Much of this book has focused on technical issues in security and their technical solutions: firewalls, encryption techniques, and more. But many threats to security involve human or natural disasters, events that should also be addressed in the security plan. For this reason, in this section we consider how to cope with the nontechnical things that can go wrong. There are two pieces to the process of dealing with nontechnical problems: preventing things that can be prevented and recovering from the things that cannot be prevented. Physical security is the term used to describe protection needed outside the computer system. Typical physical security controls include guards, locks, and fences to deter direct attacks. In addition, there are other kinds of protection against less direct disasters, such as floods and power outages; these, too, are part of physical security. As we will see, many physical security measures can be provided simply by good common sense, a characteristic that Mark Twain noted "is a most uncommon virtue."
Computers are subject to the same natural disasters that can occur to homes, stores, and automobiles. They can be flooded, burned, melted, hit by falling objects, and destroyed by earthquakes, storms, and tornadoes. Additionally, computers are sensitive to their operating environment, so excessive heat or inadequate power is also a threat. It is impossible to prevent natural disasters, but through careful planning it is possible to reduce the damage they inflict. Some measures can be taken to reduce their impact. Because many of these perils cannot be prevented or predicted, controls focus on limiting possible damage and recovering quickly from a disaster. Issues to be considered include the need for offsite backups, the cost of replacing equipment, the speed with which equipment can be replaced, the need for available computing power, and the cost or difficulty of replacing data and programs.
Water from a natural flood comes from ground level, rising gradually, and bringing with it mud and debris. Often, there is time for an orderly shutdown of the computing system; at worst, the organization loses some of the processing in progress. At other times, such as when a dam breaks, a water pipe bursts, or the roof collapses in a storm, a sudden flood can overwhelm the system and its users before anything can be saved. Water can come from above, below, or the side. The machinery may be destroyed or damaged by mud and water, but most computing systems are insured and replaceable by the manufacturer. Managers of unique or irreplaceable equipment who recognize the added risk sometimes purchase or lease duplicate redundant hardware systems to ensure against disruption of service.
Even when the hardware can be replaced, we must be concerned about the stored data and programs. The system administrator may choose to label storage media in a way that makes it easy to identify the most important data. For example, green, yellow, and red labels may show which disks are the most sensitive, so that all red disks are moved from the data center during a storm. Similarly, large plastic bags and waterproof tape can be kept near important equipment and media; they are used to protect the hardware and storage media in case of a burst pipe or other sudden flood.
The real issue is protecting data and preserving the ability to compute. The only way to ensure the safety of data is to store backup copies in one or more safe locations.
Fire is more serious than water; often there is not as much time to react, and human lives are more likely to be in immediate danger. To ensure that system personnel can react quickly, every user and manager should have a plan for shutting down the system in an orderly manner. Such a process takes only a few minutes but can make recovery much easier. This plan should include individual responsibilities for all people: some to halt the system, others to protect crucial media, others to close doors on media cabinets. Provision should be made for secondary responsibilities, so that onsite staff can perform duties for those who are not in the office.
Water is traditionally used to put out fires, but it is not a good idea for use in computer rooms. In fact, more destruction can be the result of sprinklers than of the fires themselves. A fire sensor usually activates many sprinklers, dousing an entire room, even when the fire is merely some ignited paper in a wastebasket and of no threat to the computing system. Many computing centers use carbon dioxide extinguishers or an automatic system that sprays a gas such as Halon to smother a fire but leave no residue. Unfortunately, these gas systems work by displacing the oxygen in the room, choking the fire but leaving humans unable to breathe. Consequently, when these protection devices are activated, humans must leave, disabling efforts to protect media.
The best defense for situations like these is careful placement of the computing facility. A windowless location with fire-resistant access doors and nonflammable full-height walls can prevent some fires from spreading from adjacent areas to the computing room. With a fire-and smoke-resistant facility, personnel merely shut down the system and leave, perhaps carrying out the most important media.
Fire prevention is quite effective, especially because most computer goods are not especially flammable. Advance planning, reinforced with simulation drills, can help make good use of the small amount of time available before evacuation is necessary.
Other Natural Disasters
Computers are subject to storms, earthquakes, volcanoes, and similar events. Although not natural disasters, building collapse, explosion, and damage from falling objects can be considered in the same category. These kinds of catastrophes are difficult to predict or estimate.
