COMMAND
FW-1
SYSTEMS AFFECTED
Firewall-1 Session Agent 4.1
PROBLEM
Gregory Duchemin found following. After the great revelations at
the Las Vegas black hat about many security vulnerabilities in
FW-1, Gregory was looking at this little module he uses in his
compagny and called "authentication session agent". He uses it
all over the corporate network to allow only some priviledged
users to go into Internet.
This agent is installed on the windows 9.x NT box et just listen
the 261 port for a connexion from a firewall module. When a user
wish to surf on the web or to use any other outside service, the
firewall intercept the request and three handschack the agent to
get some authentication informations: user + pass
There are at least two vulnerabilities in the agent:
1- Denial of service, when a connexion is already established with
the agent, no connexion can be carried anymore leading in a
denial of service and, if one day some malicious users decide
to type something like:
#telnet target 261
User of the target couldn't be able to get the requester asking
him for his password....too bad...no more authentication, no
more outside connection.
2- more seriously, for compatibility reason the agent show a
checkbox that permit our user to send his password in a
cleartext way because firewall modules 4.0 and below don't know
how to do encryption. It's not only possible to sniff this
password on the network segment but much interresting, it 's
really trivial to ask the user agent for giving it to us.
Example:
#nc target 261
220 FW-1 fake session authentication
331 User:
-> he answer with his username
331 *FireWall-1 p4ssw0rd pleazzz:
-> if he's an idiot, he 'll take that for a real fw prompt and u 'll get
back his password else just change the message ;)
200 User has now a clone, c3rb3r
230 OK
Note that this exploit is interactive, when u send 331 User:,
it appears straight away on the victim screen and so u should
have to wait for his answer. It's even possible to use session
agent like a funny chat with a checkpoint logo on the right
top...
The weakness is yet actual when using session agent 4.1 with
"allow clear passwords" option checked (typically for backward
compatibility mode with 4.0 inspection module and below). An IP
wrapper is coded into the agent and then when another ip source is
catched, user is prompted to accept or reject the request, most
users will certainly accept and if they don't, it should be
trivial to spoof firewall ip on the corporate LAN even in a
switched environment with arp game or icmp redirect. If the "Any
ip adress" is checked, things are worse.
A malicious user inside an internal network could be able to use
a nmap like scanner that will look for every open port 261 over
the LAN and use Andrew Danforth's perl script to exploit the flaw.
Spoofing an authorized user ip and using its login/password, our
intruder should be almost invisible in fw logs while accessing
restricted services every versions of agent are vulnerables
(3.0 -> 4.1 ) on win 9.x and NT.
SOLUTION
For the DOS, wait for checkpoint reply but for the password
weakness always use encryption (if you have a firewall module 4.1
naturally ) and use IP wrapper incorporated into the agent but
not effective by default.