Optimization of Real-Valued Self Set in Immunity-based WSN Intrusion Detection

Real-valued negative selection algorithm (RNSA) has been a main algorithm of immunity-based intrusion detection in wireless sensor networks (WSNs). However, the real-valued initial self-set which is used to train detectors has some defects: boundary invasion and overlapping among the self-samples. Detectors trained by the initial self-set may have the problem of boundary invasion, which will resulted false detection, and due to the redundancy of the self-set, the generation efficiency is low. Therefore, the self-set need to be optimized before training stage. In this paper, we proposed a new improved variable threshold self-set optimization algorithm to optimize the self-set before training the detectors based on a new concept of sample's affinity density, which is used to measure the distribution density of the sample. The experiments based on the Iris data set and wireless sensor networks intrusion detection were used to test the effectiveness of the algorithm. The results show that the optimization of self-set can solve the problem of boundary invasion, improve the detector training efficiency, and reduce the false alarm rate of the abnormal detection system.