A sniffer is any device, whether software or hardware, that grabs information traveling along a network. That network could be running any protocol: Ethernet, TCP/IP, IPX, or others (or any combination of these). The purpose of the sniffer to place the network interface--in this case, the Ethernet adapter--into promiscuous mode and by doing so, to capture all network traffic.
NOTE: Promiscuous mode refers to that mode where all workstations on a network listen to all traffic, not simply their own. In other words, non-promiscuous mode is where a workstation only listens to traffic route it its own address. In promiscuous mode, the workstation listens to all traffic, no matter what address this traffic was intended for.
When one discusses sniffers, one is not discussing key capture utilities, which grab keystrokes and nothing more. Essentially, a key capture utility is the software equivalent of peering over someone's shoulder. This peering might or might not reveal important information. True, it might capture passwords typed into the console of the local terminal, but what about other terminals? In contrast, sniffers capture network traffic. This network traffic (irrespective of what protocol is running) is composed of packets (these might be IP datagrams or Ethernet packets). These are exchanged between machines at a very low level of the operating-system network interface. However, these also carry vital data, sometimes very sensitive data. Sniffers are designed to capture and archive that data for later inspection.
As I have discussed, Ethernet was created at Xerox's Palo Alto Research Center. (Sometimes referred to as PARC Place.) You might remember an RFC document that I presented earlier in this book: It was posted over a Christmas holiday and discussed the issue of hackers gaining access to a network that would soon become the Internet. The author of that RFC was Bob Metcalfe, who, along with David Boggs (both at PARC), invented Ethernet.
In 1976, these two gentlemen presented to the computing communities a document titled Ethernet: Distributed Packet Switching for Local Computer Networks. The ideas set forth in that paper revolutionized business-based computing. Prior to the birth of Ethernet, most large networks were strung to mainframe computers (in even earlier years, most systems were based on computer time sharing).
Today, Ethernet is probably the most popular way to network machines. A group of machines within an office that are linked via Ethernet might be referred to as a local area network (LAN). These machines are strung together with high-speed cable that transfers information as quickly (or sometimes much more quickly) than most hard drives.
NOTE: You might remember that in Chapter 6, "A Brief Primer on TCP/IP," I noted that one element of TCP/IP networking was the full-duplex transmission path, which allows information to travel in both directions at one time, a common situation in TCP/IP that is especially vital to the error-reporting process during a transmission (a typical example might be during a FTP transfer). Ethernet does not truly support full-duplex transmission and therefore, although Ethernet interfaces are advertised as being capable of extremely high-speed transmission, you can expect only perhaps 50-75 percent of the actual advertised speed when using Ethernet on a high-traffic network. If you were to employ a packet sniffer, you would see that while a card is receiving a heavy transmission of data from some card elsewhere on the network, it cannot also send data out with any great reliability. That represents an interesting security issue of sorts. For example, can an Ethernet card answer an ARP request while being bombarded with data? If not, couldn't a cracker temporarily conduct an ARP spoofing session under such circumstances? At any rate, there are switching products that can remedy this limitation.
The composition of a network is complex. First, in order for each machine to be part of a network, it must have both software and hardware designed to traffic Ethernet packets. The four minimal components necessary are illustrated in Figure 12.1.
Figure 12.1.
The minimum requirements for a single workstation.
The software can either come with the operating system (Novell NetWare, UNIX, Windows NT, Windows 95), or it can be a third-party product added later (LANtastic). At a minimum, the software needed is as follows:
The network adapter driver commonly comes with the network adapter or Ethernet card. It is typically provided by the manufacturer of the card but might also be included in a total package. This is not always true. It is primarily the IBM-compatible architecture that requires an Ethernet card. Most workstations (and most Macintoshes) have on-board Ethernet support. This means that the Ethernet card is already hard-wired to the motherboard. I believe that IBM-based RS/6000 machines might be one of the few real exceptions to this. A good example would be an IBM Powerstation 320H.
NOTE: Most operating systems now come with boot drivers for various Ethernet cards. Linux certainly does, as does Windows 95 and Windows NT. Chances are, unless you have a very strange, off-beat Ethernet card, you may never need the manufacturer's driver.