But we know these catastrophes will occur. Security managers cope with them in several ways:
developing contingency plans so that people know how to react in emergencies and business can continue
insuring physical assetscomputers, buildings, devices, suppliesagainst harm
preserving sensitive data by maintaining copies in physically separated locations
Computers need their foodelectricityand they require a constant, pure supply of it. With a direct power loss, all computation ceases immediately. Because of possible damage to media by sudden loss of power, many disk drives monitor the power level and quickly retract the recording head if power fails. For certain time-critical applications, loss of service from the system is intolerable; in these cases, alternative complete power supplies must be instantly available.
Uninterruptible Power Supply
One protection against power loss is an uninterruptible power supply. This device stores energy during normal operation so that it can return the backup energy if power fails. One form of uninterruptible power supply uses batteries that are continually charged when the power is on but which then provide power when electricity fails. However, size, heat, flammability, and low output can be problems with batteries.
Some uninterruptible power supplies use massive wheels that are kept in continuous motion when electricity is available. When the power fails, the inertia in the wheels operates generators to produce more power. Size and limited duration of energy output are problems with this variety of power supply. Both forms of power supplies are intended to provide power for a limited time, just long enough to allow the current state of the computation to be saved so that no computation is lost.
Another problem with power is its "cleanness." Although most people are unaware of it, a variation of 10 percent from the stated voltage of a line is considered acceptable, and some power lines vary even more. A particular power line may always be 10 percent high or low. In many places, lights dim momentarily when a large appliance, such as an air conditioner, begins operation. When a large motor starts, it draws an exceptionally large amount of current, which reduces the flow to other devices on the line. When a motor stops, the sudden termination of draw can send a temporary surge along the line. Similarly, lightning strikes may send a momentary large pulse. Thus, instead of being constant, the power delivered along any electric line shows many brief fluctuations, called drops, spikes, and surges . A drop is a momentary reduction in voltage, and a spike or surge is a rise. For computing equipment, a drop is less serious than a surge. Most electrical equipment is tolerant of rather large fluctuations of current.
These variations can be destructive to sensitive electronic equipment, however. Simple devices called "surge suppressors" filter spikes from an electric line, blocking fluctuations that would affect computers. These devices cost from $20 to $100; they should be installed on every computer, printer, or other connected component. More sensitive models are typically used on larger systems.
As mentioned previously, a lightning strike can send a surge through a power line. To increase protection, personal computer users usually unplug their machines when they are not in use, as well as during electrical storms. Another possible source of destruction is lightning striking a telephone line. Because the power surge can travel along the phone line and into the computer or peripherals, the phone line should be disconnected from the modem during storms. These simple measures may save much work as well as valuable equipment.
Because computers and their media are sensitive to a variety of disruptions, a vandal can destroy hardware, software, and data. Human attackers may be disgruntled employees, bored operators, saboteurs, people seeking excitement, or unwitting bumblers. If physical access is easy to obtain, crude attacks using axes or bricks can be very effective. One man recently shot a computer that he claimed had been in the shop for repairs many times without success.
Physical attacks by unskilled vandals are often easy to prevent; a guard can stop someone approaching a computer installation with a threatening or dangerous object. When physical access is difficult, more subtle attacks can be tried, resulting in quite serious damage. People with only some sophisticated knowledge of a system can short-circuit a computer with a car key or disable a disk drive with a paper clip. These items are not likely to attract attention until the attack is completed.
Unauthorized Access and Use
Films and newspaper reports exaggerate the ease of gaining access to a computing system. Still, as distributed computing systems become more prevalent, protecting the system from outside access becomes more difficult and more important. Interception is a form of unauthorized access; the attacker intercepts data and either breaks confidentiality or prevents the data from being read or used by others. In this context, interception is a passive attack. But we must also be concerned about active interception, in the sense that the attacker can change or insert data before allowing it to continue to its destination.
It is hard to steal a large mainframe computer. Not only is carrying it away difficult, but finding a willing buyer and arranging installation and maintenance also require special assistance. However, printed reports, tapes, or disks can be carried easily. If done well, the loss may not be detected for some time.
Personal computers, laptops, and personal digital assistants (PDAs, such as Palms or Blackberries) are designed to be small and portable. Diskettes and tape backup cartridges are easily carried in a shirt pocket or briefcase. Computers and media that are easy to carry are also easy to conceal.
