-------[  Phrack Magazine --- Vol. 9 | Issue 55 --- 09.09.99 --- 10 of 19  ]


-------------------------[  Building Bastion Routers Using Cisco IOS  ]


--------[   brett <beldridg@best.com> / variable k <variablek@home.com>  ]


----[  Abstract

Members of the firewall and network security community are generally clueful
when it comes to the topic of bastion hosts, and the various approaches and
issues involved in constructing them on different platforms.  However, less
attention has been paid to the subject of securing routers that are exposed to
attack--or building bastion routers.

Routers, and in particular Cisco routers, are often deployed in various parts
of a firewall system, for example as border and choke packet filters.  As such,
they can be high-value targets for attackers.  This paper provides a simple
methodology and specific examples for securing Cisco routers running IOS.  Our
focus and examples are based upon the IOS versions we are most familiar with:
11.2 and 11.3.  However, the principles we present may also apply to older and
newer IOS versions (e.g., 12.0, other 11.X versions and 10.X), and possibly to
other vendors' gear.


----[  What is a Bastion Router Anyway?

Routers previously did just that: route IP.   However, modern routers have
features that permit them to be used as static packet screens, security (VPN)
gateways, and other key components in security systems.  There is even an IOS
variant called the Firewall Feature Set (this is different than the PIX
firewall), which we don't cover here because we haven't used it, that supports
stateful packet filtering, intrusion detection features and other stuff.  We
use the term bastion [0] router to refer to a router that requires some level
of special configuration to secure it against attack.

We generally focus on two areas: protecting the router itself and protecting
hosts behind the router (or possibly on other sides).


---[ Basic Methodology ]---

Our methodology is relatively simple.  We want to disable features and
services that are on by default and that we are not using.  In other words: if
we're not using something, we turn it off.  We enable features that may aid in
protecting the router or the networks behind the router.  If we need a feature
we try to protect it as best we can using the protection mechanisms that IOS
provides, for example VTY filters.  We use ACLs on each interface that permit
the specific traffic that we have decided to permit and deny everything else
(the "default deny" stance).

IOS supports many, many features; and there are many different releases and a
number of feature sets available.  Our examples assume IOS version 11.2 and
11.3, with the IP Only feature set, though we will point out exceptions (e.g.,
TCP Intercept and the Enterprise feature set) as they come up.  Also, we can't
possibly cover all the various ways to configure something-- our goal is to
present some of the things we've learned and some of the methods by which
we configure bastion routers.

So the basic methodology we will follow is:

  1. Password protection
  2. Limit remote access
  3. Limit local access
  4. Display login banner
  5. Configure SNMP
  6. Configure logging and NTP
  7. Other protection mechanisms
  8. Anti-spoofing
  9. Mitigate Denial of Service attacks
 10. Protect hosts behind the router
 11. Verify the configuration
       
For purposes of the examples, we will use a sample network with the following
topology. We will also assume that 192.168/16 is routable.

     Eth0  192.168.0.0/16            Eth1  172.16.1.0/30
                      .1 +----------+ .1     .2
   private net ----------|  Router  |---------- ISP
                         +----------+
     access-list e0-in -->          <-- access-list e1-in

The final complete configuration will be given at the end of the paper in
Appendix A.


----[ Background
  
Brief Introduction to the IOS Command Line Interface
----------------------------------------------------

Cisco's IOS (Internetworking Operating System) thankfully supports a Command
Line Interface which Cisco calls CLI. The command line interface can be
accessed via the console port, a modem, or a TELNET connection.  A command
line session is referred to as an EXEC session, and it's similar to a Unix
shell process.  There are two different kinds of traditional EXEC sessions:
user EXEC level and privileged EXEC level.  User EXEC level can be considered
similar to a non-root login account on a Unix system, and privileged EXEC
level somewhat like the super-user account, or a UID 0 process.  The prompt
even changes to end in a pound sign when you switch to privileged EXEC
level:
   
  reeb>enable
  Password:
  reeb#
  reeb#disable
  reeb>

You can also customize privilege levels.  We'll cover this a bit more later on.

