One common security truism is "Once you have physical access to a box, all bets are off." This is a good beginning assumption for this section. If an attacker has physical access to a computer, router, switch, firewall, or other device, your security options are amazingly limited. Networking devices, with few exceptions, can have their passwords reset by attaching to their console port. Hosts can be booted with a special floppy disk or CD-ROM designed to circumvent most host security on the device.
This book does not cover physical security issues in detail. Topics such as disaster recovery, site selection, and so on are not discussed at all. However, as a network designer, you must know where you are relying on physical security to augment or support your network security. There are some rules you can follow to improve your security:
The rest of this section examines these seven areas.
Control Physical Access to Facilities
Effectively controlling physical access to your organization's facilities should be the single top concern for both your physical security staff and you, the network designer. Most organizations utilize one of three mechanisms to implement physical security (presented in increasing order of security):
The most common physical security control, particularly in smaller organizations, is traditional lock-and-key access. For this method, individuals who need access to certain rooms or buildings are given keys for access. This option has the following benefits:
However, there are also several drawbacks:
Key Card Access
More common in larger organizations, key card access can alleviate some of the management problems associated with lock-and-key access and can provide increased security measures. Key card access can take the form of a magnetic card reader or a smart card. All of these systems have the same basic pros and cons once you eliminate the technical differences of the technology. These are the benefits of a key card system:
The drawbacks to a key card system are as follows:
Key Card Access with Turnstile
Although most often associated with ballparks and stadiums, turnstile access with a key card can be one of the most secure methods of controlling physical access to a building. For this method, a key card is used to activate the turnstile and allow one person into the building. These systems are most common in large multifloor buildings, where access can be controlled at the ground floor. In the following list, you can see that this option has all the benefits of the previous option plus more.
The drawbacks of a system such as this are as follows:
Solving the Single-Factor Identity Problem
A second factor can be added to either of the previous key card authentication processes. The first option is to put a personal identification number (PIN) code reader at every location where there is a card reader. After using their key card, employees must enter a PIN to unlock the door. Another option is to use some form of biometric authentication. Biometric authentication could be used as either the second factor in a key card system or the principal factor in a biometric system. In the second case, users would enter a PIN after successful biometric authentication. See Chapter 4, "Network Security Technologies," for the pros and cons of biometric authentication. Both of these alternatives add cost to the system and inconvenience for users.
Control Physical Access to Data Centers
Data-center access can utilize any of the preceding mechanisms in addition to PIN-reader-only access. The important difference with data-center access is that you are often dealing with a smaller set of operators, so issues around key management are somewhat reduced.
I once had the pleasure of experiencing a physical security audit by a client who was considering using a facility in one of my previous jobs. Needless to say, it didn't go well. One of the auditors was able to gain access to the building by tailgating. Upon entering, he asked to see the "secure" data center we had advertised. Upon reaching the entrance to the secure room, he stood on a chair and pushed up the ceiling tile outside the room. He discovered that the walls to our data center extended only 12 inches beyond the ceiling tiles, allowing access if someone climbed over them.
In the context of this discussion, data center refers to any location where centralized network resources are stored. This could include traditional data centers, wiring closets, coat closets, or someone's desk. It all depends on the size of the facility and the way it is organized.
Some ultrasecure data centers utilize sets of cameras, key card access, biometrics, and "man-traps" to catch anyone illegally trying to gain access to the room.
Separate Identity Mechanisms for Insecure Locations
Although identity design considerations are discussed in more detail in Chapter 9, "Identity Design Considerations," from a physical security perspective, it is important to ensure that passwords in physically insecure locations are not the same as those used in secure locations.
Often an organization will utilize common authentication mechanisms for the various systems that must access network resources. For example, SNMP community strings or Telnet/ SSH passwords might be set the same on all devices. From a pure security perspective, it is preferable to use two-factor authentication, when available, for each user who accesses the network device. Although this might be possible for users, it is often impossible for software management systems, which need to run scripts to make changes on several machines at once. For optimal security, different passwords should be used on each device, but this is often operationally impossible for large networks.
Therefore, at a minimum, organize your common passwords so that they are never used on systems in physically insecure locations. For example, assume you have 3 main locations (with data centers) to your organization and 10 remote sites (considered insecure). In this case, only use your shared passwords on the main sites and ensure that the passwords for each of the remote systems are unique per site at a minimum and per device ideally. As the number of insecure locations increases into the hundreds or thousands, this becomes impossible; refer to the "Business Needs" section of Chapter 2, "Security Policy and Operations Life Cycle," for guidance on calculating the costs and benefits of this and any other difficult security measure. (People generally don't compute cost/benefit on easy and cheap security measures.)
Prevent Password Recovery Mechanisms in Insecure Locations
Some devices have controls to prevent the recovery of passwords in the event that an attacker has physical access to your system. For example, on some newer Cisco routers and switches, the command is as follows:
Router(config)# no service password-recovery
When this command is entered on a router or a switch, interrupting the boot process only allows the user to reset the system to its factory default configuration. Without this command, the attacker could clear the password and have access to the original configuration. This is important because the original configuration might contain common passwords or community strings that would allow the attacker to go after other systems.
This would be particularly useful in insecure branch offices or other locations where the physical security of a network device cannot be assured.
Be Aware of Cable Plant Issues
In today's networks, there are two primary cable types: unshielded twisted pair (UTP) category 5 (or higher) and fiber optic. The risk of an attacker accessing your physical cabling is important to consider because that level of access often can bypass other security controls and provide the attacker with easy access to information (provided encryption is not used). UTP cable is very easy to tap, but it was thought years ago that fiber was immune to cable taps. We now know that this is not the case. The National Security Association (NSA) is rumored to have already tapped intercontinental network links by splicing into the cable; read about it at the following URL: http://zdnet.com.com/2100-11-529826.html.
It is also theorized that fiber cable could be bent far enough so that some light would escape if the outer layer of the cable is removed. With the right types of equipment, this information could then be read.
Additionally, if an attacker gains physical access to a wiring closet or the fiber cable as it runs in a cable tray above a drop ceiling, tapping the cable by installing couplers is another possibility.
All this being said, fiber is more secure than copper because the means to tap the signal are more expensive, difficult to execute, and often require interrupting the original flow of data to install. On the other hand, the means to tap a UTP signal can easily be purchased off of the Internet.
Be Aware of Electromagnetic Radiation
In 1985, the concerns of the paranoid among the security community were confirmed. Wim van Eck released a paper confirming that a well-resourced attacker can read the output of a cathode-ray tube (CRT) computer monitor by measuring the electromagnetic radiation (EMR) produced by the device. This isn't particularly easy to do, but it is by no means impossible. Wim's paper can be found here:
This form of attack is now commonly called van Eck phreaking. Additionally, in 2002, Markus Kuhn at the University of Cambridge published a similar method of reading data off of a CRT, this time by measuring the changes in the amount of light in a room. His paper can be found here:
And an easy-to-read FAQ on the topic can be found here:
A simple way to mitigate van Eck phreaking might just be to change the type of font you are using. Ross Anderson and Markus Kuhn did some excellent research on the topic:
I am certainly not recommending that all systems must address these sorts of security considerations, but it is good to know that such attacks are possible.
Be Aware of Physical PC Security Threats
Oftentimes, inexperienced network designers begin with an unacknowledged assumption that all the sensitive data within an organization is contained on servers. In reality, there is sensitive information about my company sitting on the laptop I am using to write this book, as well as on the servers. Like most employees at my company, server resources are used when necessary, but often interesting information is stored locally.
Several physical security issues manifest when you operate under the preceding assumption: