Review: Protection and the Control of Information Sharing in Multics.

From: Sellakumaran Kanagarathnam (sellak_at_windows.microsoft.com)
Date: Wed Jan 07 2004 - 16:15:54 PST

  • Next message: David Coleman: "Review of Multics paper"

    The author describes the design principles and functional objectives of
    the Multics system, the various mechanisms used for the protection and
    control of information (access control lists, hierarchical control of
    access specifications, authentication system, memory protection,
    supervisor-user protection) and discusses the various weaknesses as
    perceived by him at that time of the system.

    It is very interesting to note that the very first design principle talk
    about lack of access by default. Decades down the line, we are talking
    about following/enforcing this in newer OS. The principle of least
    privilege is also in the same lines. The objectives of provisions of
    decentralization of control and extensibility of default system options
    (protection scheme here) are important considerations that is applicable
    any system even today.

    The access control list gives provision to grant access to group of
    users which is very valuable. It is good to note that the same user can
    be part of different groups/roles/projects.
    But providing such a list for each object that we create could be a time
    consuming process and it is good to note that there is a default initial
    control list. It would be interesting to note how this default initial
    control list is handled in the newer versions of Multics as it makes
    more sense, at least to me, to consider this list to be a common
    appendix from the storage point of view.
    Access control list entries permitting backup and bulk I/O by default
    seemed to contradict the design principle 1. In the procedure for the
    bulk input of cards, I guess the concerned user alone is granted
    permission to move.
    The scheme that required all user names, once registered to be permanent
    was a little bit strange.

    The extra mechanisms involved around user authentication like forcing
    interactive login, proxy login, turning off printer while entering
    password, auto logout, all the logging techniques, anonymous login all
    seem to highlight the careful thought given to increase security and
    flexibility. However it would be interesting to find out if the
    utility to create the password so that it is easy to remember and to
    minimize the need for written copies of password was useful at all.

    It is interesting to note the three fields in the descriptor extension
    and how they are used for protection.
    The author lists the weaknesses and one of the major weaknesses of
    complex user interface will be impacting the user experience.

    The author concludes that the system was designed to be basic and
    extensible and that is evident in many a features in the system.
    It is interesting to note how much attention is given to security and
    attacks in this design 30 years back and still now we are dealing with
    so many vulnerabilities in the new servers.


  • Next message: David Coleman: "Review of Multics paper"

    This archive was generated by hypermail 2.1.6 : Wed Jan 07 2004 - 16:15:33 PST