Context sensitive help is also available. Typing a question mark will provide
a list of available commands and options that may be entered at that point.
For example,
   
  reeb#debug ip r?
  rip  routing  rsvp  rtp

  reeb#debug ip rip ?
  events  RIP protocol events

  reeb#debug ip rip

CLI also supports a mini Emacs-like editing mode and command history by
default.  So you have C-n for next line, C-p for previous line, C-a for
beginning of line, C-e for end of line, C-u to erase the line, C-w for erase
previous word, and also TAB to finish a partial command.  The arrow keys
should also work.
   
Configuration Settings
----------------------

One of the things that can be very confusing with IOS is how configuration
settings are presented to the user.  A default setting is not displayed when
you view the router configuration.  And the default setting can change across
different IOS versions.  For example, in IOS 11.2, the services
`udp-small-servers` and `tcp-small-servers` are enabled by default.  So when
you disable UDP and TCP small servers you will see the following in the
configuration:
   
  version 11.2
  no service udp-small-servers
  no service tcp-small-servers

And by default you would see no configuration setting.  However, the defaults
changed in 11.3 to "no service" for both.  So when no configuration setting is
displayed, UDP and TCP small servers are disabled.  You will see the following
when they are enabled:

  version 11.3
  service udp-small-servers
  service tcp-small-servers

You need to keep this in mind when building bastion Cisco's, and it may take
some investigation and detective work to determine which services and features
are enabled.

----[  Step 1 : Password protection

One of the first things we can do is configure and protect the passwords.
These include routing protocol and NTP authentication secrets, login, and
enable (privileged EXEC mode) passwords.

passwords and privileges
------------------------

There are many options available for user authentication; for example,
overriding access classes and TACACS support that we won't go into here.
However, there are some important things we wanted to mention regarding
passwords and privilege support.  First, different types of passwords have
different construction and length requirements.  For example, an OSPF simple
password can be any continuous string of characters that can be entered from
the keyboard up to 8 bytes in length, while an OSPF MD5 key can be any
alphanumeric password up to 16 bytes long.  A line password can be up to 80
characters in length and can contain any alphanumeric characters including
spaces.  An enable secret and username password can be up to 25 characters and
can contain embedded spaces.  In some cases the construction requirements are
not clearly documented so you'll have to experiment to come up with a "good"
password depending on your environment.

Earlier we mentioned "traditional" user EXEC and privileged EXEC.  There are
actually 16 privilege levels, numbered 0 through 15.  Level 1 is the normal
user EXEC mode and 15 is the default privileged EXEC mode.  To expand on
the sample earlier:

  reeb>show privilege
  Current privilege level is 1
  reeb>enable
  Password: 
  reeb#show privilege 
  Current privilege level is 15
  reeb#disable
  reeb>show privilege
  Current privilege level is 1

You can use the privilege mechanism to tailor the authentication configuration
to your specific environment.

For sample purposes, we will use separate, unique, personal login names for
each of the administrators granted access to the router for audit trail
purposes.  We will start with two users:

  username variablek password st0rk
  username brett password r0ddag
   
service password-encryption
---------------------------
 
By default, anyone with privileged access can view all of the passwords on the
router. If somebody is watching you configure the router, they can "shoulder
surf" and capture passwords.
   
You can use the "service password-encryption" command to encode or scramble
most of the various router password strings. These scrambled passwords are also
known as type 7 passwords because of the digit that precedes the encoded
password string.  Note that while technically the passwords are encrypted,
this service provides minimal protection and only serves to hide the passwords
from casual observation.  The scrambled passwords can be decoded trivially by a
simple shell script [1] or on a bar napkin while munching on a plate of
nachos or (in our case) drinking a Guinness.

Note that for some reason the password-encryption service does not encode SNMP
community names.

Granted this adds little in terms of password security, but we guess it doesn't
hurt.  We mainly point it out because its name has led to confusion regarding
its purpose and strength.
       
enable secret
-------------

The IOS equivalent of root access is privileged EXEC mode which is protected
by the enable password. There are two methods of protecting the enable
password.  The first method is to use "enable password" which only uses the
trivial Cisco encoding mechanism.
   
The second method is to use the "enable secret" command which uses MD5, a
one-way cryptographic hash function.  Passwords protected with MD5 are also
known as type 5 passwords. To use the enable secret command you can specify
the enable secret then disable the enable password if you have one:

  reeb(config)#enable secret s3kr3t
  reeb(config)#no enable password
  reeb(config)#exit

  reeb#sh running-config
  Building configuration...

  enable secret 5 $1$k2gM$4W2tuuTUqxuRd.LQxsh/v.

