Unknown non-self detection & robustness of distributed artificial immune system with normal model

Biological immune system is typical distributed parallel system for processing biological information to defense the body against viruses and diseases. Inspired from nature, a distributed artificial immune system with the normal model is proposed for detecting unknown non-selfs such as worms and software faults. Traditional approaches are used to learn unknown features and types of the unknown non-selfs, but the learning problem can not be solved for human immune system in short time, neither that for the machines. A new detecting approach is proposed with the normal model of the system, and the selfs of the system are represented and detected at first. Depending on strictness and completeness of the normal model, the selfs are known and the process for detecting the selfs is much easier and more accurate than that for the non-selfs. Not only the artificial immune system can detect the non-selfs, but also the system can eliminate the non-selfs and repair the damaged parts of the system by itself. Minimization of the non-selfs and maximization of the selfs show robustness of the artificial immune system, and robustness of the distributed artificial immune system can be reduced according to each independent module.