The packet driver negotiates packets back and forth. The network adapter driver is used to bind the Ethernet protocol to the Ethernet card. The card transmits these packets from the workstation and into wire. This wire may be one of several kinds. Some Ethernet cable transmits packets at 10MB/sec, others at 100MB/sec.
NOTE: TCP/IP can be bound to most Ethernet cards as quickly as IPX or other network protocols.
So you have a machine running Ethernet software (for both packet and card). The machine is a classic workstation, equipped with an Ethernet card that connects to a cable. But where does the data that travels down that cable lead? The answer depends on the network needs of the organization.
In general, there will be a least several other workstations and a network hub (see Figure 12.2). The workstations may be deposited throughout a building, with the wire strung through the walls.
Figure 12.2.
Basic network setting.
Figure 12.2 shows a very simple network setting. Thousands of businesses nationwide have such a setting, using any of a dozen different networked operating systems.
NOTE: In many network settings, you can take the hub out of the picture altogether. There are plenty of Novell NetWare networks that have simply a file server or a closed-circuit cabling scheme, precisely like the setup in Figure 12.2. Hubs are used for many things, including enhancement of security, as you will see later. But if you have no fear of allowing indiscriminate, network-wide broadcasts, a hub might not be necessary.
Note the line in Figure 12.2 that represents information flow. On networks without hubs, the data doesn't point in any particular direction. Instead, it travels in all directions. A typical example of this is at the moment a message needs to be sent. Each network node or workstation is an interface. When a message needs to be sent, a request is forwarded to all interfaces, looking for the intended recipient. This request is sent in the form of a general broadcast.
This broadcast issues a message to all interfaces, saying: "Hey! Which one of you is this data destined for? Will the real recipient please stand up?" All interfaces receive this message, but only one (the one for which the message is intended) actually replies. In this respect, then, there is no established flow of information until the recipient is known. As you might expect, because this broadcast is global on the network, all machines hear it. Those that are not intended recipients of the data hear the broadcast but ignore it. The request packet dies at such workstations because there is no reply.
NOTE: This all broadcast scenario only occurs in network blocks, or segments. In other words, bar hard-wiring by hub (where all machines are strung to a hub), the information will be broadcast between all machines within that network segment. As you will see, the topology of such segments can greatly enhance or debilitate your network security, at least with respect to sniffers. In general, however, all machines are sent this request.
The workstation that is the intended recipient responds, forwarding its hardware address. The information is then sent down the wire (in packets) from the issuing workstation to the recipient. You might imagine that in this scenario (and from the instant that the recipient is known), all other workstations ignore the data being sent between the bona-fide sender and recipient. This is true; they do. However, they do not necessarily have to ignore this data, and if they don't, they can still hear it. In other words, any information traveling through the network is always "hear-able" by all interfaces within a segment (barring installation of controls to prevent it).
A sniffer is nothing more than hardware or software that hears (and does not ignore) all packets sent across the wire. In this respect, every machine and every router is a sniffer (or at least, each of these devices could be a sniffer). This information is then stored on some media and archived for later viewing.
NOTE: To use your machine as a sniffer, you will need either special software (a promiscuous driver for the Ethernet card) or a version of networking software that allows promiscuous mode.
NOTE: Think of the network as a dynamic atmosphere, such as a river. In that river, packets flow freely along the path of least resistance. A sniffer is an entity that sticks its hand into the river and filters the water through its fingers.
A sniffer can be (and usually is) a combination of both hardware and software. The software might be a general network analyzer enabled with heavy debugging options, or it might be a real sniffer.
A sniffer must be located within the same network block (or net of trust) as the network it is intended to sniff. With relatively few exceptions, that sniffer could be placed anywhere within that block (see Figure 12.3).
Figure 12.3.
Possible placements for sniffers.
Notice that one of the positions I have marked as a sniffer is located in the void (along the network wire instead of within a workstation). This is possible, though unlikely. Certain tools designed for network-traffic analysis can be spliced into the cable itself. These tools are quite expensive and not something that the average cracker would employ (however, I thought I should mention them).
Cross Reference: There are also devices that are referred to as cable sniffers, which are used to diagnose problems along network cable. One such product is called the Cable Sniffer by Macally. It can be used to sniff cable problems on AppleTalk networks. Their page is located at http://www.macally.com/.
