order of no more than 5 ports in 5
> minutes) to bypass the countermeasures.

This means that you'll need 65535 min = 1092.25 h = 45.5 days to run a
full port scan. Is this acceptable? I don't think so.

> Maybe ask Fyodor what they did to compensate for the problems you are
> concerned about?

nmap randomizes the port numbers by default. We suspect that this
defeats crude countermeasures.

> fair enough, and for your network that is fine.

I nearly never scan *my* network. My experience comes from different
networks on my customers' sites. Many of them are a mix of quick LAN,
overloaded links or routers, slow leased lines or high latency
satellite based WAN. Scans often runs more than a day and the network
load is concentrated during the working hours (or at least, the nature
of the trafic changes in the evening). That's why I needed an
adaptative tool! 
It would be great if this scanner could also suit your needs, but I
am afraid that your requirements are too strict.

> I know fairly well, and have access to exactly what's happening on
> it, I can make those determinations in other ways.

Maybe Nessus SYN scanner would be better for you? It is slower but
its behaviour is more regular, IIRC.

> You asked why people prefer to use nmap instead of the built in.

I see. I am still surprised by your figures.

-- 
http://arboi.da.ru/                     http://ma75.blogspot.com/
PGP key ID : 0x0BBABA91 - 0x1320924F0BBABA91
Fingerprint: 1048 B09B EEAF 20AA F645  2E1A 1320 924F 0BBA BA91
_______________________________________________
Nessus mailing list
[email protected]
http://mail.nessus.org/mailman/listinfo/nessus

Reply via email to