Part IV.
TECHNICAL CONTROLS

Chapter 16

IDENTIFICATION AND AUTHENTICATION

For most systems, identification and authentication (I&A) is the first line of defense. I&A is a technical measure that prevents unauthorized people (or unauthorized processes) from entering a computer system.

I&A is a critical building block of computer security since it is the basis for most types of access control and for establishing user accountability.¹⁰⁷ Access control often requires that the system be able to identify and differentiate among users. For example, access control is often based on least privilege, which refers to the granting to users of only those accesses required to perform their duties. User accountability requires the linking of activities on a computer system to specific individuals and, therefore, requires the system to identify users.

Identification is the means by which a user provides a claimed identity to the system. Authentication¹⁰⁸ is the means of establishing the validity of this claim.

This chapter discusses the basic means of identification and authentication, the current technology used to provide I&A, and some important implementation issues.

Computer systems recognize people based on the authentication data the systems receive. Authentication presents several challenges: collecting authentication data, transmitting the data securely, and knowing whether the person who was originally authenticated is still the person using the computer system. For example, a user may walk away from a terminal while still logged on, and another person may start using it.

There are three means of authenticating a user's identity, which can be used alone or in combination:

• something the individual knows (a secret -- e.g., a password, Personal Identification Number (PIN), or cryptographic key);
   
• something the individual possesses (a token -- e.g., an ATM card or a smart card); and
   
• something the individual is (a biometric -- e.g., such characteristics as a voice pattern, handwriting dynamics, or a fingerprint).
  

While it may appear that any of these means could provide strong authentication, there are problems associated with each. If people wanted to pretend to be someone else on a computer system, they can guess or learn that individual's password; they can also steal or fabricate tokens. Each method also has drawbacks for legitimate users and system administrators: users forget passwords and may lose tokens, and administrative overhead for keeping track of I&A data and tokens can be substantial. Biometric systems have significant technical, user acceptance, and cost problems as well.

This section explains current I&A technologies and their benefits and drawbacks as they relate to the three means of authentication. Although some of the technologies make use of cryptography because it can significantly strengthen authentication, the explanations of cryptography appear in Chapter 19, rather than in this chapter.

16.1 I&A Based on Something the User Knows

The most common form of I&A is a user ID coupled with a password. This technique is based solely on something the user knows. There are other techniques besides conventional passwords that are based on knowledge, such as knowledge of a cryptographic key.

16.1.1 Passwords

In general, password systems work by requiring the user to enter a user ID and password (or pass phrase or personal identification number). The system compares the password to a previously stored password for that user ID. If there is a match, the user is authenticated and granted access.

Benefits of Passwords. Passwords have been successfully providing security for computer systems for a long time. They are integrated into many operating systems, and users and system administrators are familiar with them. When properly managed in a controlled environment, they can provide effective security.

Problems With Passwords. The security of a password system is dependent upon keeping passwords secret. Unfortunately, there are many ways that the secret may be divulged. All of the problems discussed below can be significantly mitigated by improving password security, as discussed in the sidebar. However, there is no fix for the problem of electronic monitoring, except to use more advanced authentication (e.g., based on cryptographic techniques or tokens).

1. Guessing or finding passwords. If users select their own passwords, they tend to make them easy to remember. That often makes them easy to guess. The names of people's children, pets, or favorite sports teams are common examples. On the other hand, assigned passwords may be difficult to remember, so users are more likely to write them down. Many computer systems are shipped with administrative accounts that have preset passwords. Because these passwords are standard, they are easily "guessed." Although security practitioners have been warning about this problem for years, many system administrators still do not change default passwords. Another method of learning passwords is to observe someone entering a password or PIN. The observation can be done by someone in the same room or by someone some distance away using binoculars. This is often referred to as shoulder surfing.
    
2. Giving passwords away. Users may share their passwords. They may give their password to a co-worker in order to share files. In addition, people can be tricked into divulging their passwords. This process is referred to as social engineering.
    
3. Electronic monitoring. When passwords are transmitted to a computer system, they can be electronically monitored. This can happen on the network used to transmit the password or on the computer system itself. Simple encryption of a password that will be used again does not solve this problem because encrypting the same password will create the same ciphertext; the ciphertext becomes the password.
    