You might ask why not protect all passwords and secrets with MD5?  This won't
work because MD5 is a one-way hash, and IOS needs to be able to access clear
text strings for stuff like the MD5-based MAC secret that NTP can use for
authentication, or OSPF simple authentication strings and so on.


----[ Step 2 : Limit remote access

Cisco routers can be remotely managed via a TELNET connection.  It is a good
idea to limit, or even disable, TELNET access. To limit access you can specify
an access class on the VTY lines:
   
  access-list 99 permit host mgmt_ip
  access-list 99 deny any
  !
  line vty 0 4
   access-class 99 in
   login local

In addition, if you are using access lists with a default deny, you will need
to allow connections to tcp/23 from specific source IP addresses on the inside:

  !
  interface Ethernet0
   ip access-group e0-in in
  !
  ip access-list extended e0-in
   permit tcp host mgmt_ip host 192.168.0.1 eq 23

If we want to disable the TELNET listener completely (a good idea for exposed
routers that are high visibility targets), the following will work:
   
  line vty 0 4
   transport input none

An ultra-paranoid configuration might even be something like:

  access-list 99 deny any
  !
  line vty 0 4
   access-class 99 in
   exec-timeout 0 1
   login local
   transport input none

This configuration may be a bit overboard but it:

  * sets a deny any access class on the VTY
  * disables the TELNET listener
  * sets the EXEC session timeout to 1 second

There have been requests to add SSH support to IOS, apparently from as long as
3 years ago.  There was even a rumor that IOS 12.0 would contain SSH support,
but it didn't make it in.  There is also Kerberos support in IOS, and a way to
do Kerberized TELNET to the router, but we haven't used that.


----[  Step 3 : Limit local access

By default, when you connect to the console or AUX port, you are given user
EXEC mode access without a password. If the router cannot be physically
secured, it is a good idea to set a user EXEC password on these ports.  Even
if the router is in a secured environment, like a locked machine room, it
doesn't hurt.

  line con 0 
   login local
   ! logout idle console access after 2 min
   exec-timeout 2 0
  line aux 0
   ! Uncomment below to disable logins on the AUX port
   ! no exec
   ! Or allow password access
   login local

This will not stop a determined attacker from gaining access to the router.  If
an attacker has physical access to the box, they can use well-known password
recovery techniques to gain access. [2]


----[  Step 4 : Display login banner

It's a good idea to configure a login banner that warns users against
unauthorized access.  This may help in the event of legal action against
an intruder.  We tend to use something like the following:

banner motd #

             This is a private system operated for and by
		       Big Phreaking Bank (BPB).

         Authorization from BPB management is required to use
                             this system.

              Use by unauthorized persons is prohibited.
#

Though you should tailor it to meet your local requirements.  BPB might also
be considered an "inviting" target. For examples and more detailed information
on the topic of login banners refer to [3].

----[ Step 5 :  SNMP

Another common method of router management is to use the Simple Network
Management Protocol (SNMP).  IOS supports SNMPv1 and SNMPv2.  SNMPv1 was
not designed with authentication and data privacy features.  Some
implementations of SNMPv2 contain security enhancements. SNMPv3 apparently
contains more security enhancements.

We generally leave SNMP disabled on our bastion routers, however if you must
enable it, we recommend the following protective steps:

  * Use a hard-to-guess community name
  * Make the MIB read only
  * Permit access only from specific hosts
       
These precautions can be implemented using the following configuration:

  ! allow SNMP reads from hosts in access-list 10
  snmp-server community h4rd2gu3ss ro 10
  ! 
  ! access list for SNMP reads
  access-list 10 permit host snmp_mgmt_ip
  access-list 10 deny any
  !
  ! send traps with community names
  snmp-server trap-authentication
  ! send all traps to the management host on the inside interface
  snmp-server trap-source Ethernet0
  snmp-server host snmp_mgmt_ip h4rd2gu3ss
  !
  interface Ethernet0
   ip access-group e0-in in
  !
  ip access-list extended e0-in
  ! allow access from a specific machine on the inside
  permit udp host snmp_mgmt_ip host 192.168.0.1 eq snmp


----[  Step 6 :  Logging data

If your security policy requires that logs be generated for access list drops
or other security events, you can use the IOS syslog facility.  Since syslog
uses UDP, which is not a reliable transport mechanism, it can be good idea to
log messages to more than one host, which may reduce the occurrence of lost
messages due to packet loss or other weirdness (and it's a simple way to
automatically create a backup of your logs).  Also, using NTP to synchronize
all of the clocks greatly aids forensic log analysis in the event of an attack
or break in.
 