We can take one of three approaches to preventing theft: preventing access, preventing portability, or detecting exit.
The surest way to prevent theft is to keep the thief away from the equipment. However, thieves can be either insiders or outsiders. Therefore, access control devices are needed both to prevent access by unauthorized individuals and to record access by those authorized. A record of accesses can help identify who committed a theft.
The oldest access control is a guard, not in the database management system sense we discussed in Chapter 6 but rather in the sense of a human being stationed at the door to control access to a room or to equipment. Guards offer traditional protection; their role is well understood, and the protection they offer is adequate in many situations. However, guards must be on duty continuously in order to be effective; providing breaks implies at least four guards for a 24-hour operation, with extras for vacation and illness. A guard must personally recognize someone or recognize an access token, such as a badge. People can lose or forget badges; terminated employees and forged badges are also problems. Unless the guard makes a record of everyone who has entered a facility, there is no way to know who (employee or visitor) has had access in case a problem is discovered.
The second oldest access control is a lock. This device is even easier, cheaper, and simpler to manage than a guard. However, it too provides no record of who has had access, and difficulties arise when keys are lost or duplicated. At computer facilities, it is inconvenient to fumble for a key when your hands are filled with tapes or disks, which might be ruined if dropped. There is also the possibility of piggybacking: a person walks through the door that someone else has just unlocked. Still, guards and locks provide simple, effective security for access to facilities such as computer rooms.
More exotic access control devices employ cards with radio transmitters, magnetic stripe cards (similar to 24-hour bank cards), and smart cards with chips containing electronic circuitry that makes them difficult to duplicate. Because each of these devices interfaces with a computer, it is easy for the computer to capture identity information, generating a list of who entered and left the facility, when, and by which routes. Some of these devices operate by proximity, so that a person can carry the device in a pocket or clipped to a collar; the person obtains easy access even when hands are full. Because these devices are computer controlled, it is easy to invalidate an access authority when someone quits or reports the access token lost or stolen.
The nature of the application or service determines how strict the access control needs to be. Working in concert with computer-based authentication techniques, the access controls can be part of defense in depthusing multiple mechanisms to provide security.
Portability is a mixed blessing. We can now carry around in our pockets devices that provide as much computing power as mainframes did twenty years ago. Portability is in fact a necessity in devices such as PDAs and mobile phones. And we do not want to permanently affix our personal computers to our desks, in case they need to be removed for repair or replacement. Thus, we need to find ways to enable portability without promoting theft.
One antitheft device is a pad connected to cable, similar to those used to secure bicycles. The pad is glued to the desktop with extremely strong adhesive. The cables loop around the equipment and are locked in place. Releasing the lock permits the equipment to be moved. An alternative is to couple the base of the equipment to a secure pad, in much the same way that televisions are locked in place in hotel rooms. Yet a third possibility is a large, lockable cabinet in which the personal computer and its peripherals are kept when they are not in use. Some people argue that cables, pads, and cabinets are unsightly and, worse, they make the equipment inconvenient to use.
Another alternative is to use movement-activated alarm devices when the equipment is not in use. Small alarms are available that can be locked to a laptop or PDA. When movement is detected, a loud, annoying whine or whistle warns that the equipment has been disturbed. Such an alarm is especially useful when laptops must be left in meeting or presentation rooms overnight or during a break. Used in concert with guards, the alarms can offer reasonable protection at reasonable cost.
For some devices, protection is more important than detection. We want to keep someone from stealing certain systems or information at all costs. But for other devices, it may be enough to detect that an attempt has been made to access or steal hardware or software. For example, chaining down a disk makes it unusable. Instead, we try to detect when someone tries to leave a protected area with the disk or other protected object. In these cases, the protection mechanism should be small and unobtrusive.
One such mechanism is similar to the protection used by many libraries, bookstores, or department stores. Each sensitive object is marked with a special label. Although the label looks like a normal pressure-sensitive one, its presence can be detected by a machine at the exit door if the label has not been disabled by an authorized party, such as a librarian or sales clerk. Similar security code tags are available for vehicles, people, machinery, and documents. Some tags are enabled by radio transmitters. When the detector sounds an alarm, someone must apprehend the person trying to leave with the marked object.