Sniffers are a significant threat because of the following:
You are likely to find a sniffer almost anywhere. However, there are some strategic points that a cracker might favor. One of those points is anywhere adjacent to a machine or network that receives many passwords. This is especially so if the targeted machine is the gateway of a network, or a path of data going to or coming from the outside world. If your network goes out to the Internet (and that's really what I'm getting at here), the cracker will want to capture authentication procedures between your network and other networks. This could exponentially expand the cracker's sphere of activity.
Sniffers represent a high level of risk. In fact, the existence of a sniffer in itself shows a high level of compromise. In fact, if a sniffer has been placed on your network (by folks other than those authorized to undertake such an action), your network is already compromised. That is, taking the case study out of the LAN and into the Internet, if your Internet-enabled network has a sniffer, someone has breached your network security. One scenario is that he or she has come from the outside and placed a monitoring device on your network. The other scenario is that one of your own is up to no good. Either way, the situation is grave.
Security analysts characterize a sniffer attack as a second-level attack. The cracker has already worked his or her way into your network and now seeks to further compromise the system. To do so, he must begin by capturing all the user IDs and passwords. For that reason (and for the information a sniffer gathers), a sniffer represents an extremely high level of risk.
However, sniffers can catch more than simply user IDs and passwords; they can capture sensitive financial data (credit-card numbers), confidential information (e-mail), and proprietary information. Depending on the resources available to the cracker, a sniffer is capable of capturing nearly all traffic on a network.
NOTE: I do not believe that, in practice, any sniffer can catch absolutely all traffic on a network. This is because as the number of packets increases, the chances of lost packets is high. If you examine technical reports on sniffers, you will discover that at high speeds and in highly trafficked networks, a more-than negligible amount of data can be lost. This suggests that sniffers employed by the good guys might be vulnerable to attacks themselves. In other words, just how many packets per second can a sniffer take before it starts to fail in its fundamental mission? That is a subject probably worth investigating.Security technology has evolved considerably. Some operating systems now employ encryption at the packet level, and, therefore, even though a sniffer attack can yield valuable data, that data is encrypted. This presents an additional obstacle likely to be passed only by those with deeper knowledge of security, encryption, and networking.
Sniffers are designed as devices that can diagnose a network connection. You will remember that in Chapter 9, "Scanners," I referred to a UNIX command called traceroute. traceroute examines the route between two points and is used to determine whether problems exist along that route (for example, if one of the machines aÚ g that route has died).
Tools such as traceroute are sniffers of sorts. However, a hard-core sniffer is designed to examine the packet traffic at a very deep level. Again, this--like the scanner--has a perfectly legitimate purpose. Sniffers were designed by those aiding network engineers (and not for the purpose of compromising networks).
Some companies produce entire suites of sniffer applications designed to diagnose network problems. The leading company in this industry is Network General Corporation (NGC), which offers a wide variety of sniffer products, including
Sniffers now exist for every network platform, but even if they did not, they would still be a threat to you. Here is why: Sniffers sniff packets, not machines. Unless your network is entirely homogenous, a sniffer could exist there. As I pointed out, a sniffer need be only on a single node of a network (or at a gateway) to sniff traffic. This is because of the manner in which Ethernet broadcasts occur. Because the traffic is broadcasted to all nodes on a network segment, any platform that you have will do. Also, more sniffers for different operating systems emerge every few months; because source is now available for a wide variety of systems, it seems likely that trend will continue. Eventually, you will see the ultimate sniffer written for Windows 95 with some sort of VB front end. You can bet on it.
There have been many sniffer attacks executed over the Internet; these attacks were disparate in terms of target and scope. Consider this security advisement update:
1Naval Computer & Telecommunications Area Master Station LANT advisory.
Cross Reference: You can access the Naval Computer & Telecommunications Area Master Station LANT advisory at http://www.chips.navy.mil/chips/archives/94_jul/file14.html.
Naturally, institutions and private companies are reluctant to state what level of compromise might have occurred. But, there are many such victims:
Cross Reference: For more information about the Stanislaus incident, visit http://yahi.csustan.edu/studnote.html.For more information about the U.S. Army missile research lab and White Sands Missile Range incidents, see the GAO report at http://www.securitymanagement. com/library/000215.html.
