基于多源域偏移数据特征融合的数字化校园网络IoT入侵检测方法

An IoT Intrusion Detection Method for Digital Campus Network Based on Feature Fusion of Offset Data in Multi-source Domain

目前常规的数字化校园网络入侵检测方法主要通过构建自编码器对数据进行批量处理,通过对恶意攻击的数据特征进行学习,从而构建出检测标准,常常易忽略不同数据源的偏移问题,导致检测精度不佳。为此,提出基于多源域偏移数据特征融合的数字化校园网络IoT入侵检测方法。首先,针对数字化校园网络数据进行特征维度判断,从而明确可提取的数据类型;然后,通过构建偏移系数矫正指标,对多源域偏移数据进行融合处理,采用优化迭代的方式,对不同层次的特征参数进行优化;最后,提取特征偏移标准,并将融合后的特征向量与偏移标准进行对比,实现对入侵数据的有效判断。实验结果表明,所提方法检测精度较高,检测效果较好。

Currently conventional intrusion detection methods for digital campus networks mainly batch process the data by constructing a autoencoder and construct a detection criterion by learning the data features of malicious attacks,which leads to poor detection accuracy due to neglecting the offsets of different data sources.In this regard,An IoT intrusion detection method for digital campus network based on feature fusion of offset data in multi-source domain was proposed.Firstly,the feature dimension was judged,so as to clarify the type of data that can be extracted.Then by constructing the offset coefficient correction index,the offset data from multi-source domain was fused and processed,and the optimization iteration was used to optimize the feature parameters at different levels.Finally,the feature offset criteria are extracted,and the fused feature vectors were compared with the offset criteria to realize the effective judgment of the intrusion data.The experimental results show that,the proposed method has higher detection accuracy and better detection effect detection effect.