Effective Uses of Packet-Filtering Devices

Because packet filtering is older technology and lacks the capability to differentiate between types of network traffic, you might be wondering why we are discussing it. What could be the possible use for this technology in a world that is filled with hi-tech firewalls that can track protocols using knowledge of the way they work to intelligently differentiate between incoming and outgoing traffic streams? Good question! Why use a PC when we have supercomputers? Sometimes a lighter-weight, less expensive means to get things done is a major advantage. Because packet filters don't go to great depth in their analysis of traffic streams, they are faster than other firewall technologies. This is partially due to the speed at which the header information can be checked and partially due to the fact that packets don't have to be "decoded" to the application level for a decision to be made on them. Complex decisions are not necessary, simply a comparison of bits in a packet to bits in an ACL.

Filtering Based on Source Address: The Cisco Standard ACL

One of the things that packet-filtering technology is great for is the blocking or allowing of traffic based on the IP address of the source system. Some ways that this technology can be usefully applied are filters blocking specific hosts (blacklisting), filters allowing specific hosts (such as business partners), and in the implementation of ingress and egress filters. Any of these examples can be implemented on a Cisco router by using a "standard" access list.

The standard access list is used to specifically allow or disallow traffic from a given source IP address only. It cannot filter based on destination or port number. Because of these limitations, the standard access list is fast and should be preferred when the source address is the only criteria on which you need to filter.

The syntax for a standard access list is as follows:

 access-list list number 1-99 or 1300-1999 permit |deny source address mask log 

Notice that when ACLs were first created, the list number had to be 199. This range was expanded in IOS version 12.0(1) to include the numbers 13001999. The only way that the Cisco IOS can identify the list as a standard ACL is if a list number in one of these two ranges is used. The mask option is a required wildcard mask, which tells the router whether this is a single host we are filtering or an entire network range. (For more information on wildcard masks, check out the sidebar "The Cisco Wildcard Mask Explained" later in this chapter). The log option can be appended to tell the router to specifically log any matches of this filter. These log entries can be saved in local memory or more appropriately sent to a remote Syslog server. For more information on router logging, see Chapter 6, "The Role of a Router," and Chapter 20, "Network Log Analysis."

The previously listed access list notation is entered in global configuration mode and can be applied to the interface in the interface configuration mode with the access-group statement, as shown here:

 ip access-group list number in|out 

The access-group command is used to specifically apply an ACL to an interface (by its list number) either inbound or outbound. Only one access list can be applied in one direction (in or out) per interface. This means a maximum of two applied ACLs per interface: one inbound and one outbound.

One of the confusing concepts of router ACLs is the way that applying filters "in" or "out" works. This is confusing because people normally visualize "in" as traffic moving toward their internal network and "out" as traffic moving away from their network toward outside entities. However, this premise does not necessarily hold true when talking about the in and out keywords in Cisco router access lists. Specifically, the keywords tell the router to check the traffic moving toward (in) or away from (out) the interface listed.

In a simple dual-interface router, this concept is more easily illustrated. Let's assume you have an interface called e1 (hooked to your internal network) and an external interface called s1 (hooked up to the Internet, for example). Traffic that comes into the s1 interface moves toward your internal network, whereas traffic that goes out of the s1 interface moves toward the Internet.

So far this seems to be pretty logical, but now let's consider the internal e1 interface. Traffic that comes into e1 moves away from your internal network (toward the Internet), and traffic that goes out of e1 goes toward your internal network.

VLAN Interfaces and Direction

When determining direction, VLAN interfaces are a little more confusing than physical router interfaces. Applying an access group "in" on a VLAN interface means that traffic moving away from the network will be filtered, whereas "out" means that traffic coming into the VLAN will be filtered.

Keep in mind when you apply access-group commands to your interfaces that you apply the access list in the direction that the traffic is traveling, in regards to the router's interface. You might be thinking, "What is the difference, then, between inbound on the s1 interface and outbound on the e1 interface? Both refer to traffic that is moving in the same direction. Which is more appropriate to use?"


To maximize performance, filter traffic as it enters the router.

The less work the router has to do, the better. Always try to filter traffic at the first interface it enters, or apply your filters "inbound" as much as possible.

Returning to our example, if something should be blocked (or permitted, for that matter) coming in from the Internet, it would make the most sense to block it coming in to the s1 (outside) interface. In addition, if something should be filtered leaving your network, it would be best to filter it inbound on the e1 (inside) interface. Basically, you should show preference to the in keyword in your access lists. Some specific exceptions to this rule exist, which involve certain access list types (such as reflexive ACLs) that require being placed outbound.

These examples often list the actual prompt for a router named "router" to remind you of the command modes that the router will be in when entering the various commands.

The following is an example of an actual filter that uses the previous syntax and 190.190.190.x as the source network's IP address that you want to deny:

 router(config)#access-list 11 deny 

This filter's list number is 11. It denies any packet with a source network address of 190.190.190 with any source host address. It is applied to the interface inbound, so it filters the traffic on the way into the router's interface.

