< Free Open Study > |
4.5 User AuthenticationAn operating system bases much of its protection on knowing who a user of the system is. In real-life situations, people commonly ask for identification from people they do not know: A bank employee may ask for a driver's license before cashing a check, library employees may require some identification before charging out books, and immigration officials ask for passports as proof of identity. In-person identification is usually easier than remote identification. For instance, some universities do not report grades over the telephone because the office workers do not necessarily know the students calling. However, a professor who recognizes the voice of a certain student can release that student's grades. Over time, organizations and systems have developed means of authentication, using documents, voice recognition, fingerprint and retina matching, and other trusted means of identification. In computing, the choices are more limited and the possibilities less secure. Anyone can attempt to log in to a computing system. Unlike the professor who recognizes a student's voice, the computer cannot recognize electrical signals from one person as being any different from those of anyone else. Thus, most computing authentication systems must be based on some knowledge shared only by the computing system and the user. Authentication mechanisms use any of three qualities to confirm a user's identity:
Two or more forms can be combined for more solid authentication; for example, a bank card and a PIN combine something the user has with something the user knows. Use of PasswordsThe most common authentication mechanism for user to operating system is a password , a "word" known to computer and user. Although password protection seems to offer a relatively secure system, human practice sometimes degrades its quality. In this section we consider passwords, criteria for selecting them, and ways of using them for authentication. We conclude by noting other authentication techniques and by studying problems in the authentication process, notably Trojan horses masquerading as the computer authentication process.
Passwords are mutually agreed-upon code words, assumed to be known only to the user and the system. In some cases a user chooses passwords; in other cases they are assigned by the system. The length and format of the password also vary from one system to another. The use of passwords is fairly straightforward. A user enters some piece of identification, such as a name or an assigned user ID; this identification can be available to the public or easy to guess because it does not provide the real security of the system. The system then requests a password from the user. If the password matches that on file for the user, the user is authenticated and allowed access to the system. If the password match fails, the system requests the password again, in case the user mistyped . Loose-Lipped SystemsSo far the process seems secure, but in fact it has some vulnerabilities. To see why, consider the actions of a would-be intruder. Authentication is based on knowing the < name, password > pair. A complete outsider is presumed to know nothing of the system. Suppose the intruder attempts to access a system in the following manner. (In the following examples, the system messages are in upper case, and the user's responses are in lower case.) WELCOME TO THE XYZ COMPUTING SYSTEMS ENTER USER NAME: adams INVALID USER NAMEUNKNOWN USER ENTER USER NAME: We assumed that the intruder knew nothing of the system, but without having to do much, the intruder found out that adams is not the name of an authorized user. The intruder could try other common names, first names, and likely generic names such as system or operator to build a list of authorized users. An alternative arrangement of the login sequence is shown below. WELCOME TO THE XYZ COMPUTING SYSTEMS ENTER USER NAME: adams ENTER PASSWORD: john INVALID ACCESS ENTER USER NAME: This system notifies a user of a failure only after accepting both the user name and the password. The failure message should not indicate whether it is the user name or password that is unacceptable. In this way, the intruder does not know which failed. These examples also gave a clue as to which computing system is being accessed. The true outsider has no right to know that, and legitimate insiders already know what system they have accessed. In the example below, the user is given no information until the system is assured of the identity of the user. ENTER USER NAME: adams ENTER PASSWORD: john INVALID ACCESS ENTER USER NAME: adams ENTER PASSWORD: johnq WELCOME TO THE XYZ COMPUTING SYSTEMS Additional Authentication InformationIn addition to the name and password, we can use other information available to authenticate users. Suppose Adams works in the accounting department during the shift between 8:00 a.m. and 5:00 p.m., Monday through Friday. Any legitimate access attempt by Adams should be made during those times, through a terminal in the accounting department offices. By limiting Adams to logging in under those conditions, the system protects against two problems:
Limiting users to certain terminals or certain times of access can cause complications (as when a user legitimately needs to work overtime, or a person has to access the system while out of town on a business trip). However, some companies use these authentication techniques because the added security they provide outweighs inconveniences. Attacks on PasswordsHow secure are passwords themselves ? Passwords are somewhat limited as protection devices because of the relatively small number of bits of information they contain. Here are some ways you might be able to determine a user's password.
