1.3 The Meaning of Computer Security
We have seen that any computer-related system has both theoretical and real weaknesses. The purpose of computer security is to devise ways to prevent the weaknesses from being exploited. To understand what preventive measures make the most sense, we consider what we mean when we say that a system is "secure."
We use the term "security" in many ways in our daily lives. A "security system" protects our house, warning the neighbors or the police if an unauthorized intruder tries to get in. "Financial security" involves a set of investments that are adequately funded; we hope the investments will grow in value over time so that we have enough money to survive later in life. And we speak of children's "physical security," hoping they are safe from potential harm. Just as each of these terms has a very specific meaning in the context of its use, so too does the phrase "computer security."
When we talk about computer security, we mean that we are addressing three important aspects of any computer-related system: confidentiality, integrity, and availability.
- Confidentiality ensures that computer-related assets are accessed only by authorized parties. That is, only those who should have access to something will actually get that access. By "access," we mean not only reading but also viewing, printing, or simply knowing that a particular asset exists. Confidentiality is sometimes called secrecy or privacy.
- Integrity means that assets can be modified only by authorized parties or only in authorized ways. In this context, modification includes writing, changing, changing status, deleting, and creating.
- Availability means that assets are accessible to authorized parties at appropriate times. In other words, if some person or system has legitimate access to a particular set of objects, that access should not be prevented. For this reason, availability is sometimes known by its opposite, denial of service.
Security in computing addresses these three goals. One of the challenges in building a secure system is finding the right balance among the goals, which often conflict. For example, it is easy to preserve a particular object's confidentiality in a secure system simply by preventing everyone from reading that object. However, this system is not secure, because it does not meet the requirement of availability for proper access. That is, there must be a balance between confidentiality and availability.
But balance is not all. In fact, these three characteristics can be independent, can overlap (as shown in Figure 1-3), and can even be mutually exclusive. For example, we have seen that strong protection of confidentiality can severely restrict availability. Let us examine each of the three qualities in depth.
Figure 1-3 Relationship Between Confidentiality, Integrity, and Availability.
You may find the notion of confidentiality to be straightforward: Only authorized people or systems can access protected data. However, as we see in later chapters, ensuring confidentiality can be difficult. For example, who determines which people or systems are authorized to access the current system? By "accessing" data, do we mean that an authorized party can access a single bit? the whole collection? pieces of data out of context? Can someone who is authorized disclose those data to other parties?
Confidentiality is the security property we understand best because its meaning is narrower than the other two. We also understand confidentiality well because we can relate computing examples to those of preserving confidentiality in the real world.
Integrity is much harder to pin down. As Welke and Mayfield [WEL90, MAY91, NCS91b] point out, integrity means different things in different contexts. When we survey the way some people use the term, we find several different meanings. For example, if we say that we have preserved the integrity of an item, we may mean that the item is
- modified only in acceptable ways
- modified only by authorized people
- modified only by authorized processes
- internally consistent
- meaningful and usable
Integrity can also mean two or more of these properties. Welke and Mayfield recognize three particular aspects of integrity—authorized actions, separation and protection of resources, and error detection and correction. Integrity can be enforced in much the same way as can confidentiality: by rigorous control of who or what can access which resources in what ways. Some forms of integrity are well represented in the real world, and those precise representations can be implemented in a computerized environment. But not all interpretations of integrity are well reflected by computer implementations.
Availability applies both to data and to services (that is, to information and to information processing), and it is similarly complex. As with the notion of confidentiality, different people expect availability to mean different things. For example, an object or service is thought to be available if
- It is present in a usable form.
- It has capacity enough to meet the service's needs.
- It is making clear progress, and, if in wait mode, it has a bounded waiting time.
- The service is completed in an acceptable period of time.
We can construct an overall description of availability by combining these goals. We say a data item, service, or system is available if
- There is a timely response to our request.
- Resources are allocated fairly so that some requesters are not favored over others.
- The service or system involved follows a philosophy of fault tolerance, whereby hardware or software faults lead to graceful cessation of service or to work-arounds rather than to crashes and abrupt loss of information.
- The service or system can be used easily and in the way it was intended to be used.
- Concurrency is controlled; that is, simultaneous access, deadlock management, and exclusive access are supported as required.