4. Accessing the password file. If the password file is not protected by strong access controls, the file can be downloaded. Password files are often protected with one-way encryption¹⁰⁹ so that plain-text passwords are not available to system administrators or hackers (if they successfully bypass access controls). Even if the file is encrypted, brute force can be used to learn passwords if the file is downloaded (e.g., by encrypting English words and comparing them to the file).

Passwords Used as Access Control. Some mainframe operating systems and many PC applications use passwords as a means of restricting access to specific resources within a system. Instead of using mechanisms such as access control lists (see Chapter 17), access is granted by entering a password. The result is a proliferation of passwords that can reduce the overall security of a system. While the use of passwords as a means of access control is common, it is an approach that is often less than optimal and not cost-effective.

16.1.2 Cryptographic Keys

Although the authentication derived from the knowledge of a cryptographic key may be based entirely on something the user knows, it is necessary for the user to also possess (or have access to) something that can perform the cryptographic computations, such as a PC or a smart card. For this reason, the protocols used are discussed in the Smart Tokens section of this chapter. However, it is possible to implement these types of protocols without using a smart token. Additional discussion is also provided under the Single Log-in section.

16.2 I&A Based on Something the User Possesses

Although some techniques are based solely on something the user possesses, most of the techniques described in this section are combined with something the user knows. This combination can provide significantly stronger security than either something the user knows or possesses alone.¹¹⁰

Objects that a user possesses for the purpose of I&A are called tokens. This section divides tokens into two categories: memory tokens and smart tokens.

16.2.1 Memory Tokens

Memory tokens store, but do not process, information. Special reader/writer devices control the writing and reading of data to and from the tokens. The most common type of memory token is a magnetic striped card, in which a thin stripe of magnetic material is affixed to the surface of a card (e.g., as on the back of credit cards). A common application of memory tokens for authentication to computer systems is the automatic teller machine (ATM) card. This uses a combination of something the user possesses (the card) with something the user knows (the PIN).

Some computer systems authentication technologies are based solely on possession of a token, but they are less common. Token-only systems are more likely to be used in other applications, such as for physical access. (See Chapter 15.)

Benefits of Memory Token Systems. Memory tokens when used with PINs provide significantly more security than passwords. In addition, memory cards are inexpensive to produce. For a hacker or other would-be masquerader to pretend to be someone else, the hacker must have both a valid token and the corresponding PIN. This is much more difficult than obtaining a valid password and user ID combination (especially since most user IDs are common knowledge).

Another benefit of tokens is that they can be used in support of log generation without the need for the employee to key in a user ID for each transaction or other logged event since the token can be scanned repeatedly. If the token is required for physical entry and exit, then people will be forced to remove the token when they leave the computer. This can help maintain authentication.

Problems With Memory Token Systems. Although sophisticated technical attacks are possible against memory token systems, most of the problems associated with them relate to their cost, administration, token loss, user dissatisfaction, and the compromise of PINs. Most of the techniques for increasing the security of memory token systems relate to the protection of PINs. Many of the techniques discussed in the sidebar on Improving Password Security apply to PINs.

1. Requires special reader. The need for a special reader increases the cost of using memory tokens. The readers used for memory tokens must include both the physical unit that reads the card and a processor that determines whether the card and/or the PIN entered with the card is valid. If the PIN or token is validated by a processor that is not physically located with the reader, then the authentication data is vulnerable to electronic monitoring (although cryptography can be used to solve this problem).
    

Attacks on memory-card systems have sometimes been quite creative. One group stole an ATM machine that they installed at a local shopping mall. The machine collected valid account numbers and corresponding PINs, which the thieves used to forge cards. The forged cards were then used to withdraw money from legitimate ATMs.

2. Token loss. A lost token may prevent the user from being able to log in until a replacement is provided. This can increase administrative overhead costs.
    
   The lost token could be found by someone who wants to break into the system, or could be stolen or forged. If the token is also used with a PIN, any of the methods described above in password problems can be used to obtain the PIN. Common methods are finding the PIN taped to the card or observing the PIN being entered by the legitimate user. In addition, any information stored on the magnetic stripe that has not been encrypted can be read.
    
3. User Dissatisfaction. In general, users want computers to be easy to use. Many users find it inconvenient to carry and present a token. However, their dissatisfaction may be reduced if they see the need for increased security.