NTP Configuration
-----------------

Without synchronized time on the various hosts within your firewall complex
and network, event correlation from log message timestamps is nearly
impossible.  The NTP protocol and the Cisco NTP implementation support
cryptographic authentication using MD5 (DES is also supported by the protocol
as the authentication hash but MD5 doesn't suffer from US export bogosity).
This allows the NTP client to authenticate its time sources, and should
prevent attackers from spoofing NTP servers and playing with the system
clock.  If your budget can handle it, consider a network-based GPS stratum
1 NTP time server that supports MD5 authentication.  Below we configure
NTP to allow updates only from our internal time servers and authenticate
the messages using MD5 for the message authentication code (MAC).

  ! Setup our clock environment
  clock timezone PST -8
  clock summer-time zone recurring
  ! Configure NTP
  ntp authenticate
  ntp authentication-key 1 md5 ntpk3y
  ntp trusted-key 1
  ntp access-group peer 20
  ntp server ntp_server1_ip key 1 prefer
  ntp server ntp_server2_ip key 1
  !
  ! Allow selected ntp hosts
  access-list 20 permit host ntp_server1_ip
  access-list 20 permit host ntp_server2_ip
  access-list 20 deny any

Syslog setup
------------
 
In this case, we will send syslog messages to two hosts and stamp the messages
with the local date and time:

  ! Send syslog messages to the mgmt host and log with localtime
  service timestamps log datetime localtime
  logging syslog1_ip
  logging syslog2_ip

By default, the router will send syslog messages with a local7 facility.
If you want to store router messages in a separate file, your syslog.conf
should include the line:

  # router messages
  local7.*                                 /var/adm/router.log

The exact syntax and log file location may vary depending upon the syslogd you
are using.
   
You can change the facility using:
   
  logging facility facility-type

----[  Step 7 : Other protection mechanisms

no ip source-route
------------------

Some attacks use the IP source route option.  The attacks rely on the ability
of the attacker to specify the path a packet will take.  An attacker can send
a source routed packet to a victim host behind the router which will then
send back packets along the same path.  This allows replies to spoofed packets
to return to the attacker.  Many modern operating systems allow you to drop IP
packets with source route options set.  However, it is a good idea to drop
these packets at the edge using the "no ip source-route" option.
  
Limiting ICMP
-------------
 
Several DoS attacks use the ICMP protocol.  It is a good idea to limit what
types of ICMP messages are allowed.  At a minimum, in order to allow for Path
MTU discovery (PMTU), you should consider permitting packet-too-big messages.
The other types of ICMP messages allowed will depend upon the local security
policy.
   
  ip access-list extended e1-in
  ! Allow fragmentation needed messages (type 3 code 4)
   permit icmp any 192.168.0.0 0.0.255.255 packet-too-big
  ! Allow outbound ping and MS style traceroute (type 0)
   permit icmp any 192.168.0.0 0.0.255.255 echo-reply
  ! Uncomment to allow ping to the inside net (type 8)
  ! permit icmp any 192.168.0.0 0.0.255.255 echo
  ! Uncomment to allow traceroute
  ! permit icmp any 192.168.0.0 0.0.255.255 ttl-exceeded

Disable unnecessary services
----------------------------
 
Next we can disable unnecessary services.  By default, IOS has some services
enabled which will allow attackers to gain information and perform Denial of
Service attacks (though see above for issues with changing defaults in newer
IOS versions and determining what is really enabled).
   
We will disable these:
   
  no service udp-small-servers
  no service tcp-small-servers
  no service finger
  no ip bootp server
  ! not enabled by default but be paranoid
  no ip http server

no cdp run
----------

Cisco Discovery Protocol (CDP) is a media independent protocol which, by
default, runs on all Cisco equipment. The protocol is used for network
management and to discover other Cisco devices.  The Cisco documentation says:

   "CDP allows network management applications to discover Cisco
   devices that are neighbors of already known devices, in particular,
   neighbors running lower-layer, transparent protocols."
   
