基于GA-RBF神经网络的物联网业务安全态势感知  

Security situation awareness of IoT business based on GA-RBF neural network

作　　者：张进军 李育泽 ZHANG Jinjun;LI Yuze(Department of Information Management,Anhui Vocational College of Police Officers,Hefei 230031,China;School of Languages and Media,Anhui University of Finance&Economics,Bengbu 233000,China)

机构地区：[1]安徽警官职业学院信息管理系,合肥230031 [2]安徽财经大学文学院,安徽蚌埠233000

出　　处：《黑龙江工程学院学报》2025年第1期29-34,共6页Journal of Heilongjiang Institute of Technology

基　　金：安徽省教育厅高等学校省级质量工程项目(2022cyxy010)。

摘　　要：在物联网设备动态加入或离开网络的背景下,网络拓扑结构的持续变化对静态安全感知方法的可扩展性构成了挑战。同时,现有部分识别模型过度依赖专家知识和模型训练,导致其动态适应性不足,难以确保物联网业务的流程安全。因此,对基于GA-RBF神经网络的物联网业务安全态势感知方法展开研究。该方法基于经典粗糙集理论,引入并行约简思想,利用条件熵评估特征要素的重要度,并采用约简规则去除冗余特征,从而高效提取物联网业务安全态势的关键特征。采用遗传算法对RBF神经网络进行优化,将编码后的物联网业务运行数据直接应用于网络中,通过计算适应度构建安全态势特征要素样本矩阵,并输入至RBF神经网络进行训练,以确立映射关系并获取最终的安全态势感知结果。实验结果显示,将该方法应用于大数据环境下的物联网业务链中,其态势感知结果与实际环境高度吻合,感知误差低至0.5%,平均迭代次数不超过110,均方误差仅为0.03%,响应时间最短可达120 ms。这些数据充分证明了所提方法在感知误差、精度、训练效率以及响应速度方面的优越性能,为物联网业务安全态势感知提供了有力支持。In the context of IoT devices dynamically joining or leaving the network,the continuous changes in network topology pose a challenge to the scalability of static security awareness methods.Meanwhile,some existing recognition models overly rely on expert knowledge and model training,resulting in insufficient dynamic adaptability and difficulty in ensuring the security of IoT business processes.Therefore,research is conducted on the security situational awareness method for IoT business based on GA-RBF neural network.This method is based on classical rough set theory,introduces parallel reduction ideas,uses conditional entropy to evaluate the importance of feature elements,and adopts reduction rules to remove redundant features,thereby efficiently extracting key features of the security situation of networked business.Using genetic algorithm to optimize RBF neural network,the encoded IoT business operation data is directly applied to the network.The fitness is calculated to construct a sample matrix of security situation feature elements,which is input to the RBF neural network for training to establish the mapping relationship and obtain the final security situation awareness result.The experimental results show that applying this method to the IoT business chain in a big data environment,the situational awareness results are highly consistent with the actual environment,with a perception error as low as 0.5%,an average iteration count of no more than 110,a mean square error of only 0.03%,and a response time as short as 120ms.These data fully demonstrate the superior performance of the proposed method in terms of perception error,accuracy,training efficiency,and response speed,providing strong support for IoT business security situational awareness.

关 键 词：物联网业务 安全态势感知 RBF网络 遗传算法 要素特征 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]