16.2.2 Smart Tokens

A smart token expands the functionality of a memory token by incorporating one or more integrated circuits into the token itself. When used for authentication, a smart token is another example of authentication based on something a user possesses (i.e., the token itself). A smart token typically requires a user also to provide something the user knows (i.e., a PIN or password) in order to "unlock" the smart token for use.

There are many different types of smart tokens. In general, smart tokens can be divided three different ways based on physical characteristics, interface, and protocols used. These three divisions are not mutually exclusive.

Physical Characteristics. Smart tokens can be divided into two groups: smart cards and other types of tokens. A smart card looks like a credit card, but incorporates an embedded microprocessor. Smart cards are defined by an International Standards Organization (ISO) standard. Smart tokens that are not smart cards can look like calculators, keys, or other small portable objects.

Interface. Smart tokens have either a manual or an electronic interface. Manual or human interface tokens have displays and/or keypads to allow humans to communicate with the card. Smart tokens with electronic interfaces must be read by special reader/writers. Smart cards, described above, have an electronic interface. Smart tokens that look like calculators usually have a manual interface.

Protocol. There are many possible protocols a smart token can use for authentication. In general, they can be divided into three categories: static password exchange, dynamic password generators, and challenge-response.

• Static tokens work similarly to memory tokens, except that the users authenticate themselves to the token and then the token authenticates the user to the computer.
   
• A token that uses a dynamic password generator protocol creates a unique value, for example, an eight-digit number, that changes periodically (e.g., every minute). If the token has a manual interface, the user simply reads the current value and then types it into the computer system for authentication. If the token has an electronic interface, the transfer is done automatically. If the correct value is provided, the log-in is permitted, and the user is granted access to the system.
   
• Tokens that use a challenge-response protocol work by having the computer generate a challenge, such as a random string of numbers. The smart token then generates a response based on the challenge. This is sent back to the computer, which authenticates the user based on the response. The challenge-response protocol is based on cryptography. Challenge-response tokens can use either electronic or manual interfaces.

There are other types of protocols, some more sophisticated and some less so. The three types described above are the most common.

Benefits of Smart Tokens

Smart tokens offer great flexibility and can be used to solve many authentication problems. The benefits of smart tokens vary, depending on the type used. In general, they provide greater security than memory cards. Smart tokens can solve the problem of electronic monitoring even if the authentication is done across an open network by using one-time passwords.

1. One-time passwords. Smart tokens that use either dynamic password generation or challenge-response protocols can create one-time passwords. Electronic monitoring is not a problem with one-time passwords because each time the user is authenticated to the computer, a different "password" is used. (A hacker could learn the one-time password through electronic monitoring, but would be of no value.)
    
2. Reduced risk of forgery. Generally, the memory on a smart token is not readable unless the PIN is entered. In addition, the tokens are more complex and, therefore, more difficult to forge.
    
3. Multi-application. Smart tokens with electronic interfaces, such as smart cards, provide a way for users to access many computers using many networks with only one log-in. This is further discussed in the Single Log-in section of this chapter. In addition, a single smart card can be used for multiple functions, such as physical access or as a debit card.

Problems with Smart Tokens

Like memory tokens, most of the problems associated with smart tokens relate to their cost, the administration of the system, and user dissatisfaction. Smart tokens are generally less vulnerable to the compromise of PINs because authentication usually takes place on the card. (It is possible, of course, for someone to watch a PIN being entered and steal that card.) Smart tokens cost more than memory cards because they are more complex, particularly challenge-response calculators.

1. Need reader/writers or human intervention. Smart tokens can use either an electronic or a human interface. An electronic interface requires a reader, which creates additional expense. Human interfaces require more actions from the user. This is especially true for challenge-response tokens with a manual interface, which require the user to type the challenge into the smart token and the response into the computer. This can increase user dissatisfaction.
    
2. Substantial Administration. Smart tokens, like passwords and memory tokens, require strong administration. For tokens that use cryptography, this includes key management. (See Chapter 19.)

16.3 I&A Based on Something the User Is

Biometric authentication technologies use the unique characteristics (or attributes) of an individual to authenticate that person's identity. These include physiological attributes (such as fingerprints, hand geometry, or retina patterns) or behavioral attributes (such as voice patterns and hand-written signatures). Biometric authentication technologies based upon these attributes have been developed for computer log-in applications.

