针对WSN异常数据检测改进的孤立森林方法

异常数据检测一直是无线传感器网络安全的重要防护手段.针对现有方案计算复杂度高和检测精度低等问题,提出一种离散二进制粒子群优化孤立森林算法(BPSO-iForest).依据选择性集成思想,利用离散二进制粒子群算法改进由孤立森林算法生成的初始森林,选取初始森林中精度高、差异性大的隔离树,构建最优孤立森林,提升异常数据的检测精度和算法的执行效率.在无线传感器网络数据集上,与传统孤立森林、随机森林算法及其改进算法进行对比实验,结果表明本算法的检测精度和执行效率有明显的提升.

Improved Isolation Forest Method for WSN Anomaly Data Detection

Abnormal data detection has always been an important means of protection for wireless sensor network security.Aiming at the problems of high computational complexity and low detection accuracy of the existing schemes,a discrete binary particle swarm optimization isolation forest algorithm(BPSO-iForest)was proposed.Based on the idea of selective integration,the discrete binary particle swarm algorithm is used to improve the initial forest generated by the isolation forest algorithm,the isolation tree with high accuracy and large difference in the initial forest is selected to construct the optimal isolation forest to improve the accuracy of abnormal data detection and algorithm execution efficiency.The wireless sensor network dataset is compared with traditional isolation forest,random forest algorithms and its revised method.The results show that the detection accuracy and execution efficiency of the algorithm have been significantly improved.