Universities seem to be consistent targets, mainly because of the sheer volume of usernames and passwords that can be gleaned from such an attack. This also translates into bigger and more complex networks. Network administration in a university is quite a job, even if crackers aren't prowling around. How many times have you fingered an account at a university only to find that the target was discharged or graduated a year or more before? Two days before writing this chapter, I encountered exactly that situation. Except that the individual had been gone 18 months. Even so, his account was still active!
A sniffer attack is not as easy as you might think. It requires some knowledge of networking before a cracker can effectively launch one. Simply setting up a sniffer and leaving it will lead to problems because even a five-station network transmits thousands of packets an hour. Within a short time, the outfile of a sniffer could easily fill a hard disk drive to capacity (if you logged every packet).
To circumvent this problem, crackers typically sniff only the first 200-300 bytes of each packet. Contained within this portion is the username and password, which is really all most crackers want. However, it is true that you could sniff all the packets on a given interface; if you have the storage media to handle that kind of volume, you would probably find some interesting things.
There are many sniffers available on many platforms. As you might expect, the majority of these are commercial. Commercial sniffing applications are a good idea if you have a real need to diagnose your network (or catch a cracker). They are probably a poor idea if you simply want to learn about networking.
Gobbler, shown in Figure 12.4, is probably the best sniffer for someone wanting to learn a bit about network traffic. It was designed to work on the MS-DOS platform but can be run in Windows 95.
Figure 12.4.
Gobbler's opening screen.
Operation of Gobbler might seem a little confusing at first. There are no menus in the traditional sense (that is, the menus are not immediately apparent when you start the application); the application just pops up, as shown in Figure 12.4. (The menus are there; it is just that Gobbler is not the most user-friendly application.) Depending on what package you get, you may or may not receive documentation. If you do, it will be a PostScript document titled Paper.gs. Of the four locations where I have found Gobbler, only one has the document. It is the first of the addresses that follow.
Cross Reference: Gobbler is no longer widely distributed; these links are quite remote. Expect some download time. Gobbler can be found at
- http://www.cse.rmit.edu.au/~rdssc/courses/ds738/watt/other/gobbler.zip
- http://cosmos.ipc.chiba-u.ac.jp/~simizu/ftp.ipc.chiba-u.ac.jp/.0/network/noctools/sniffer/gobbler.zip
- ftp://ftp.mzt.hr/pub/tools/pc/sniffers/gobbler/gobbler.zip
- ftp://ftp.tordata.se/www/hokum/gobbler.zip
Press the F1 key after booting the application to view a legend that provides information about the program's functions (see Figure 12.5).
Figure 12.5.
Gobbler's function and navigation help screen.
Gobbler runs on any PC running DOS, Windows, Windows 95, and perhaps NT. It can be run from a single workstation, analyzing only local packets, or it can be used remotely over a network (this is an especially useful function).
Contained within the program are some fairly complex functions for packet filtering as well as an event-triggered mechanism (that is, one can specify a particular type of packet that must be encountered before the deep logging process starts or stops). Perhaps most importantly, Gobbler allows you to view both the source and destination addresses for each packet without further effort (these are printed to the screen in a very convenient manner).
The program allows you to view the recording process as it happens. This is a vital element of its usefulness. As noted in one of the case studies presented with the application:
1T.V. Rijn and J.V. Oorschot, The Gobbler, An Ethernet Troubleshooter/Protocol Analyzer. November 29, 1991. Delft University of Technology, Faculty of Electrical Engineering, the Netherlands.
A freeware packet sniffer written in C for Ethernet and token ring networks, ETHLOAD runs well atop or in conjunction with any of the following interfaces:
Further, it analyzes the following protocols:
One thing that is not available in the standard distribution is the source code. This is unfortunate because some time ago, the source was available. However, as the authors explain:
What is interesting is that the program did have the capability to sniff rlogin and Telnet sessions, though only with a special key that had to be obtained from the author. As one might expect, even when this key was available, the author restricted its access to those who could provide some form of official certification.
For a free sniffer executable on a DOS/Novell platform, ETHLOAD is probably the most comprehensive currently available (this is certainly so for the PC platforms). It is also more easily found than others (altavista.digital.com returns approximately one hundred instances of the file name, and more than half of those sites have the product).