To turn CDP off on a specific interface, you can use:

     interface Ethernet1
      no cdp enable

To disable CDP on all interfaces, you can use the global command:

     no cdp run

no ip unreachables
------------------
 
By default, when an access list drops a packet, the router returns a type 3,
code 13 ICMP (administratively prohibited) message.  This allows potential
attackers to know that the router implements access list filters. Also, most
UDP scans rely on the target sending back unreachable messages.  To thwart
UDP scans we can prevent the router from sending any ICMP type 3 (unreachable)
messages by specifying the following on each interface:

  no ip unreachables

no ip proxy-arp
---------------

By default, IOS enables proxy ARP on all interfaces.  Since we don't need
the service, we will disable it:

  interface Ethernet0
   no ip proxy-arp
  interface Ethernet1
   no ip proxy-arp

no ip redirects
---------------

In cases where we have no need to send redirects, we will disable them:

  interface Ethernet0
   no ip redirects
  interface Ethernet1
   no ip redirects

----[  Step 8 : Anti-spoofing

The idea behind anti-spoofing is that nobody from the outside network should
be sending packets to you with a source address of either your inside network
address, or certain well-known and reserved addresses. We will use access
lists to drop and log any of these packets.  A recent Internet draft is
available (draft-manning-dsua-00.txt) which discusses the reserved netblocks
that should be blocked at the edge.

  ip access-list extended e1-in
  ! Anti-spoofing: no packets with a src address = our inside net
  ! Normally, this would not be a RFC 1918 net
   deny ip 192.168.0.0 0.0.255.255 any log
  !
  ! Deny first octet zeros, all ones, and loopback network
   deny ip 0.0.0.0 0.255.255.255 any log
   deny ip host 255.255.255.255 any log
   deny ip 127.0.0.0 0.255.255.255 any log
  !
  ! Deny class D (multicast) and class E (reserved for future use)
   deny ip 224.0.0.0 15.255.255.255 any log
   deny ip 240.0.0.0 7.255.255.255 any log
  !
  ! Deny RFC 1918 addresses
   deny ip 10.0.0.0 0.255.255.255 any log
   deny ip 172.16.0.0 0.15.255.255 any log
  ! included above in this example
  ! deny ip 192.168.0.0 0.0.255.255 any log
  !
  ! Deny test-net
   deny ip 192.0.2.0 0.0.0.255 any log
  !
  ! Deny end node autoconfig 
   deny ip 169.254.0.0 0.0.255.255 any log

What you really want is a switch that will drop packets arriving on an
interface with a source address that is not routed out that interface.  Some
IOS releases have the ability to do this by using something called Cisco
Express Forwarding (CEF) in conjunction with the "ip verify unicast
reverse-path" interface command.  This requires strictly symmetric routing
patterns and a 7500 Series (any 7000 with IOS 11.3) or a 12000 Gigabit switch
router to run CEF.


----[  Step 9 : Mitigating Denial of Service attacks
 
There have been a rash of new Denial of Service (DoS) attacks over the past
few years.  We can use access lists and other mechanisms to prevent or at
least increase our ability to withstand some common DoS attacks.
   
SYN Floods
----------
 
A SYN flood occurs when an attacker sends a TCP SYN segment with an
unreachable spoofed source address to an open port on the target.  The victim
responds with a SYN,ACK to the unreachable host and the TCP handshake never
completes. The victim's connection queue quickly gets filled with half-open
connections in the SYN_RCVD state.  At some point, the server TCP will start
to drop new SYNs.
   
SYN floods are discussed in the Cisco publication "Defining Strategies to
Protect Against TCP SYN Denial of Service Attacks" [4].  Cisco IOS has a
mechanism called TCP Intercept [5] which can be used to help protect against
SYN floods.  TCP Intercept was introduced in IOS 11.3 and requires a specific
feature set; it's in the Enterprise feature set and we hear some service
provider feature sets and maybe others.
   
We have found that TCP Intercept works well in practice (protecting against
real SYN floods); however, configuring it can be very confusing and the
specifics will vary depending on a number of factors.  We recommend reading
the Cisco documentation and if you are susceptible to SYN floods you may
consider implementing TCP Intercept to mitigate the effects.
   
Land attack
-----------
 
The land program sends a packet to the victim with identical source and
destination port, and identical IP addresses. This causes many network devices
with to panic, including Unix hosts, Windows hosts,  routers, etc.

