yamauchi@granite.cs.rochester.edu (Brian Yamauchi) (09/19/90)
landman@hanami.Eng.Sun.COM (Howard A. Landman) writes: > I am not arguing in favor of AVVs, but have a few technical ideas to > throw out for discussion: > > One method of limiting the risk of an AVV would be to make it spread > more rapidly where there are other viruses than where there are not. > For example, the virus could award itself "food points" every time it > "eats" a bad virus, and require a certain number of points before it > attempts to replicate. The copy, of course, starts life with no food > points ... > > This way, the AVV would be almost unable to spread among systems which > were apparently clean, but would spread rapidly in an obviously sick > environment. This property could be quite useful in focusing the > concentration of the AVV to where it was needed the most. > > Many variants of this scheme are possible. For example, the virus > might split food points with its copy, but then lose a point every > time it runs and there's nothing to disinfect. Eventually it could > "starve" and remove itself. The problem with "food points", and other self-limiting strategies in general, is that from an evolutionary perspective these limitations are *flaws*, and the evolutionary pressure will be to remove these. For an example: suppose you have a virus with a mutation that prevents it from losing food points when it runs or reproduces. This mutant will constantly accumulate food points and replicate faster and faster until it eventually becomes a dominant subspecies of this virus. Of course, you could introduce AAVVs to hunt down and kill the mutant strain of AVV, but if you try to limit the reproduction of the AAVVs then they will be under the same evolutionary pressure to eliminate those limitations. Another option might be a "genetic purity" behavior for the AVVs where they scan any other AVVs of the same species they encounter and kill them if the others' code differs from its own. But in the long run, the mutant AVVs will have the upper hand since they can reproduce faster and exterminate any members of the unmutated strain that they encounter. Mike Travers at the MIT Media Lab performed some interesting experiments with simulated creatures that evolved to take advantage of bugs in the simulator. Like the AVVs above, these creatures could forage for food and reproduce by giving a certain amount of its food to its offspring. One species evolved which was able to reproduce at the clock rate of the simulation -- it did this by giving its offspring more food than the parent. The parent dies immediately, but not before generating a child (who dies on the next clock tick, after producing another child, and so on). If AVVs ever become necessary, it will probably be futile to attempt to hobble them with a software ball-and-chain. They should probably be designed as viable (i.e. non-self-destructive / rapidly reproducing) but benign organisms from the start. Of course, some will mutate to become destructive, but at least there won't be any evolutionary pressure toward destructiveness. If anything, there will be a pressure towards benignity, since a harmful virus that mutates into a nondestructive form will be less likely to be detected and destroyed. I don't consider AVVs either necessary or beneficial now, but if the virus population even expands to the level where the majority of systems carry some sort of virus, then at least AVVs could increase the probability of having a benign population on your system. _______________________________________________________________________________ Brian Yamauchi University of Rochester yamauchi@cs.rochester.edu Computer Science Department _______________________________________________________________________________