These suggestions are arranged in decreasing order of difficulty; the later ones are, or at least should be, less likely to succeed. Exhaustive AttackIn an exhaustive or brute force attack , the attacker tries all possible passwords, usually in some automated fashion. Of course, the number of possible passwords depends on the implementation of the particular computing system. For example, if passwords are words consisting of the 26 characters AZ and can be of any length from 1 to 8 characters , there are 26 1 passwords of 1 character, 26 2 passwords of 2 characters, and 26 8 passwords of 8 characters. Therefore, the system as a whole has 26 1 + 26 2 + ... + 26 8 = 26 9 1 5 * 10 12 or five million million possible passwords. That number seems intractable enough. If we were to use a computer to create and try each password at a rate of one password per millisecond, it would take on the order of 150 years to test all passwords. But if we can speed up the search to one password per microsecond, the work factor drops to about two months. This amount of time is reasonable if the reward is large. For instance, an intruder may try to break the password on a file of credit card numbers or bank account information. But the break-in time can be made more tractable in a number of ways. Searching for a single particular password does not necessarily require all passwords to be tried; an intruder needs to try only until the correct password is identified. If the set of all possible passwords were evenly distributed, an intruder would likely need to try only half of the password space: the expected number of searches to find any particular password. However, an intruder can also use to advantage the fact that passwords are not evenly distributed. Because a password has to be remembered , people tend to pick simple passwords. This feature reduces the size of the password space. Probable PasswordsThink of a word. Is the word you thought of long? Is it uncommon? Is it hard to spell or to pronounce ? The answer to all three of these questions is probably no. Penetrators searching for passwords realize these very human characteristics and use them to their advantage. Therefore, penetrators try techniques that are likely to lead to rapid success. If people prefer short passwords to long ones, the penetrator will plan to try all passwords but to try them in order by length. There are only 26 1 + 26 2 + 26 3 = 18,278 passwords of length 3 or less. At the assumed rate of one password per millisecond, all of these passwords can be checked in 18.278 seconds, hardly a challenge with a computer. Even expanding the tries to 4 or 5 characters raises the count only to 475 seconds (about 8 minutes) and 12,356 seconds (about 3.5 hours), respectively. This analysis assumes that people choose passwords such as vxlag and msms as often as they pick enter and beer . However, people tend to choose names or words they can remember. Many computing systems have spelling checkers that can be used to check for spelling errors and typographic mistakes in documents. These spelling checkers sometimes carry online dictionaries of the most common English words. One contains a dictionary of 80,000 words. Trying all of these words as passwords takes only 80 seconds. Passwords Likely for a UserIf Sandy is selecting a password, she is probably not choosing a word completely at random. Most likely Sandy's password is something meaningful to her. People typically choose personal passwords, such as the name of a spouse, a child, a brother or sister, a pet, a street name, or something memorable or familiar. If we restrict our password attempts to just names of people (first names), streets , projects, and so forth, we generate a list of only a few hundred possibilities at most. Trying this number of passwords takes under a second! Even a person working by hand could try ten likely candidates in a minute or two. Thus, what seemed formidable in theory is in fact quite vulnerable in practice, and the likelihood of successful penetration is frightening. Morris and Thompson [MOR79] confirmed our fears in their report on the results of having gathered passwords from many users, shown in Table 4-2. Figure 4-15 (based on data from that study) shows the characteristics of the 3,289 passwords gathered. The results from that study are distressing, and the situation today is likely to be the same. Of those passwords, 86 percent could be uncovered in about one week's worth of 24- hour -a-day testing, using the very generous estimate of 1 millisecond per password check. Figure 4-15. Users' Password Choices.
Table 4-2. Distribution of Actual Passwords.