Cross Reference: Here are a few sites that offer ETHLOAD:
- ftp://oak.oakland.edu/SimTel/msdos/lan/ethld104.zip
- http://www.med.ucalgary.ca:70/1/ftp/dos/regular
- ftp://ftp.vuw.ac.nz/simtel/msdos/lan/ethld104.zip
- http://www.apricot.co.uk/ftp/bbs/atsbbs/allfiles.htm
Netman is a little different from ETHLOAD in that you can obtain the source, although the process is more complex than "ask and ye shall receive." It involves money ($500 for educational institutions, $1,000 for private firms), and the development team makes it clear that that source is not to be used for commercial purposes.
The team at Curtin University has developed a whole suite of applications in the Netman project:
Etherman is of main interest in tracing Ethernet activity. It is important to note that this tool is no ordinary ASCII-to-outfile packet sniffer. As the authors explain in Homebrew Network Monitoring: A Prelude to Network Management, Etherman takes a whole new approach that is completely distinct from its counterparts:
True to their claims, these individuals created an extraordinary tool. The program presents a black screen on which addresses, traffic, and interfaces are all represented as points within the network (connection points or flows of data between these are represented in red). This accurate graphical model is altered as traffic varies.
The entire suite of applications constitutes a formidable arsenal for network analysis and management. In the hands of a cracker, this suite could prove quite a weapon. However, the main features of the Etherman program, at least, run in X. It is extremely unlikely that a cracker would be running X apps on your network without your knowledge. If this is going on, you better wake up and mind your network; your problems are deeper than a sniffer.
Cross Reference: The Netman project, papers, and all binaries for these programs are located at http://www.cs.curtin.edu.au/~netman/.
NOTE: The Netman suite of applications was reportedly coded on the Sun and DEC platforms (SPARC and Decstation 5000, respectively). Information about porting is scarce, but this much is certain: This application runs only on UNIX platforms. Moreover, remember when I suggested that some sniffers might lose data on high-speed, high-volume networks? Packetman is apparently one such application, although the problem is reportedly limited to the SunOS platform. This application is probably the most functional sniffer suite for the UNIX platform (if not in terms of superb functionality, at least in design).
Esniff.c is a sniffer program that is always distributed in source form (C language), designed primarily to sniff packet traffic on the SunOS platform. It is probably the most popular among crackers. It is already coded to capture only the beginning portion of each packet (and thereby capture user login IDs and passwords).
Cross Reference: Esniff.c is available at many locations, including
- ftp.infonexus.com
- http://pokey.nswc.navy.mil/Docs/Progs/ensnif.txt
- http://www.catch22.com/Twilight.NET/phuncnet/hacking/proggies/sniffers/
Sunsniff is also designed specifically for the SunOS platform. It consists of 513 lines of C source, coded reportedly by crackers who wish to remain anonymous. It works reasonably well on Sun, and is probably not easily portable to another flavor. This program is good for experimentation.
Cross Reference: Sunsniff is available at
- www.catch22.com/Twilight.NET/phuncnet/hacking/proggies/sniffers/
- http://mygale.mygale.org/08/datskewl/elite/
- http://hacked-inhabitants.com/warez/SUNSNIFF.C
This program's name pretty much says it all. It consists of 175 lines of C code, distributed primarily at cracker sites on the Net. This program is Linux specific. It is another utility that is great for experimentation on a nice Sunday afternoon; it's a free and easy way to learn about packet traffic.
Cross Reference: linux_sniffer.c is available at
- www.catch22.com/Twilight.NET/phuncnet/hacking/proggies/sniffers/
- http://mygale.mygale.org/08/datskewl/elite/
- http://www.hacked-inhabitants.com/warez/
This C source (159 lines, excluding comments) is distributed primarily at cracker sites. When compiled, it runs as a NIT (Network Interface Tap) sniffer. It is yet another SunOS-only utility. The authors anonymously claim that the utility is:
I would closely examine the source before employing this utility. This utility emerged from the back alleys of the Net.
Cross Reference: Nitwit.c can be found at www.catch22.com/Twilight.NET/phuncnet/hacking/proggies/sniffers/nitwit.c.
The short answer to this question is: You don't. Here lies one of the reasons sniffers are so threatening to security. They are largely passive applications and generate nothing. In other words, they leave no trace on the system.