The command to apply the access list to your serial 1 interface would be

 router(config-if)#ip access-group 11 in 

where we are in interface configuration mode for the interface to which we are applying the access list, inbound.

The Cisco Wildcard Mask Explained

The wildcard mask is one of the least understood portions of the Cisco ACL syntax. Take a look at the following example:

 access-list 12 permit 

In this case, represents the wildcard mask. It looks like a reverse subnet mask and represents the portion of the listed IP address range to filter against the traffic in question. Zeros mean, "Test this portion of the address," and ones mean, "Ignore this portion of the address when testing."

In our example, let's say a packet comes in with a source address of Because the first octet of the wildcard mask is a zero, the router compares the first octet of the incoming packet to the value 192, listed in the access list. In this case, they are the same, so the router continues to the second octet of the wildcard mask, which is also a zero. Again, the second octet of the value of the source address of the incoming packet is compared to the value 168. Because they are also the same, the router continues to the third octet. Because the wildcard mask specifies a zero in the third octet as well, it continues to test the address, but the value of the third octet does not match, so the packet is dropped.

For the sake of example, let's continue to look at the fourth octet, even though in actuality the packet would have been dropped at this point. The wildcard's fourth octet is valued at 255. In binary, this equates to 11111111. In this example, the value 0 in the access list does not match the value 27 of the compared packet; however, because the wildcard wants us to ignore this octet, the access list allows the packet to pass (assuming it hadn't failed on the previous octet). The concept might seem pretty easy with a wildcard value that deals with entire octets, but it gets tricky when you need to deal with an address range that is smaller than 255. The reality is that the router doesn't test octet by octet, but bit by bit through each of the octets.

What if you want to allow traffic from systems in the address range only? The first three octets of the wildcard are easy: 0.0.0. It's the last octet that is difficult. It's time to grab your handy-dandy binary calculator. Consider what 16 looks like in binary: 0001 0000. Now look at 31: 0001 1111. The difference between these two binary values occurs in the last four bits. Therefore, the values between 16 and 31 are covered in the range 1000011111. To allow those values, you need to place zeros in your wildcard mask for the portions that need to match exactly, and ones for the binary values that change. Because the last four bits are the only bits that change in the desired range, the wildcard mask reflects those four bits with ones. In binary, our wildcard mask is as follows:


Translated with our binary calculator, that is

This wildcard mask works for any range of 15 addresses you are comparing, so to make it work for 1631, you must properly reflect the range in the IP address portion of the access list. Your final access list looks like this:

 access-list 10 permit 

A wildcard mask is always contiguous zeros and ones, without interruption, as in the example listed previously. In some cases, you need more than one ACL statement and wildcard mask to cover a range of network addresses. For example, if you want to block addresses in the range 232255, you need the command

 access-list 110 deny ip any 

to block the range 232239, and you also need to specify

 access-list 110 deny ip any 

to block the range 240255.

Blacklisting: The Blocking of Specific Addresses

One popular use of the standard access list is the "blacklisting" of particular host networks. This means that you can block a single host or an entire network from accessing your network. The most popular reason for blocking a given address is mischief. If your intrusion detection system (IDS) shows that you are being scanned constantly by a given address, or you have found that a certain IP address seems to be constantly trying to log in to your systems, you might simply want to block it as a preventative measure.

For example, if your intranet web server should only be offering its information to your business locations in the continental United States, and you are getting a lot of hits from a range of IPs in China, you might want to consider blocking those addresses.


The blocking of "spoofed" addresses can lead to a denial of service condition. Always research IP addresses before uniformly blocking them.

The blocking of address ranges is also a popular way to "band-aid" your system against an immediate threat. For example, if one of your servers was being attacked from a certain IP address, you could simply block all traffic from that host or network number. As another example, if you just found out that you had a widespread infection of a Trojan that contacted a remote IRC server, you could block all traffic coming from that IRC server. It would be more appropriate to block the traffic leaving your network to that destination, however, but that would require an extended access list because standard ACLs only filter on source address.

A sample access list to block access from an outside address range would be

 router(config)#access-list 11 deny router(config-if)# ip access-group 11 in 

where the network number of the outside parties to be blocked would be (Of course, this address range is part of the ranges reserved for private addressing, and it's simply used as an example in this instance.) This access list would be applied to the external router interface, inbound.


Once I was perusing my logs to check for Internet connections from my network during off-hours to see if anything peculiar was going on. I noticed some connections that were initiated in the middle of the night from various stations, repeatedly contacting a similar network address. I did some research on the address and determined that the maker of a popular freeware program owned it.

