Biomedical Engineering Reference
In-Depth Information
• Too low or too high memory usage.
• Inappropriate disk activity.
• Unexpected frequency of file changes as measured for example by checksums or
file size.
• SIGABRT signal from abnormally terminated UNIX processes.
• Presence of non-self.
Of course, it would also be possible to use 'positive' signals, as discussed in the
previous section, such as the absence of some normal 'health' signals.
Consequently, those antibodies or detectors that match (first signal) those
antigens within a radius, defined by a measure such as the above (second signal),
will proliferate. Having thereby identified the dangerous components, further
confirmation could then be sought by sending it to a special part of the system
simulating another attack. This would have the further advantage of not having to
send all detectors to confirm danger. In conclusion, using these ideas from the
Danger Theory has provided a better grounding of danger labels in comparison to
self/non-self, whilst at the same time relying less on human competence.
Conclusion
To conclude, the Danger Theory is not about the way AIS represent data. Instead,
it provides ideas about which data the AIS should represent and deal with. They
should focus on dangerous, i.e., interesting data.
It could be argued that the shift from non-self to danger is merely a symbolic
label change that achieves nothing. We do not believe this to be the case, since
danger is a grounded signal, and non-self is (typically) a set of feature vectors with
no further information about their meaning.
The danger signal helps us to identify which subset of feature vectors is of
interest. A suitably defined danger signal thus overcomes many of the limitations
of self-non-self selection. It restricts the domain of non-self to a manageable size,
removes the need to screen against all self, and deals adaptively with scenarios
where self (or non-self) changes over time.
The challenge is clearly to define a suitable danger signal, a choice that might
prove as critical as the choice of fitness function for an evolutionary algorithm. In
addition, the physical distance in the biological system should be translated into a
suitable proxy measure for similarity or causality in an AIS. We have made some
suggestions in this paper about how to tackle these challenges in a variety of
domains, but the process is not likely to be trivial. Nevertheless, if these challenges
are met, then future AIS applications might derive considerable benefit, and new
insights, from the Danger Theory.
Acknowledgments We would like to thank the two anonymous reviewers, whose comments
greatly improved this paper.
Search WWH ::




Custom Search