earticle

영어

A novel technique for intrusion detection based on real-valued dual negative selection scheme is proposed in this paper. In traditional real-valued negative selection algorithms, whether the candidate detectors can detect self-set or not totally relies on the affinity extent and the constant-sized mechanism is unfavorable to eliminating the black holes with irregular sizes. The proposed technique introduces the mechanism of variable-sized dual negative selection, in which each mutual detector has to pass three tests. Firstly, the new mutual detector should not be detected by the current existing ones. In other words, the existence of the new detector is necessary. Secondly, those detectors which can detect self-set will be eliminated. Thirdly, the detectors distribution has to be optimized aiming at enhancing the detecting efficiency. Experimental results demonstrate that the proposed technique has much less black holes, fewer detectors and higher detecting rates.