PINs and passkeys

(i) PINs

Personal identification numbers (PINs) fall under the category of fixed (time-invariant) passwords. They are most often used in conjunction with “something possessed”, typically a physical token such as a plastic banking card with a magnetic stripe, or a chipcard. To prove one’s identity as the authorized user of the token, and gam access to the privileges associated therewith, entry of the correct PIN is required when the token is used. This provides a second level of security if the token is lost or stolen. PINs may also serve as the second level of security for entry to buildings which have an independent first level of security (e.g., a security guard or video camera).

For user convenience and historical reasons, PINs are typically short (relative to fixed password schemes) and numeric, e.g., 4 to 8 digits. To prevent exhaustive search through such a small key space (e.g., 10 000 values for a 4-digit numeric PIN), additional procedural constraints are necessary. For example, some automated cash dispenser machines accessed by banking cards confiscate a card if three incorrect PINs are entered successively; for others, incorrect entry of a number of successive PINs may cause the card to be “locked” or deactivated, thereafter requiring a longer PIN (e.g., 8 digits) for reactivation following such suspicious circumstances.

In an on-line system using PINs or reusable passwords, a claimed identity accompanied by a user-entered PIN may be verified by comparison to the PIN stored for that identity in a system database. An alternative is to use the PIN as a key for a MAC (see Chapter 9).

In an off-line system without access to a central database, information facilitating PIN verification must be stored on the token itself. If the PIN need not be user-selected, this may be done by defining the PIN to be a function of a secret key and the identity associated with the token; the PIN is then verifiable by any remote system knowing this master key.

In an off-line system, it may also be desirable to allow the PIN to be user-selectable, to facilitate PIN memorization by users. In this case, the PIN may be encrypted under a master key and stored on the token, with the master key known to all off-line terminals that need to be capable of verifying the token. A preferable design is to store a one-way function of the PIN, user identity, and master key on the token.

(ii) Two-stage authentication and password-derived keys

Human users have difficulty remembering secret keys which have sufficient entropy to provide adequate security. Two techniques which address this issue are now described.

When tokens are used with off-line PIN verification, a common technique is for the PIN to serve to verify the user to the token, while the token contains additional independent information allowing the token to authenticate itself to the system (as a valid token representing a legitimate user). The user is thereby indirectly authenticated to the system by a two-stage process. This requires the user have possession of the token but need remember only a short PIN, while a longer key (containing adequate entropy) provides cryptographic security for authentication over an unsecured link.

A second technique is for a user password to be mapped by a one-way hash function into a cryptographic key (e.g., a 56-bit DES key). Such password-derived keys are called passkeys. The passkey is then used to secure a communications link between the user and a system which also knows the user password. It should be ensured that the entropy of the user’s password is sufficiently large that exhaustive search of the password space is not more efficient than exhaustive search of the passkey space (i.e., guessing passwords is not easier than guessing 56-bit DES keys); see Table 10.1 for guidance.

An alternative to having passkeys remain fixed until the password is changed is to keep a running sequence number on the system side along with each user’s password, for use as a time-variant salt communicated to the user in the clear and incremented after each use. A fixed per-user salt could also be used in addition to a running sequence number.

Passkeys should be viewed as long-term keys, with use restricted to authentication and key management (e.g., rather than also for bulk encryption of user data). A disadvantage of using password-derived keys is that storing each user’s password within the system requires some mechanism to protect the confidentiality of the stored passwords.

< Prev   CONTENTS   Source   Next >