We recommend that you run one of the newer IOS releases which contains fixes
for this defect. A Cisco field notice provides details on which IOS versions
are vulnerable. [6] If you can't update to a newer IOS, the field notice
also contains information on how to configure access lists for protection.

Stop malicious insiders (Ingress Filtering)
-------------------------------------------
 
If the inside network has untrusted hosts or users, you might want to use
Ingress Filtering [7].  By denying packets with spoofed source addresses,
Ingress Filtering prevents malicious inside users from launching some
Denial of Service attacks.
   
In our case, this would be achieved by allowing the valid inside
addresses out and then denying all others:

  ! Ingress filter: only allow our net outbound
  ip access-list extended e0-in   
   permit ip 192.168.0.0 0.0.255.255 any
   deny ip any any log
  !
  ! apply to inbound packets on the inside interface
  interface Ethernet0
   ip access-group e0-in in

Smurf attacks
-------------
 
Smurf attacks continue to plague the Internet. If you don't take appropriate
steps, you can be either a victim or an amplifier in a Smurf attack.  Craig
Huegen has written a paper that details Smurf attacks and defenses [8].
   
To prevent your network from being used as a smurf amplifier, you need
to filter packets sent to the broadcast address of your network.
   
  interface Ethernet0
   no ip directed-broadcast

  interface Ethernet1
   no ip directed-broadcast


----[  Step 10 : Protect hosts behind the router

The router can also provide additional protection to any hosts behind it.
This may include bastion hosts running web, FTP, mail, and DNS servers.  As an
example, we will implement access lists to screen access to an HTTP server
host (192.168.0.5). We think it is generally a good idea to filter both
inbound and outbound packets (using inbound "in" access lists of each
interface--we rarely come across cases where we use outbound "out" access
lists).

  ip access-list extended e1-in
  ! allow tcp/80 to the web server
   permit tcp any host 192.168.0.5 eq www
  !
  interface Ethernet1
   ip access-group e1-in in

  ip access-list extended e0-in
  ! allow established connections from the web server
   permit tcp host 192.168.0.5 eq www any established
  !
  interface Ethernet0
   ip access-group e0-in in

Note that this will not protect against command channel attacks directed at
the permitted services.


----[  Step 11 : Verify the configuration

As mentioned earlier, depending upon the IOS version, a "sh running-config"
might not display whether TCP and UDP small-servers are enabled.  You should,
at a minimum, run a port scan against the router to verify the basic
configuration. Note that if you have disabled IP unreachables, you will have
to temporarily re-enable them to perform a UDP scan.

You can use Fyodor's nmap program [9] to perform the scans.

TCP scan
--------

[root@fuel src]# nmap -sT 192.168.0.1 -p 1-65535

Starting nmap V. 2.12 by Fyodor (fyodor@dhp.com, www.insecure.org/nmap/)
Interesting ports on  (192.168.0.1):
Port    State       Protocol  Service
23      open        tcp        telnet          

If you do not allow VTY access, there shouldn't be any ports open.  In this
case, we are allowing TELNET access from the same host that performed the scan.

UDP scan
--------

[root@fuel config]# nmap -sU 192.168.0.1
WARNING:  -sU is now UDP scan -- for TCP FIN scan use -sF

Starting nmap V. 2.12 by Fyodor (fyodor@dhp.com, www.insecure.org/nmap/)
Interesting ports on  (192.168.0.1):
Port    State       Protocol  Service
123     open        udp        ntp             
161     open        udp        snmp            
387     open        udp        aurp            
611     open        udp        npmp-gui        
727     open        udp        unknown         
910     open        udp        unknown         
   
Note: We have seen false positives when using nmap for router UDP scans.  It
can be a good approach to use multiple scanners for these tests.  Below we
point udp_scan from SATAN at the router.  In this case, it turns out that
611/udp and 727/udp are not really open:

[root@fuel bin]# ./udp_scan 192.168.0.1 1-1024
123:ntp:
161:snmp:
387:UNKNOWN:
910:UNKNOWN:

Also, we have noticed that IOS versions 11.2 and 11.3 have 387/udp and 910/udp
open.  If someone at Cisco could explain this, we sure would like to hear it.
We don't have Appletalk enabled so that doesn't explain the udp/387.  We
tested IOS 12.0 with the exact same configuration and they are not open.


