TO UNSUBSCRIBE: email "unsubscribe issforum" in the body of your message to
[EMAIL PROTECTED]  Contact [EMAIL PROTECTED] for help with any problems!
----------------------------------------------------------------------------

ISS,

i have a doubt about two issues topics.

1 - PacketsPerEvent:
Is The value defined at PacketsPerEvent by second or by destination. How can
i know when value is reached ?


Tnks in advance .

Adilson Lemos.



Hello Adilson,

The following article explains how RealSecure uses TCP reset packets to
terminate specified connections.

This information applies to:
RealSecure Network Sensors 3.2 and higher 

Fix Version:
N/A

Related Articles:
None  
  
Solution:
 
When you specify an RSKill as a response for an event in your policy, this
tells the RealSecure Network Sensor to attempt to terminate or reset the
connection responsible for generating the event. The RSKill is the only
RealSecure response capable of going back through the stealth/promiscuous
interface, which makes this response work well whether the NIC is in
"Stealth Mode" or not. 

An RSKill is, simply put, a TCP packet with the RST or reset flag set. The
RealSecure Network Sensor sends two of these for each RSKill. One of these
is sent to the machine originating the connection, with the IP address of
the target spoofed by RealSecure. The other packet is sent to the target
machine, with the IP address of the originating machine also spoofed by
RealSecure. 

What this does, in effect, is to fool both machines into closing the
connection, as if the other machine requested it. A TCP Packet with the RST
flag set is used for this same purpose in normal TCP communications. The
result, is that after initial connection negotiations occur, RealSecure will
terminate any TCP traffic for which this response is specified. 

An RSKill can be used to prohibit unauthorized hosts, or networks from
connecting to services on specified machines, or to enforce the ban of a
service such as Telnet. There are many valid uses for this, and a creative
administrator can make this a valuable tool. If misconfigured, however, an
RSKill can quickly become a nuisance or even a detriment to an otherwise
healthy network.

Care must be taken when setting this up so as not to interfere with
legitimate traffic such as email. A common use of the RSKill is to setup a
UDE (user-defined event) to monitor SMTP traffic for particular text in the
body or header of a message, and when a match is found, to terminate that
connection with an RSKill. This is mainly used to terminate virus-infected
email, and is effective, unless the string happens to be common, i.e. the
infamous "I Love You" VBS worm. 

The result is, legitimate traffic being terminated on the segment that
RealSecure is monitoring, and a local Denial-of-Service attack launched by
the same system that is supposed to be preventing such things. This can also
interfere with mail delivery to other systems, especially if there is an
email relay on the same segment. This is only one example, but there are
several possible ways to misuse the RSKill feature.

As a general rule, it is a good idea not to set the RSKill as a response
unless it is a matter of some importance. Any UDE's should be verified for
proper configuration, and an RSKill should be set as a response, only if it
is absolutely necessary.

As a safety precaution, the RealSecure Network Sensor also embeds what is
known as a "RealSecure Kill Tag" into the packet payload. This can be read
by any other RealSecure Network Sensor that the packet passes. This
identifies the license holder of the Sensor from which the RSKill
originated, and also contains IP information, among other things. This is to
prevent a malicious user from employing this powerful tool as a DoS against
a host or network.

Due to the nature of the TCP protocol, an RSKill cannot be used to terminate
ICMP or UDP-based traffic. Therefore, an RSKill will not work for events
such as Ping Flood, Smurf, UDPBomb, or other, non TCP-based traffic. An
RSKill is also ineffective against portscans, simply because, once the TCP
connection is opened (if it is a full-connect portscan), the portscanner is
moving to another port, and the RSKill sent by the Network Sensor does
nothing. The portscanner would already have the information it needed by the
initial response from the port.

While the RealSecure Kill can be a powerful tool, it is also important to
remember that it is not a replacement for proper firewall configuration or
network security policies. 


How can I reduce the volume of SYNFlood alerts? 

Synopsis:
 
Although the SYNFlood event is non-filterable in RealSecure, the number of
events can be controlled by editing the advanced properties for SYNFlood.

This information applies to:
RealSecure Network Sensor 3.2 and higher only

Fix Version:
N/A

Related Articles:
001120-0007  
  
Solution:
 
Due to differences in network environments, the default installation of the
RealSecure Network Sensor configures the SNYFlood decode's advanced
properties with low values.

To correctly configure the SYNFlood decode advanced properties:

1) Baseline the approximate half-open TCP connections that occur on the
network being monitored by the sensor.

2) In the RealSecure Workgroup Manager, go to Sensor-->Properties-->select
your policy and customize it using the Policy Editor. 

3) On the Security Events tab under the Denial of Service category,
highlight SYNFlood. 

4) On the right side of the screen there is an "Advanced Properties" button.
By selecting that, there will be optional parameters that are configurable.
These parameters are used for "filtering" the SYNFlood events so that the
Workgroup Manager is not flooded with unnecessary events.

Below is a description of what the optional parameters are and their
purpose:

--"HighWaterMark" is the number of SYNs (i.e. half-open TCP connections) to
allow to wait in each port's queue for a response before implementing a
random drop algorithm to free up queue entries. This number should be
smaller than the size of the listen queue for your machines by some
percentage. A guideline for this is 70% of the size of your listen queue,
but you will need to tweak this value to find what works best for your
individual network.

*To configure this value, double click on the number displayed. This is the
event count that RealSecure will until it triggers the first SYNFlood event.

"PacketsPerEvent" is the number of SYNFlood packets seen before a single
event is sent to console/database/etc. This is typically much greater than 1
to prevent a SYNFlood from also causing the sensor to flood the console with
events.

*To configure this value, double click on the number. This is the number
that RealSecure will count until it triggers the next SYNFlood event after
the "HighWaterMark" is reached. All events after that will contain this
number of half-open TCP connections. 




Thank you,

=================================================
GEORGE NAVARRETE
Technical Support Engineer

Internet Security Systems: www.iss.net
Phone:  (404)-236-2700  or  888-447-4861
Technical Support email: [EMAIL PROTECTED]

Training
http://education.iss.net/

Internet Security Systems Product Knowledgebase
http://www.iss.net/customer_care/knowledgebase/

=================================================


-----Original Message-----
From: Adilson Lemos [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, March 06, 2002 1:22 PM
To: ISS Technical Support
Cc: Adilson Lemos
Subject: 522877 SynFlood Configuration


Dear Iss,

We are receiving many attacks of SynFlood in our Site. I would like know if
you will have any documentation about SynFlood Configuration on RealSecure
6.0 ( Network Sensor ). How can i do RSKILL and some advanced configuration
?

Tnks in advance,

Adilson Lemos
Analista de Seguran�a - Globo.com
[EMAIL PROTECTED]
(21) 2233-1022
(21) 7842-7538 - ID 199


Reply via email to