After a brief inspection of one of the "beaconing" stations, I found that the freeware program in question was loaded on the system. Being completely paranoid (as all good security professionals are), I immediately set up a rule on my perimeter router to block all access to the network address in question. Many software packages search for updates regularly and do not use these automatic "phoning home" sessions for malicious activity. However, as a firewall administrator, I have to be in control of my network's traffic flow. Besides, I do question why these packages need to "call out" several times a night.

Of course, my block was just a quick fix until I could unload the software at all the "infected" stations, but if it somehow appeared again, I had no fear of the software gaining outside access. The moral of the story: It can be a lot more efficient to block a single network address at your perimeter router than to run from station to station to audit software.

"Friendly Net": Allowing Specific Addresses

Another way you can use a standard access list is to permit traffic from a given IP address. However, this is not recommended. Allowing access to an address in this manner, without any kind of authentication, can make you a candidate for attacks and scans that use spoofed addresses. Because we can only filter on source address with a standard ACL, any inside device with an IP address can be accessed. Also, it's impossible to protect individual services on those devices. If you need to set up access like this and can't do it through a solution that requires authentication or some type of VPN, it is probably best to at least use an access list that considers more than the source address. We'll discuss this more in the section on extended access lists. However, this type of access may be suitable in situations requiring less security, such as access between internal network segments. For example, if Bob in accounting needs access to your intranet server segment, a standard ACL would be a simple way to allow his station access.

Ingress Filtering

RFC 1918 pertains to reserved addresses. Private/reserved addresses are ranges of IP addresses that will never be distributed for public use. This way, you can use these addresses in internal networks without worry of accidentally picking network addresses that might correspond with a public address you might want to access some day.

For example, imagine that you are in a world in which reserved private addresses don't exist. You are installing a new private network for your business that will have access to the Internet and will be running TCP/IP, so you will have to come up with a range of IP addresses for your stations. You don't want these stations to have public addressing because they won't be serving information to the Internet, and you will be accessing the Internet through a proxy server. You pick a range of IP addresses at random (say, You configure your stations and set up Internet access.

Everything is working great until the first time you attempt to access your bank's website and receive an error. You call the bank, and the bank says that everything is fine on its end. You eventually come to discover that the bank is right. The bank's system you were trying to contact has a public IP address of, the same address you have configured for one of your own stations. Every time your web browser goes to send information to the bank, it never even leaves your internal network. You've been asking your CAD station for a web page, and because your CAD station doesn't run web server software, you've just been getting an error. This example paints a clearer picture of why reserved addresses are so important for the interrelationship of public and private TCP/IP networks.

The reserved ranges are as follows:

  • Class A:

  • Class B:

  • Class C:

Because these ranges are often used as internal network numbers, they are good candidates for someone who is crafting packets or doing other malicious packet-transmitting behavior, including denial of service. Therefore, these ranges should be blocked at the outside of your network. In addition, the loopback address (the default address that all IP stations use to "address" themselves) is another candidate for being blocked, for the same reason. While you are blocking invalid addresses, you should also block the multicast address range and the invalid address Following is a sample access list to accomplish this:

 router(config)#access-list 11 deny router(config)#access-list 11 deny router(config)#access-list 11 deny router(config)#access-list 11 deny router(config)#access-list 11 deny router(config)#access-list 11 deny host router(config-if)# ip access-group 11 in 

These access lists are similar to the last one, denying access to the IP address ranges listed in an inbound direction on the applied interface. Notice the host keyword when blocking in the previous example. When blocking a single host, instead of following the IP address with the wildcard, you can precede the address with the keyword host. These lists have the same "implicit deny" that any access list does. This means that somewhere in access list number 11, a permit statement would have to exist; otherwise, all inbound traffic would be denied! An example of an appropriate permit statement might be one that allows the return of established traffic, like the following:

 router(config)# access-list 111 permit tcp any any established 

This is, however, an extended access list, which we'll talk more about later in the section "Filtering by Port and Destination Address: The Cisco Extended ACL."

It is also advisable that you create a rule to block traffic coming into your network that claims to have a source address matching that of your internal network, to complete your ingress access list. Valid traffic from the outside world doesn't have to have the same addressing as your stations. However, if this traffic is allowed to pass, it could bypass security mechanisms that think the traffic is local. If you are using one of the standard private address ranges, this is already done. If you're not, it would look like this:

 router(config)#access-list 11 deny 

Here, your network address range is This rule would be added to the previous list before the line that allows return traffic.

Ingress filters are an excellent example of a means to use packet-filtering technology to its fullest, on any network. Even if you have a stateful or proxy firewall, why not let your perimeter router use packet filtering to strip off this unwanted traffic? Let perimeter routers be the "bouncers" of your network, stripping off the undesirables before they even reach other internal protection devices.

    Inside Network Perimeter Security
    Inside Network Perimeter Security (2nd Edition)
    ISBN: 0672327376
    EAN: 2147483647
    Year: 2005
    Pages: 230

    Similar book on Amazon

    flylib.com © 2008-2017.
    If you may any questions please contact us: flylib@qtcs.net