Artificial immune system based on normal model and immune learning

Inspired form natural immune system, a new artificial immune system was proposed to detect, recognize and eliminate the non-selfs such as computer worms and software faults. Because unknown non-selfs are very difficult to detect only by recognizing the features of the non-selfs, a normal model was built to provide an easy and effective tool for completely detecting the unknown non-selfs by detecting the known selfs. The probability for detecting unknown non-selfs with traditional approaches depends on the complexity for the features of unknown non-selfs, and the usage of the selfs for detecting the non-selfs in some systems has been neglected. After the normal model is built with the space-time properties of the selfs for the systems, the probability for detecting the unknown non-selfs can be improved with the normal model. To overcome the bottleneck for finding which to recognize and how to learn the unknown worms, an adaptive immune learning model was proposed against the unknown worms, by searching in the multi-dimension feature space of worms with random evolutionary computation. The goal of the adaptive immune learning was to find the most similar known worm to the unknown worm or establish a new class for the unknown worm. The normal model and the innate immune tier on the normal model provided a better source of unknown non-selfs so that the probability for recognizing the unknown worms was increased. To recognize and learn the unknown non-selfs with uncertainty in the artificial immune system, the evolutionary immune algorithm was used to search and reason with uncertainty. At last, a prototype for using the artificial immune system in anti-worm and fault diagnosis applications validated the models.