As you can see, expectations of availability are far-reaching. Indeed, the security community is just beginning to understand what availability implies and how to ensure it. A small, centralized control of access is fundamental to preserving confidentiality and integrity, but it is not clear that a single access control point can enforce availability. Much of computer security's past success has focused on confidentiality and integrity; full implementation of availability is security's next great challenge.
When we prepare to test a system, we usually try to imagine how the system can fail; we then look for ways in which the requirements, design, or code can enable such failures. In the same way, when we prepare to specify, design, code, or test a secure system, we try to imagine the vulnerabilities that would prevent us from reaching one or more of our three security goals.
It is sometimes easier to consider vulnerabilities as they apply to all three broad categories of system resources (hardware, software, and data), rather than to start with the security goals themselves. Figure 1-4 shows the types of vulnerabilities we might find as they apply to the assets of hardware, software, and data. These three assets and the connections among them are all potential security weak points. Let us look in turn at the vulnerabilities of each asset.
Figure 1-4 Vulnerabilities of Computing Systems.
Hardware is more visible than software, largely because it is composed of physical objects. Because we can see what devices are hooked to the system, it is rather simple to attack by adding devices, changing them, removing them, intercepting the traffic to them, or flooding them with traffic until they can no longer function. However, designers can usually put safeguards in place.
But there are other ways that computer hardware can be attacked physically. Computers have been drenched with water, burned, frozen, gassed, and electrocuted with power surges. People have spilled soft drinks, corn chips, ketchup, beer, and many other kinds of food on computing devices. Mice have chewed through cables. Particles of dust, and especially ash in cigarette smoke, have threatened precisely engineered moving parts. Computers have been kicked, slapped, bumped, jarred, and punched. Although such attacks might be intentional, most are not; this abuse might be considered "involuntary machine slaughter": accidental acts not intended to do serious damage to the hardware involved.
A more serious attack, "voluntary machine slaughter" or "machinicide," usually involves someone who actually wishes to harm the computer hardware or software. Machines have been shot with guns, stabbed with knives, and smashed with all kinds of things. Bombs, fires, and collisions have destroyed computer rooms. Ordinary keys, pens, and screwdrivers have been used to short-out circuit boards and other components. Devices and whole systems have been carried off by thieves. The list of the kinds of human attacks perpetrated on computers is almost endless.
In particular, deliberate attacks on equipment, intending to limit availability, usually involve theft or destruction. Managers of major computing centers long ago recognized these vulnerabilities and installed physical security systems to protect their machines. However, the proliferation of PCs, especially laptops, as office equipment has resulted in several thousands of dollars'worth of equipment sitting unattended on desks outside the carefully protected computer room. (Curiously, the supply cabinet, containing only a few hundred dollars' worth of pens, stationery, and paper clips, is often locked.) Sometimes the security of hardware components can be enhanced greatly by simple physical measures such as locks and guards.
Laptop computers are especially vulnerable because they are designed to be easy to carry. (See Sidebar 1-3 for the story of a stolen laptop.) Safeware Insurance reported 600,000 laptops stolen in 2003. Credent Technologies reported that 29 percent were stolen from the office, 25 percent from a car, and 14 percent in an airport. Stolen laptops are almost never recovered: The FBI reports 97 percent were not returned [SAI05].
Computing equipment is of little use without the software (operating system, controllers, utility programs, and application programs) that users expect. Software can be replaced, changed, or destroyed maliciously, or it can be modified, deleted, or misplaced accidentally. Whether intentional or not, these attacks exploit the software's vulnerabilities.
Sometimes, the attacks are obvious, as when the software no longer runs. More subtle are attacks in which the software has been altered but seems to run normally. Whereas physical equipment usually shows some mark of inflicted injury when its boundary has been breached, the loss of a line of source or object code may not leave an obvious mark in a program. Furthermore, it is possible to change a program so that it does all it did before, and then some. That is, a malicious intruder can "enhance" the software to enable it to perform functions you may not find desirable. In this case, it may be very hard to detect that the software has been changed, let alone to determine the extent of the change.
A classic example of exploiting software vulnerability is the case in which a bank worker realized that software truncates the fractional interest on each account. In other words, if the monthly interest on an account is calculated to be $14.5467, the software credits only $14.54 and ignores the $.0067. The worker amended the software so that the throw-away interest (the $.0067) was placed into his own account. Since the accounting practices ensured only that all accounts balanced, he built up a large amount of money from the thousands of account throw-aways without detection. It was only when he bragged to a colleague of his cleverness that the scheme was discovered.
