Password Security: A Case History -- Morris and Thompson, 1979
--------------------------------------------------------------

Presents a history of the password mechanism in UNIX.

Goal: high security with minimal burden to the users.

Remote access systems are more vulnerable to attacks than local systems.

The security of the password encryption algorithm is only a small part
of the security problem. physical security of the system,
communications security of the network, casual unauthorized access,
and lack of security awareness among users are more serious issues in
practice.

Initial implementation: 
      - cleartext password file with heavily restricted read/write access
      - problems include copies of the passwd file made by sysadms
      which can be used later, copies of the file on backup media, and
      dispersal of user information among many files which could actually
      go into the passwd file

First approach: 
      - encrypt passwd; used M-209 cipher machine in which given the
      plaintext and ciphertext, it was difficult to find the key; so
      passwd is used as a key to encrypt a constant
      - Attacks: attack on the algorithm (unsuccessful); key search 
      (very successful, as many as 86% of the passwds could be found in
      the authors experience) solutions to this could be to force
      users to choose long passwds with special characters, or making
      the system generate good passwds
      
Enhancements:
      - Slower encryption (DES), so that key search becomes slower 
      - Less predictable passwds, enforced by the passwd entry system
      - Salted passwds: A random salt appended by the system to the
      passwd. This prevents attacks based on dictionaries of encrypted
      words.
      - Slight modification to DES to make fast DES chips unusable in
      passwd-cracking.
      - subtle point: perform encryption even in the case of invalid
      user name so that the 'bad guy' cannot guess valid user names

Other:
      - 'external security code' for remote logins
      - logging all successful and unsuccessful accesses to the system

* What is the meaning of the line "Salts make it nearly impossible to
  find out whether a person has used the same passwd on all the
  machines"?

* Why haven't other mechanisms for authentication become more
  popular. For example, recognizing the speech of a person as he reads
  a random sentence, recognizing the pattern of inter-key-stroke
  delays as a person types out a sentence etc., Could these mechanisms
  augument the passwd mechanism to provide greater security ?