----[ Thanks to...

Thanks to everybody who reviewed the paper and provided valuable
feedback. You know who you are.


----[  References
  
General References
------------------
 
Increasing Security on IP Networks is at
http://www.cisco.com/univercd/cc/td/doc/cisintwk/ics/cs003.htm
   
Cisco Internet Security Advisories can be found at
http://www.cisco.com/warp/public/707/advisory.html

Specific References
-------------------

[0] Marcus J. Ranum, "Thinking About Firewalls V2.0: Beyond
    Perimeter Security"
    http://www.clark.net/pub/mjr/pubs/think/index.htm
 
[1] Decoding type 7 passwords
    http://geek-girl.com/bugtraq/1997_4/0156.html

[2] Password Recovery Techniques
    http://www.cisco.com/warp/public/701/22.html

[3] CIAC bulletin on login banners
    http://ciac.llnl.gov/ciac/bulletins/j-043.shtml
          
[4] "Defining Strategies to Protect Against TCP SYN Denial of
    Service Attacks"
    http://www.cisco.com/warp/public/707/4.html
          
[5] Information on TCP Intercept
    http://www.cisco.com/univercd/cc/td/doc/product/software/ios113ed/
    113ed_cr/secur_c/scprt3/scdenial.htm
          
[6] Information on land attacks
    http://www.cisco.com/warp/public/770/land-pub.shtml
          
[7] RFC 2267: Network Ingress Filtering: Defeating Denial of
    Service Attacks by P. Ferguson and D. Senie
    ftp://ftp.isi.edu/in-notes/rfc2267.txt
          
[8] Craig Huegen's paper
    http://users.quadrunner.com/chuegen/smurf.cgi
       
    Cisco has a paper Minimizing the Effects of "Smurfing" Denial
    of Service (DOS) Attacks
    http://www.cisco.com/warp/public/707/5.html

[9] Fyodor's nmap
    http://www.insecure.org/nmap/


----[  Appendix A

The complete router configuration is given below.