Software is surprisingly easy to delete. Each of us has, at some point in our careers, accidentally erased a file or saved a bad copy of a program, destroying a good previous copy. Because of software's high value to a commercial computing center, access to software is usually carefully controlled through a process called configuration management so that software cannot be deleted, destroyed, or replaced accidentally. Configuration management uses several techniques to ensure that each version or release retains its integrity. When configuration management is used, an old version or release can be replaced with a newer version only when it has been thoroughly tested to verify that the improvements work correctly without degrading the functionality and performance of other functions and services.
Software is vulnerable to modifications that either cause it to fail or cause it to perform an unintended task. Indeed, because software is so susceptible to "off by one" errors, it is quite easy to modify. Changing a bit or two can convert a working program into a failing one. Depending on which bit was changed, the program may crash when it begins or it may execute for some time before it falters.
With a little more work, the change can be much more subtle: The program works well most of the time but fails in specialized circumstances. For instance, the program may be maliciously modified to fail when certain conditions are met or when a certain date or time is reached. Because of this delayed effect, such a program is known as a logic bomb. For example, a disgruntled employee may modify a crucial program so that it accesses the system date and halts abruptly after July 1. The employee might quit on May l and plan to be at a new job miles away by July.
Another type of change can extend the functioning of a program so that an innocuous program has a hidden side effect. For example, a program that ostensibly structures a listing of files belonging to a user may also modify the protection of all those files to permit access by another user.
Other categories of software modification include
- Trojan horse: a program that overtly does one thing while covertly doing another
- virus: a specific type of Trojan horse that can be used to spread its "infection" from one computer to another
- trapdoor: a program that has a secret entry point
- information leaks in a program: code that makes information accessible to unauthorized people or programs
More details on these and other software modifications are provided in Chapter 3.
Of course, it is possible to invent a completely new program and install it on a computing system. Inadequate control over the programs that are installed and run on a computing system permits this kind of software security breach.
This attack includes unauthorized copying of software. Software authors and distributors are entitled to fair compensation for use of their product, as are musicians and book authors. Unauthorized copying of software has not been stopped satisfactorily. As we see in Chapter 11, the legal system is still grappling with the difficulties of interpreting paper-based copyright laws for electronic media.
Hardware security is usually the concern of a relatively small staff of computing center professionals. Software security is a larger problem, extending to all programmers and analysts who create or modify programs. Computer programs are written in a dialect intelligible primarily to computer professionals, so a "leaked" source listing of a program might very well be meaningless to the general public.
Printed data, however, can be readily interpreted by the general public. Because of its visible nature, a data attack is a more widespread and serious problem than either a hardware or software attack. Thus, data items have greater public value than hardware and software because more people know how to use or interpret data.
By themselves, out of context, pieces of data have essentially no intrinsic value. For example, if you are shown the value "42," it has no meaning for you unless you know what the number represents. Likewise, "326 Old Norwalk Road" is of little use unless you know the city, state, and country for the address. For this reason, it is hard to measure the value of a given data item.
On the other hand, data items in context do relate to cost, perhaps measurable by the cost to reconstruct or redevelop damaged or lost data. For example, confidential data leaked to a competitor may narrow a competitive edge. Data incorrectly modified can cost human lives. To see how, consider the flight coordinate data used by an airplane that is guided partly or fully by software, as many now are. Finally, inadequate security may lead to financial liability if certain personal data are made public. Thus, data have a definite value, even though that value is often difficult to measure.
Typically, both hardware and software have a relatively long life. No matter how they are valued initially, their value usually declines gradually over time. By contrast, the value of data over time is far less predictable or consistent. Initially, data may be valued highly. However, some data items are of interest for only a short period of time, after which their value declines precipitously.
To see why, consider the following example. In many countries, government analysts periodically generate data to describe the state of the national economy. The results are scheduled to be released to the public at a predetermined time and date. Before that time, access to the data could allow someone to profit from advance knowledge of the probable effect of the data on the stock market. For instance, suppose an analyst develops the data 24 hours before their release and then wishes to communicate the results to other analysts for independent verification before release. The data vulnerability here is clear, and, to the right people, the data are worth more before the scheduled release than afterward. However, we can protect the data and control the threat in simple ways. For example, we could devise a scheme that would take an outsider more than 24 hours to break; even though the scheme may be eminently breakable (that is, an intruder could eventually reveal the data), it is adequate for those data because confidentiality is not needed beyond the 24-hour period.