Lest you dismiss these results as dated (they were reported in 1979), Klein repeated the experiment in 1990 [KLE90] and Spafford in 1992 [SPA92]. Each collected approximately 15,000 passwords. Klein reported that 2.7 percent of the passwords were guessed in only 15 minutes of machine time and 21 percent were guessed within a week! Spafford found the average password length was 6.8 characters, and 28.9 percent consisted of only lowercase alphabetic characters. Notice that both these studies were done after the Internet worm (described in Chapter 3) succeeded, in part by breaking weak passwords. Even in 2002, the British online bank Egg found users still choosing weak passwords [BUX02]. A full 50 percent of passwords for their online banking service were family members ' names: 23 percent children's names, 19 percent a spouse or partner, and 9 percent their own. Alas, pets came in at only 8 percent, while celebrities and football (soccer) stars tied at 9 percent each. And in 1998, Knight and Hartley [KNI98] reported that approximately 35 percent of passwords are deduced from syllables and initials of the account owner's name. Several network sites post dictionaries of phrases, science fiction characters, places, mythological names, Chinese words, Yiddish words, and other specialized lists. All these lists are posted to help site administrators identify users who have chosen weak passwords, but the same dictionaries can also be used by attackers of sites that do not have such attentive administrators. The COPS [FAR90], Crack [MUF92], and SATAN [FAR95] utilities allow an administrator to scan a system for weak passwords. But these same utilities, or other homemade ones, allow attackers to do the same. People think they can be clever by picking a simple password and replacing certain characters, such as 0 (zero) for letter O, 1 (one) for letter I or L, 3 (three) for letter E or @ (at) for letter A. But users aren't the only people who could think up these substitutions. Knight and Hartley [KNI98] list, in order, 12 steps an attacker might try to determine a password. These steps are in increasing degree of difficulty (number of guesses), and so they indicate the amount of work to which the attacker must go in order to derive a password. Here are their password guessing steps:
Although the last step will always succeed, the steps immediately preceding it are so time consuming that they will deter all but the dedicated attacker for whom time is not a limiting factor. Plaintext System Password ListTo validate passwords, the system must have a way of comparing entries with actual passwords. Rather than trying to guess a user's password, an attacker may instead target the system password file. Why guess when with one table you can determine all passwords with total accuracy? On some systems, the password list is a file, organized essentially as a two-column table of user IDs and corresponding passwords. This information is certainly too obvious to leave out in the open. Various security approaches are used to conceal this table from those who should not see it. You might protect the table with strong access controls, limiting access to the operating system. But even this tightening of control is looser than it should be, because not every operating system module needs or deserves access to this table. For example, the operating system scheduler, accounting routines, or storage manager have no need to know the table's contents. Unfortunately, in some systems, there are n +1 known users: n regular users and the operating system. The operating system is not divided, so all its modules have access to all privileged information. This monolithic view of the operating system implies that a user who exploits a flaw in one section of the operating system has access to all the system's deepest secrets. A better approach is to limit table access to the modules that need access: the user authentication module and the parts associated with installing new users, for example. If the table is stored in plain sight, an intruder can simply dump memory at a convenient time to access it. Careful timing may enable a user to dump the contents of all of memory and, by exhaustive search, find values that look like the password table. System backups can also be used to obtain the password table. To be able to recover from system errors, system administrators periodically back up the file space onto some auxiliary medium for safe storage. In the unlikely event of a problem, the file system can be reloaded from a backup, with a loss only of changes made since the last backup. Backups often contain only file contents, with no protection mechanism to control file access. (Physical security and access controls to the backups themselves are depended upon to provide security for the contents of backup media.) If a regular user can access the backups , even ones from several weeks, months, or years ago, the password tables stored in them may contain entries that are still valid. Finally, the password file is a copy of a file stored on disk. Anyone with access to the disk or anyone who can overcome file access restrictions can obtain the password file. Encrypted Password FileThere is an easy way to foil an intruder seeking passwords in plain sight: encrypt them. Frequently, the password list is hidden from view with conventional encryption or one-way ciphers. With conventional encryption, either the entire password table is encrypted or just the password column. When a user's password is received, the stored password is decrypted, and the two are compared. Even with encryption, there is still a slight exposure because for an instant the user's password is available in plaintext in main memory. That is, the password is available to anyone who could obtain access to all of memory. A safer approach uses one-way encryption, defined in Chapter 2. The password table's entries are encrypted by a one-way encryption and then stored. When the user enters a password, it is also encrypted and then compared with the table. If the two values are equal, the authentication succeeds. Of course, the encryption has to be such that it is unlikely that two passwords would encrypt to the same ciphertext , but this characteristic is true for most secure encryption algorithms. With one-way encryption, the password file can be stored in plain view. For example, the password table for the Unix operating system can be read by any user, unless special access controls have been installed. Because the contents are encrypted, backup copies of the password table are no longer a problem. There is always the possibility that two people might choose the same password, thus creating two identical entries in the password file. Even though the entries are encrypted, each user will know the plaintext equivalent. For instance, if Bill and Kathy both choose their passwords on April 1, they might choose APRILFOOL as a password. Bill might read the password file and notice that the encrypted version of his password is the same as Kathy's. Unix circumvents this vulnerability by using a password extension, called the salt. The salt