Biometric authentication is technically complex and expensive, and user acceptance can be difficult. However, advances continue to be made to make the technology more reliable, less costly, and more user-friendly.

Biometric systems can provide an increased level of security for computer systems, but the technology is still less mature than that of memory tokens or smart tokens. Imperfections in biometric authentication devices arise from technical difficulties in measuring and profiling physical attributes as well as from the somewhat variable nature of physical attributes. These may change, depending on various conditions. For example, a person's speech pattern may change under stressful conditions or when suffering from a sore throat or cold.

Due to their relatively high cost, biometric systems are typically used with other authentication means in environments requiring high security.

16.4 Implementing I&A Systems

Some of the important implementation issues for I&A systems include administration, maintaining authentication, and single log-in.

16.4.1 Administration

Administration of authentication data is a critical element for all types of authentication systems. The administrative overhead associated with I&A can be significant. I&A systems need to create, distribute, and store authentication data. For passwords, this includes creating passwords, issuing them to users, and maintaining a password file. Token systems involve the creation and distribution of tokens/PINs and data that tell the computer how to recognize valid tokens/PINs. For biometric systems, this includes creating and storing profiles.

The administrative tasks of creating and distributing authentication data and tokens can be a substantial. Identification data has to be kept current by adding new users and deleting former users. If the distribution of passwords or tokens is not controlled, system administrators will not know if they have been given to someone other than the legitimate user. It is critical that the distribution system ensure that authentication data is firmly linked with a given individual. Some of these issues are discussed in Chapter 10 under User Administration.

In addition, I&A administrative tasks should address lost or stolen passwords or tokens. It is often necessary to monitor systems to look for stolen or shared accounts.

Authentication data needs to be stored securely, as discussed with regard to accessing password files. The value of authentication data lies in the data's confidentiality, integrity, and availability. If confidentiality is compromised, someone may be able to use the information to masquerade as a legitimate user. If system administrators can read the authentication file, they can masquerade as another user. Many systems use encryption to hide the authentication data from the system administrators.¹¹¹ If integrity is compromised, authentication data can be added or the system can be disrupted. If availability is compromised, the system cannot authenticate users, and the users may not be able to work.

16.4.2 Maintaining Authentication

So far, this chapter has discussed initial authentication only. It is also possible for someone to use a legitimate user's account after log-in.¹¹² Many computer systems handle this problem by logging a user out or locking their display or session after a certain period of inactivity. However, these methods can affect productivity and can make the computer less user-friendly.

16.4.3 Single Log-in

From an efficiency viewpoint, it is desirable for users to authenticate themselves only once and then to be able to access a wide variety of applications and data available on local and remote systems, even if those systems require users to authenticate themselves. This is known as single log-in.¹¹³ if the access is within the same host computer, then the use of a modern access control system (such as an access control list) should allow for a single log-in. If the access is across multiple platforms, then the issue is more complicated, as discussed below. There are three main techniques that can provide single log-in across multiple computers: host-to-host authentication, authentication servers, and user-to-host authentication.

Host-to-Host Authentication. Under a host-to-host authentication approach, users authenticate themselves once to a host computer. That computer then authenticates itself to other computers and vouches for the specific user. Host-to-host authentication can be done by passing an identification, a password, or by a challenge-response mechanism or other one-time password scheme. Under this approach, it is necessary for the computers to recognize each other and to trust each other.

Authentication Servers. When using authentication server, the users authenticate themselves to a special host computer (the authentication server). This computer then authenticates the user to other host computers the user wants to access. Under this approach, it is necessary for the computers to trust the authentication server. (The authentication server need not be a separate computer, although in some environments this may be a cost-effective way to increase the security of the server.) Authentication servers can be distributed geographically or logically, as needed, to reduce workload.

User-to-Host. A user-to-host authentication approach requires the user to log-in to each host computer. However, a smart token (such as a smart card) can contain all authentication data and perform that service for the user. To users, it looks as though they were only authenticated once.

16.5 Interdependencies

There are many interdependencies among I&A and other controls. Several of them have been discussed in the chapter.