<++> P55/Bastion-router/cisco-conf.txt !75510e67
! We have replaced the mnemonic names with the following addresses:
!
! ntp_server1_ip: 192.168.1.100
! ntp_server2_ip: 192.168.1.101
! syslog1_ip: 192.168.1.102
! syslog1_ip: 192.168.1.103
! mgmt_ip: 192.168.1.104
! snmp_mgmt_ip: 192.168.1.105
!
version 11.3
service timestamps debug uptime
service timestamps log datetime localtime
!
! protect passwords
service password-encryption
enable secret 5 $1$k2gM$4W2tuuTUqxuRd.LQxsh/v.
!
username variablek password 7 110F0B0012
username brett password 7 15190E1A0D24
ip subnet-zero
!
hostname reeb
!
interface Ethernet0
 description Inside Interface
 ip address 192.168.0.1 255.255.0.0
 ip access-group e0-in in
 no ip directed-broadcast
 no ip unreachables
 no ip proxy-arp
 no ip redirects
!
interface Ethernet1
 description Outside Interface
 ip address 172.16.1.1 255.255.255.252
 ip access-group e1-in in
 no ip directed-broadcast
 no ip unreachables
 no ip proxy-arp
 no ip redirects
!
! turn off unnecessary services
no ip bootp server
! the http server is disabled by default. but be paranoid.
no ip http server
no service tcp-small-servers
no service udp-small-servers
no service finger
no cdp run
!
! disable source routed packets
no ip source-route
!
! setup the clock
clock timezone PST -8
clock summer-time zone recurring
! setup NTP
ntp authenticate
ntp authentication-key 1 md5 151C1F1C0F7932 7
ntp trusted-key 1
ntp access-group peer 20
ntp server 192.168.1.100 key 1 prefer
ntp server 192.168.1.101 key 1
!
! configure logging
service timestamps log datetime localtime
logging buffered 4096 informational
logging console informational
logging 192.168.1.102
logging 192.168.1.103
!
! configure SNMP
! allow SNMP reads from hosts in access-list 10
snmp-server community h4rd2gu3ss ro 10
! send traps with community names
snmp-server trap-authentication
! send all traps to the management host on the inside interface
snmp-server trap-source Ethernet0
snmp-server host 192.168.1.105 h4rd2gu3ss
!
! simple static routing. default to the ISP
ip route 0.0.0.0 0.0.0.0 172.16.1.2 
ip route 192.168.0.0 255.255.0.0 192.168.0.2
!
! standard ip access-lists
!
! allowed hosts for SNMP reads
no access-list 10
access-list 10 permit host 192.168.1.105
access-list 10 deny any
!
! ntp hosts
no access-list 20
access-list 20 permit host 192.168.1.100
access-list 20 permit host 192.168.1.101
access-list 20 deny any
!
! hosts allowed to telnet to the router
no access-list 99
access-list 99 permit host 192.168.1.104
access-list 99 deny any
!
! extended ip access-lists
!
no ip access-list extended e1-in
ip access-list extended e1-in
! Anti-spoofing
! Deny packets on outside with src address = our inside nets
! This normally wouldn't be a RFC 1918 network
 deny ip 192.168.0.0 0.0.255.255 any log
!
! Deny first octet zeros, all ones, and loopback
 deny ip 0.0.0.0 0.255.255.255 any log
 deny ip host 255.255.255.255 any log
 deny ip 127.0.0.0 0.255.255.255 any log
!
! Deny class D (multicast) and class E (reserved for future use)
 deny ip 224.0.0.0 15.255.255.255 any log
 deny ip 240.0.0.0 7.255.255.255 any log
!
! Deny RFC 1918 addresses
 deny ip 10.0.0.0 0.255.255.255 any log
 deny ip 172.16.0.0 0.15.255.255 any log
! included above in this example
! deny ip 192.168.0.0 0.0.255.255 any log
!
! Deny test-net
 deny ip 192.0.2.0 0.0.0.255 any log
! Deny end node autoconfig 
 deny ip 169.254.0.0 0.0.255.255 any log
!
! ICMP allows
! Allow fragmentation needed messages (type 3 code 4)
 permit icmp any 192.168.0.0 0.0.255.255 packet-too-big
! Allow outbound ping and MS style traceroute (type 0)
 permit icmp any 192.168.0.0 0.0.255.255 echo-reply
! Uncomment to allow ping to the inside net (type 8)
! permit icmp any 192.168.0.0 0.0.255.255 echo
! Uncomment to allow traceroute
! permit icmp any 192.168.0.0 0.0.255.255  ttl-exceeded
!
! permit certain connections
! example: permit connections from the outside to a web server
 permit tcp any host 192.168.0.5 eq 80
!
! explicit default deny
 deny ip any any log
!
no ip access-list extended e0-in
ip access-list extended e0-in
!
! our policy is only allow replies from the inside web server,
! some ICMP and specific inside hosts to access the router.
!
! permit certain connections
! example: allow responses from the web server
 permit tcp host 192.168.0.5 eq www any established
!
! allow connections from ntp, mgmt, etc. to the router
 permit udp host 192.168.1.105 host 192.168.0.1 eq snmp
 permit udp host 192.168.1.100 host 192.168.0.1 eq ntp
 permit udp host 192.168.1.101 host 192.168.0.1 eq ntp
 permit tcp host 192.168.1.104 host 192.168.0.1 eq telnet
!
! allow specific ICMP out
 permit icmp 192.168.0.0 0.0.255.255  any packet-too-big
 permit icmp 192.168.0.0 0.0.255.255 any echo
! Uncomment to allow inbound ping responses
! permit icmp 192.168.0.0 0.0.255.255 any echo-reply
! Uncomment to allow traceroute
! permit icmp 192.168.0.0 0.0.255.255 any ttl-exceeded
!
! Ingress filtering: uncomment to deny connections to router and 
! then allow outbound if source address = our net. In this case, 
! we don't allow any traffic out and go directly to explicit deny.
! deny ip any host 192.168.0.1 log
! permit ip 192.168.0.0 0.0.255.255 any 
!
! explicit deny
 deny ip any any log
!
!
line con 0
 login local
! logout idle console access after two min
 exec-timeout 2 0
line aux 0
! Uncomment below to disable logins on the AUX port
! no exec
! Or allow password access
 login local
line vty 0 4
! uncomment to disable telnet listener
! transport input none
 access-class 99 in
 login local
end

$Id: bastion-ios.txt,v 1.26 1999/06/24 17:06:21 beldridg Exp $
<-->

----[  EOF