Authentication | Research & Encyclopedia Articles

Authentication

One of the needs of social or business interaction is the authentication of the principals involved in it. Just because a man says his name is John Smith does not mean that he indeed is John Smith; the fact must be proved in other ways. In the real world, the authentication of a person's identity is achieved in various ways--using a birth certificate, a state-issued picture ID, a passport, a social security card, knowledge of one's mother's maiden name, and so on. A document can be authenticated as having been authored or agreed to by a principal, using a simple handwritten signature--this age-old technique is still, for the most part, very reliable. If additional verification is needed, then the signature can be backed up with attestation using a notary public, or a sworn affidavit, or similar established legal procedures.

In an online environment where the principal in an interaction is not physically available for inspection, authentication is a more involved affair. It is not then possible to authenticate a principal by her appearance, or by her possession of real-life credentials. Specifically, in the context of secure electronic message transfer and digital signatures, it is necessary to show that a certain message or signed document is:

• Authentic--the recipient of the message accepts that it was sent by the claimed sender, with the same contents as it had upon receipt; the message was neither sent nor altered by another.
• Unforgeable--there is proof that only the claimed sender, and none else, deliberately signed the message, and that the signature has not been copied over from another document.
• Non-repudiable--the sender of the message cannot credibly deny being its signatory.

These are all issues that also arise in offline interactions, and there the techniques used have matured over the ages to resolve these issues. To do likewise in an online environment, cryptography is applied to create mechanisms for authentication of communications between principals. A principal who is able to decrypt a coded message using a particular key is able to assume that the message is authentic if it contains a valid checksum, for instance. The receiving principal is able to infer that only the purported sender, presumed to be the sole possessor of the corresponding encryption key, could have sent the message. Thus, if the encryption and decryption keys are held in private, it is easy to come up with a protocol for authenticating a message as coming from a particular sender. However, the disadvantage to this method is that it requires two pairs of keys for each pair of principals in the system, which may lead to impractical problems of key management since the number of pairs grows very quickly as the number of principals increases. Asymmetric key (also called public-key) encryption can solve this problem to a certain extent. Key management still remains a serious issue, but it is not quite as bad.

Historically, the first successful authentication protocol was described by R.M. Needham and M.D. Schroeder in a paper published in the Communications of the ACM, volume 21, in 1978. Their protocol is now known as the Needham-Schroeder protocol, and serves as the basis for most of the authentication techniques in widespread use. Its most well-known use is in forming the basis of the Kerberos system, which provides authentication between clients and servers in single-domain networks such as university computing systems and company intranets. Needham and Schroeder described, among other things, the concept of an authentication server, whose role is to provide a secure way for pairs of principals on a network to obtain shared keys. Another important aspect of these researchers' work was the notion of a cryptographic challenge--a user should not have to submit a password to an authentication service over an open communication channel; rather, the authentication server issues the user a ticket encrypted with his secret key. Only the legitimate user can decrypt the ticket with the secret key, hence only the right principal finds it of any use.

Passwords and PIN numbers are common forms of authentication for access to login accounts, bank accounts, etc. The basic idea involved there is that the system stores an encrypted form of the password in a password file, against the user entry. When a person supplies a password to a terminal or other authentication unit, it is immediately encrypted, and the encrypted form is compared with the entry stored in the file; if there is a match, then the person is authenticated, else not. Obviously, this method works only when there is an easy way to encrypt a password, but when decrypting the ciphertext version back to the original password is very hard. This is an application of the notion of a one-way function. A similar application is made in case of message or document signatures, by computing a message-digest or hash for the whole message. Given the message, it should be easy to obtain the hash using a one-way hash function, but it should be impossible to obtain the message back using just the hash. The hashes for two messages should never be the same.

The technique of using a human's features such as voice, appearance, or the like for authentication is called biometric authentication. Currently passwords, PIN numbers, electronic keycards, and the like are the common means of interaction for authentication, and fancy ideas such as retina scans (scanning of the retina in the eye--every retina is distinct!) or palm scans are too expensive and limited in availability. Even so, it is certain that biometric authentication will gradually increase in importance and become more and more common. There are some advantages to this--although it is conceivable that one could fake (or replay using a sophisticated device) another's voice or appearance, it is not so easy to fake the distance between the ischial tuberosities (sitting bones). Biometric authentication, if designed and applied with care, thus is not easily fooled. The downside to this in some people's minds is, of course, the aspect of Big Brother always watching: would the government or other powerful entities be able to keep tabs on people, see where they went and what they did, if biometric authentication were widespread?