🚀 go-pugleaf

Perhaps I am a bit naive, not being a virus researcher, but could the
problems related to benign virus-like (i.e. self-replicating) programs
be solved by implementing a protocol to control the installation of
self replicating programs? What I envision is something like the
following:

     1. Self replicator sends a message to the target machine, asking if
        self replicators are allowed to be installed there. One of several
        things happens:
        A. The target has been set up to allow this to happen automatically,
           go to step 2.
        B. The target might allow it, but must ask for judgement from the
           system administrator (mechanics of "asking" could include E-mail
           to and from administrator, direct interaction, or whatever). If
           the administrator OKs it, go to 2.
           If the administrator doesn't okay it, go to 6.
        C. The system has been set to always deny these requests. Go to 6.
        D. The system doesn't support the protocol. It should not be at risk
           unless it always executes things sent to it from other systems,
           but I don't think this system could be considered completely safe.
           In any event, it cannot participate further in the protocol.
     2. Send an affirmative message to the source (of the replicator) machine,
        inviting the transmission of the program.
     3. Receive the program to a directory that no users except the system
        administrators can access.
     4. Execute whatever security scanning and checking mechanisms are desired
        for this program. If malignancy is detected, notify the administrators
        and stop the process. Do not send negative response messages to the
        source, since this could allow someone to send many small bits of
        malignancy over time and determine exactly what holes your scanning
        and checking procedures allow, then write a self replicator which
        passes all your tests but is still malignant.
     5. If it looks OK, move the program the place where it should be executed.
        I would propose that all such self-replicating programs from one area
        under one user identification, and that that user have the lowest
        possible rights on that system, and _no_ rights to create or modify
        files. Reading files and sending messages should be allowed.
     6. Error situations: Send a negative response and stop.

The advantages I see to this are that self-replicating programs can be used
where they will be benificial, but _cannot_ be used without the permission
of the system administrators. Step 4 will help make sure that the program is
not malignant (obviously dependent upon the quality of your checking), and
following the caveats in step 5 should prevent it from modifying files or
data (well at least it will be no more capable than a program installed by the
system administratos). Once the program has been installed, in fact, the
security problem is about the same as that for programs written at the site,
and looks identical (to me) to the security problems for programs obtained
from other sites by FTP, download, or purchase. Allowing it to send messages
does mean that someone could send all your sensitive data back to themselves,
but your sensitive data should already be protected from being read by any
old program, anyways.

Like I said earlier, I'm not a researcher, so my theoretical groundings are not
very good. Any and all comments on this are welcomed. My thoughts on this are
mainly due to a desire to someday see the fabled "knowbot" which can hunt
around on "the net" for information on a particular topic, following leads to
different sites as necessary.

Regards
Tim Hare
Systems Programmer
Florida Department of Transportation ("Potholes Built and Fuilt While-U-Wait")