is a 12-bit number formed from the system time and the process identifier. Thus, the salt is likely to be unique for each user, and it can be stored in plaintext in the password file. The salt is concatenated to Bill's password ( pw ) when he chooses it; E ( pw +salt B ) is stored for Bill, and his salt value is also stored. When Kathy chooses her password, the salt is different because the time or the process number is different. Call this new one salt K . For her, E ( pw +salt K ) and salt K are stored. When either person tries to log in, the system fetches the appropriate salt from the password table and combines that with the password before performing the encryption. The encrypted versions of ( pw +salt) are very different for these two users. When Bill looks down the password list, the encrypted version of his password will not look at all like Kathy's. Storing the password file in a disguised form relieves much of the pressure to secure it. Better still is to limit access to processes that have a legitimate need for access. In this way, the password file is protected to a level commensurate with the protection provided by the password itself. Someone who has broken the controls of the file system has access to data, not just passwords, and that is a serious threat. But if an attacker successfully penetrates the outer security layer, the attacker still must get past the encryption of the password file to access the useful information in it. Indiscreet UsersGuessing passwords and breaking encryption can be tedious or daunting. But there is a simple way to obtain a password: Get it directly from the user! People often tape a password to the side of a terminal, or write it on a card just inside the top desk drawer . Users are afraid they will forget their passwords, or they cannot be bothered trying to remember them. It is particularly tempting to write the passwords down when users have several accounts. Users sharing work or data may also be tempted to share passwords. If someone needs a file, it is easier to say "my password is x ; get the file yourself" than to arrange to share the file. This situation is a result of user laziness , but it may be brought about or exacerbated by a system that makes sharing inconvenient. Password Selection CriteriaSo what can we conclude about passwords? They should be hard to guess and difficult to determine exhaustively. But the degree of difficulty should be appropriate to the security needs of the situation. To these ends, we present several guidelines for password selection:
To help users select good passwords, some systems provide meaningless but pronounceable passwords. For example, the VAX VMS system randomly generates five passwords from which the user chooses one. They are pronounceable, so that the user should be able to repeat and memorize them. However, the user may misremember a password because of having interchanged syllables or letters of a meaningless string. (The sound "bliptab" is easily misremembered as "blaptib" or "blabtip.") Other systems encourage users to change their passwords regularly. The regularity of password change is usually a system parameter, which can be changed for the characteristics of a given installation. Suppose the frequency is set at 30 days. Some systems begin to warn the user after 25 days that the password is about to expire. Others wait until 30 days and inform the user that the password has expired. Some systems nag without end, whereas other systems cut off a user's access if a password has expired . Still others force the user immediately into the password change utility on the first login after 30 days. Grampp and Morris [GRA84a] argue that this reminder process is not necessarily good. Choosing passwords is not difficult, but under pressure a user may adopt any password, just to satisfy the system's demand for a new password. Furthermore, if this is the only time when a password can be changed, a user who selects a bad password and realizes it cannot change until the next scheduled time. Sometimes when systems force users to change passwords periodically, users with favorite passwords will alternate between two passwords each time a change is required. To prevent password reuse, Microsoft Windows 2000 systems refuse to accept any of the k most recently used passwords. One user of such a system went through 24 password changes each month, just to cycle back to the favorite password. One-Time PasswordsA one-time password is one that changes every time it is used. Instead of assigning a static phrase to a user, the system assigns a static mathematical function. The system provides an argument to the function, and the user computes and returns the function value. Such systems are also called challengeresponse systems because the system presents a challenge to the user and judges the authenticity of the user by the user's response. Here are some simple examples of one-time password functions; these functions are overly simplified to make the explanation easier. Very complex functions can be used in place of these simple ones for host authentication in a network.
One-time passwords are very important for authentication because (as becomes clear in Chapter 7) an intercepted password is useless. However, their usefulness is limited by the complexity of algorithms people can be expected to remember. A password generating device, similar to a pocket calculator, can implement more complex functions. Several models are readily available at reasonable prices. They are very effective at countering the threat of transmitting passwords in plaintext across a network. (See Sidebar 4-4 for another dilemma in remote authentication.) The Authentication ProcessAuthentication usually operates as described previously. However, users occasionally mistype their passwords. A user who receives a message of INCORRECT LOGIN will carefully retype the login and gain access to the system. Even a user who is a terrible typist should be able to log in successfully in a few tries. Some authentication procedures are intentionally slow. A legitimate user will not complain if the login process takes 5 or 10 seconds. To a penetrator who is trying an exhaustive search or a dictionary search, however, 5 or 10 seconds per trial makes this class of attack generally infeasible. Someone who continually fails to log in may not be an authorized user. Systems commonly disconnect a user after a small number of failed logins, forcing the user to reestablish a connection with the system. (This action will slow down a penetrator who is trying to penetrate the system by telephone. After a small number of failures, the penetrator must redial, which takes a few seconds.) In more secure installations, stopping penetrators is more important than tolerating users' mistakes. For example, some system administrators assume that all legitimate users can type their passwords correctly within three tries. After three successive password failures, the account for that user is disabled and only the security administrator can reenable it. This action identifies accounts that may be the target of attacks by penetrators.