Logical Access Controls. Access controls are needed to protect the authentication database. I&A is often the basis for access controls. Dial-back modems and firewalls, discussed in Chapter 17, can help prevent hackers from trying to log-in.

Audit. I&A is necessary if an audit log is going to be used for individual accountability.

Cryptography. Cryptography provides two basic services to I&A: it protects the confidentiality of authentication data, and it provides protocols for proving knowledge and/or possession of a token without having to transmit data that could be replayed to gain access to a computer system.

16.6 Cost Considerations

In general, passwords are the least expensive authentication technique and generally the least secure. They are already embedded in many systems. Memory tokens are less expensive than smart tokens, but have less functionality. Smart tokens with a human interface do not require readers, but are more inconvenient to use. Biometrics tends to be the most expensive.

For I&A systems, the cost of administration is often underestimated. Just because a system comes with a password system does not mean that using it is free. For example, there is significant overhead to administering the I&A system.

References:

Alexander, M., ed. "Keeping the Bad Guys Off-Line." Infosecurity News. 4(6), 1993. pp. 54-65.

American Bankers Association. American National Standard for Financial Institution Sign-On Authentication for Wholesale Financial Transactions. ANSI X9.26-1990. Washington, D.C., February 28, 1990.

CCITT Recommendation X.509. The Directory - Authentication Framework. November 1988 (Developed in collaboration, and technically aligned, with ISO 9594-8).

Department of Defense. Password Management Guideline. CSC-STD-002-85. April 12, 1985.

Feldmeier, David C., and Philip R. Kam. "UNIX Password Security - Ten Years Later." Crypto '89 Abstracts. Santa Barbara, CA: Crypto '89 Conference, August 20-24, 1989.

Haykin, Martha E., and Robert B. J. Warnar. Smart Card Technology: New Methods for Computer Access Control. Special Publication 500-157. Gaithersburg, MD: National Institute of Standards and Technology, September 1988.

Kay, R. "Whatever Happened to Biometrics?" Infosecurity News. 4(5), 1993. pp. 60-62.

National Bureau of Standards. Password Usage. Federal Information Processing Standard Publication 112. May 30, 1985.

National Institute of Standards and Technology. Automated Password Generator. Federal Information Processing Standard Publication 181. October, 1993.

National Institute of Standards and Technology. Guideline for the Use of Advanced Authentication Technology Alternatives. Federal Information Processing Standard Publication 190. October, 1994.

Salamone, S. "Internetwork Security: Unsafe at Any Node?" Data Communications. 22(12), 1993. pp. 61-68.

Sherman, R. "Biometric Futures." Computers and Security. 11(2), 1992. pp. 128-133.

Smid, Miles, James Dray, and Robert B. J. Warnar. "A Token-Based Access Control System for Computer Networks." Proceedings of the 12th National Computer Security Conference. National Institute of Standards and Technology, October 1989.

Steiner, J.O., C. Neuman, and J. Schiller. "Kerberos: An Authentication Service for Open Network Systems." Proceedings Winter USENIX. Dallas, Texas, February 1988. pp. 191-202.

Troy, Eugene F. Security for Dial-Up Lines. Special Publication 500-137, Gaithersburg, MD: National Bureau of Standards, May 1986.

Footnotes:

107. Not all types of access control require identification and authentication.
108. Computers also use authentication to verify that a message or file has not been altered and to verify that a message originated with a cetain person. This chapter only addresses user authentication. The other forms of authentication are addressed in the Chapter 19.
109. One-way encryption algorithms only provide for the encryption of data. The resulting ciphertext cannot be decrypted. When passwords are entered into the system they are one-way encrypted, and the result is compared with the stored ciphertext. (See the Chapter 19.)
110. For the purpose of understanding how possession-based I&A works, it is not necessary to distinguish whether possession of a token in various systems is identification or authentication.
111. Masquerading by system administrators cannot be prevented entirely. However, controls can be set up so that improper actions by the system administrator can be detected in audit records.
112. After a user signs on, the computer treats all commands originating from the user's physical device (such as a PC or terminal) as being from the user.
113. Single log-in is somewhat of a misnomer. It is currently not feasible to have one sign-on for every computer system a user might wish to access. The types of single log-in described apply mainly to groups of systems (e.g., within an organization or a consortium).