Data security suggests the second principle of computer security.
- Principle of Adequate Protection: Computer items must be protected only until they lose their value. They must be protected to a degree consistent with their value.
This principle says that things with a short life can be protected by security measures that are effective only for that short time. The notion of a small protection window applies primarily to data, but it can in some cases be relevant for software and hardware, too.
Sidebar 1-4 confirms that intruders take advantage of vulnerabilities to break in by whatever means they can.
Figure 1-5 illustrates how the three goals of security apply to data. In particular, confidentiality prevents unauthorized disclosure of a data item, integrity prevents unauthorized modification, and availability prevents denial of authorized access.
Figure 1-5 Security of Data.
Data can be gathered by many means, such as tapping wires, planting bugs in output devices, sifting through trash receptacles, monitoring electromagnetic radiation, bribing key employees, inferring one data point from other values, or simply requesting the data. Because data are often available in a form people can read, the confidentiality of data is a major concern in computer security.
Data are not just numbers on paper; computer data include digital recordings such as CDs and DVDs, digital signals such as network and telephone traffic, and broadband communications such as cable and satellite TV. Other forms of data are biometric identifiers embedded in passports, online activity preferences, and personal information such as financial records and votes. Protecting this range of data types requires many different approaches.
Stealing, buying, finding, or hearing data requires no computer sophistication, whereas modifying or fabricating new data requires some understanding of the technology by which the data are transmitted or stored, as well as the format in which the data are maintained. Thus, a higher level of sophistication is needed to modify existing data or to fabricate new data than to intercept existing data. The most common sources of this kind of problem are malicious programs, errant file system utilities, and flawed communication facilities.
Data are especially vulnerable to modification. Small and skillfully done modifications may not be detected in ordinary ways. For instance, we saw in our truncated interest example that a criminal can perform what is known as a salami attack: The crook shaves a little from many accounts and puts these shavings together to form a valuable result, like the meat scraps joined in a salami.
A more complicated process is trying to reprocess used data items. With the proliferation of telecommunications among banks, a fabricator might intercept a message ordering one bank to credit a given amount to a certain person's account. The fabricator might try to replay that message, causing the receiving bank to credit the same account again. The fabricator might also try to modify the message slightly, changing the account to be credited or the amount, and then transmit this revised message.
Other Exposed Assets
We have noted that the major points of weakness in a computing system are hardware, software, and data. However, other components of the system may also be possible targets. In this section, we identify some of these other points of attack.
Networks are specialized collections of hardware, software, and data. Each network node is itself a computing system; as such, it experiences all the normal security problems. In addition, a network must confront communication problems that involve the interaction of system components and outside resources. The problems may be introduced by a very exposed storage medium or access from distant and potentially untrustworthy computing systems.
Thus, networks can easily multiply the problems of computer security. The challenges are rooted in a network's lack of physical proximity, use of insecure shared media, and the inability of a network to identify remote users positively.
Access to computing equipment leads to three types of vulnerabilities. In the first, an intruder may steal computer time to do general-purpose computing that does not attack the integrity of the system itself. This theft of computer services is analogous to the stealing of electricity, gas, or water. However, the value of the stolen computing services may be substantially higher than the value of the stolen utility products or services. Moreover, the unpaid computing access spreads the true costs of maintaining the computing system to other legitimate users. In fact, the unauthorized access risks affecting legitimate computing, perhaps by changing data or programs. A second vulnerability involves malicious access to a computing system, whereby an intruding person or system actually destroys software or data. Finally, unauthorized access may deny service to a legitimate user. For example, a user who has a time-critical task to perform may depend on the availability of the computing system. For all three of these reasons, unauthorized access to a computing system must be prevented.
People can be crucial weak points in security. If only one person knows how to use or maintain a particular program, trouble can arise if that person is ill, suffers an accident, or leaves the organization (taking her knowledge with her). In particular, a disgruntled employee can cause serious damage by using inside knowledge of the system and the data that are manipulated. For this reason, trusted individuals, such as operators and systems programmers, are usually selected carefully because of their potential ability to affect all computer users.
We have described common assets at risk. In fact, there are valuable assets in almost any computer system. (See Sidebar 1-5 for an example of exposed assets in ordinary business dealings.)
Next, we turn to the people who design, build, and interact with computer systems, to see who can breach the systems' confidentiality, integrity, and availability.