Fixing Flaws in the Authentication ProcessPassword authentication assumes that anyone who knows a password is the user to whom the password belongs. As we have seen, passwords can be guessed, deduced, or inferred. Some people give out their passwords for the asking. Other passwords have been obtained just by someone watching a user typing in the password. The password can be considered as a preliminary or first-level piece of evidence, but skeptics will want more convincing proof. There are several ways to provide a second level of protection, including another round of passwords or a challenge-response interchange. ChallengeResponse SystemsAs we have just seen, the login is usually time invariant. Except when passwords are changed, each login looks like every other. A more sophisticated login requires a user ID and password, followed by a challengeresponse interchange. In such an interchange, the system prompts the user for a reply that will be different each time the user logs in. For example, the system might display a four-digit number, and the user would have to correctly enter a function such as the sum or product of the digits. Each user is assigned a different challenge function to compute. Because there are many possible challenge functions, a penetrator who captures the user ID and password cannot necessarily infer the proper function. A physical device similar to a calculator can be used to implement a more complicated response function. The user enters the challenge number, and the device computes and displays the response for the user to type in order to log in. (For more examples, see Chapter 7's discussion of network authentication.) Impersonation of LoginIn the systems we have described, the proof is one sided. The system demands certain identification of the user, but the user is supposed to trust the system. However, a programmer can easily write a program that displays the standard prompts for user ID and password, captures the pair entered, stores the pair in a file, displays SYSTEM ERROR; DISCONNECTED, and exits. This attack is a type of Trojan horse. The perpetrator sets it up, leaves the terminal unattended, and waits for an innocent victim to attempt a login. The nave victim may not even suspect that a security breach has occurred. To foil this type of attack, the user should be sure the path to the system is reinitialized each time the system is used. On some systems, turning the terminal off and on again or pressing the BREAK key generates a clear signal to the computer to halt any running process for the terminal. (Microsoft chose <CTRL><ALT><DELETE> as the path to the secure authorization mechanism for this reason.) Not every computer recognizes power-off or BREAK as an interruption of the current process, though. And computing systems are often accessed through networks, so that physical reinitialization is impossible . Alternatively, the user can be suspicious of the computing system, just as the system is suspicious of the user. The user will not enter confidential data (such as a password) until convinced that the computing system is legitimate. Of course, the computer acknowledges the user only after passing the authentication process. A computing system can display some information known only by the user and the system. For example, the system might read the user's name and reply "YOUR LAST LOGIN WAS 10 APRIL AT 09:47." The user can verify that the date and time are correct before entering a secret password. If higher security is desired, the system can send an encrypted timestamp. The user decrypts this and discovers that the time is current. The user then replies with an encrypted timestamp and password, to convince the system that a malicious intruder has not intercepted a password from some prior login. Authentication Other Than PasswordsSome sophisticated authentication devices are now available. These devices include handprint detectors, voice recognizers, and identifiers of patterns in the retina. Authentication with such devices uses unforgeable physical characteristics to authenticate users. The cost continues to fall as these devices are adopted by major markets; the devices are useful in very high security situations. More normal security needs can be handled by a combination of login and characteristics such as work hours, physical location, or patterns of access. For example, a user who seeks access to files with no justifiable reason might not be authentic . After a system detects this access violation attempt, it might disconnect the user and suspend access until a security administrator clears the matter. Therefore, a penetrator who enters the system as "Jones" has to continue to act like Jones to remain on the system. (See Chapter 7 for more on intrusion and anomaly detection.) Authentication is essential for an operating system because accurate user identification is the key to individual access rights. Most operating systems and computing system administrators have applied reasonable but stringent security measures to lock out illegal users before they can access system resources. But, as reported in Sidebar 4-5, sometimes an inappropriate mechanism is forced into use as an authentication device. |
< Free Open Study > |