One way to detect a sniffer is to search all current processes being run. This isn't entirely reliable, of course, but you can at least determine whether a process is being run from your machine. Commands differ from platform to platform in performing this operation. Those with DOS, Windows for Workgroups, or Windows 95 might have a problem. However, those using UNIX or Windows NT can easily obtain a list of current processes. In UNIX, issue the following command:
ps -aux
or
ps -augx
This command results in a listing of all processes, who initiated those processes, what percentage of the CPU those processes are using, what percentage of memory, and so on. The output comes in standard table form on STDOUT. If a process is still running, it should be listed here (unless your ps or other binaries have been trojaned).
Another method is to go searching for the sniffer. There are only so many sniffers in existence. Chances are a cracker is using a freeware version. There is a possibility that the cracker has written his or her own. In this case, you are in trouble and will spend much time reconciling your directories. This is a complicated procedure, and I am unaware of a utility that does expressly this. On the UNIX platform, you likely will have to hack a program for yourself.
NOTE: Programs like ps (in fact, most programs) can be hidden from the ps query by changing their argv[0] (their first argument) to the name of a program one that is innocuous and not so suspicious.
NOTE: Directory reconciliation is a fancy way of saying you need to perform frequent backups (ideally, once a day). The trick is to hack a program that takes the list of files on each backup and compares them to the backup on the following day. Include a type of file field, which contains the information you normally glean from the file command. This command reports the status of the file (whether it is binary, text, sound, and so on). If a file in a user's directory was a compiled binary one day and a shell script the next, it might not necessarily mean anything is wrong, but it is worth noting. A number of programs can help you per-form file reconciliation and are treated in Chapter 11, "Trojans." Some of these programs are Tripwire, ATP, and Hobgoblin.
Some utilities, however, can identify whether your system has been thrown into promiscuous mode. These can at least detect whether a running sniffer would even work under your current configuration. Nitwit.c is one such utility.
Foiling a sniffer attack is not a difficult task. You have quite a few choices and what you pick will probably depend on how paranoid you truly are. The rest will depend on cost.
Your main concern is probably the transmission of truly sensitive data (namely, user IDs and passwords). Some of these cross the network in plain (or clear) text and, when captured with a sniffer, are easily read. Solutions to this problem are straightforward and cost effective.
At various points in this book, I mention a product called Secure Shell, or SSH. SSH is a protocol that provides secure communications in an application environment such as Telnet. It is built on the client/server model, as are most protocols out there. The official port that the SSH server binds to is 22. Connections are negotiated using an algorithm from RSA. After the authentication procedure is complete, all subsequent traffic is encrypted using IDEA technology. This is typically strong encryption and is suitable for just about any nonsecret, nonclassified communication.
For quite some time, the original SSH has been lauded (rightly so) for being the chief communications protocol that provided security by encrypted session. However, that all changed in mid-1996. SSH forged an alliance with Data Fellows, and F-SSH currently provides high-level, military-grade encryption to communication sessions. It provides the strongest encryption available to the general public for communications across TCP/IP.
If you employ F-SSH on your site, usernames and passwords become less of an issue. To my knowledge, there have been no instances of anyone cracking such an encryption scheme. If you employ this product, even if a sniffer is present, the value of the information gleaned would be negligible. The hard part is getting your staff to use it.
People sometimes receive new policies and authentication procedures poorly. In short, you might have to demonstrate to your local users exactly how easy it is to use SSH.
Both free and commercial versions of SSH and F-SSH exist. The free version is a UNIX utility; commercial versions for Windows 3.11, Windows 95, and Windows NT are available.
The generally accepted way to defeat sniffer attacks is to employ safe topology. That sounds easy enough, but involves some cost.
Are you familiar with that puzzle game that consists of a series of numbered tiles? The object of the game is to arrange the numbers so they appear in sequential, ascending order using the fewest possible moves. When working with network topology (under cost constraints by management), you are playing a game much like the tile game.
Here are the rules:
That established, let's have a look. The first point is this: a network block should be composed of only those machines that need to trust one another. These typically belong in the same room or, at worst, within the same office. Your accounting staff, for example, should be bunched together in some portion of the building (see Figure 12.6).
Figure 12.6.
The accounting office.
From the diagram in Figure 12.6, you can immediately see one difference in this configuration as compared to the others earlier in this chapter. Notice that each of the stations is hardwired to the hub. (There is no closed-circuit wrap, like you often see in small Novell networks. I see that kind of configuration all the time in medical and legal offices.) Furthermore, the hub is wired to a switch. The major difference is that because the segment is hardwired in this fashion, packets can only be sniffed within that network segment. Thus, the remaining portion of the network (beyond the switch) is protected from sniffing activity. This technique is commonly referred to as compartmentalization or segmentation.
TIP: You can also use bridges or routers to perform this segmentation. This may be more suitable, depending upon your configuration and finances. In fact, an older PC or workstation can be made to act as a bridge or a router.
In segmentation, costs rise dramatically. Suppose you have 50 departments. Does that mean you need 50 hubs, 50 switches, and a router to tie them together? Possibly. It depends on how paranoid you really are. If you are doing sensitive work, then yes, you will be spending money on hardware. But consider the advantages: If an evil accounting employee wants to plant a sniffer, he can get no more than he could by physically tampering with his coworker's workstation. Moreover, if a sniffer is found on one of the three stations in accounting, there are only a limited number of individuals who could have placed it there.
The problem is a matter of trust. Some machines must trust one another in order to traffic information between themselves. Your job as a system administrator is to determine ways in which to create the fewest trust relationships possible. In this way, you build a framework that can tell you when a sniffer is placed, where it is, and who could have placed it.
The problem is, in most offices, there is no real level of trust. The average American business is in the business of making money. Tech support is expensive and so is the downtime to restring a network. Additionally, there can be serious costs involved in that restringing. What if all the wiring is embedded in the walls? These are all issues that you must consider. In legacy networks, these are real problems.
Also, you must consider the level of risk. What are you protecting? What are you planning to do regarding the Internet? These are the real issues. If you intend to connect your LAN to the Net, a firewall is not going to be enough. Relying solely on a firewall is a bad idea because new cracking techniques are always emerging. Are firewalls impenetrable? Vendors say yes, as long as they are properly configured. However, think about that statement for a moment. There was a time, not long ago, when shadowed password schemes were touted as pretty close to infallible (in spite of the fact that everyone deep in security knew that NIS would remain a weakness that could render the best shadowing a wet noodle). Crackers can already scan behind a firewall and determine the services running there. That is the first and most important step.
It will not be long before firewalls get cracked, so be prepared. Your first concern should be the worst case: If an intruder cuts through your armor, how far can he or she get? Try to think of it in terms of a path or a trajectory. Starting at your Web server and working your way back, how deep do the various levels of trust go? For example, the Web server probably trusts one machine, which we'll call workstation1. How many machines does workstation1 trust? How many of those machines trust others? In other words, worst case scenario, where will the cracker finally run out of machines to compromise?
Your job is to prevent that worst-case scenario from becoming a disaster. You do so by ensuring that if an intruder places a sniffer, that sniffing activity is confined to a very small area.
If I ran a large LAN connected to the Net, I would be sniffing the traffic on it. There are products that can reliably and conveniently present the results of such sniffing in tabular form. A good storage device, such as a Jazz drive, makes an excellent target to save those sniffer logs.
In this chapter, you learned a bit about sniffers:
I assert that you can benefit greatly by running a sniffer on your network, even if only for a day. This will familiarize you with what a cracker is facing to implement such an attack. Also, after you are proficient with a sniffer, you can see for yourself what type of information can actually be gleaned from your particular network configuration.
Lastly, sniffer or no sniffer, trace the levels and relationships of trust on your network. You might be surprised to find that this path extends through the larger portion of your network for one reason or another. This becomes more complicated, depending on how many interfaces you are running and how many protocols run on them. For example, if your firm is running Novell in one area, AppleTalk in another, TCP/IP in another, DECnet in another, NFS in another, and so forth, you have your job cut out for you. Starting at any given point, how far can you travel before you reach a trust roadblock?
Cross Reference: Levels of trust and relationships between network segments will be examined further in Chapter 28, "Spoofing Attacks." Spoofing relies almost solely on trust relationships and has little to do with passwords. (After all, who needs a password if two machines already trust one another?)
These considerations are all relevant to the sniffer issue. In closing, sniffers are very powerful tools for crackers, but only if you let them be. Moreover, if you find one on your network, do not immediately remove it. Instead, install one of your own and find out who is pulling the strings. Successful conclusions to network break-ins almost never begin with confrontations. They begin with stealth. You cannot go to the halls of justice without evidence.