摘要
当前物联网终端软件缺陷数据在挖掘过程中,没有深入考虑不同类别与属性分布的因素,对异常数据采用了容忍的思路,导致神经网络缺陷判断过程需要大量迭代,检测效果不佳.为此,提出了基于混合神经网络的物联网终端软件缺陷检测方法 .根据变异概率对异常物联网终端软件运行数据进行剔除,结合数据类别与属性概率分布对剔除异常后的数据进行融合处理,通过预测软件运行数据融合结果进行数据特征提取后,将提取到的终端软件运行数据特征输入由神经网络与长短期记忆网络混合组成的混合神经网络中,利用长短期记忆网络预测软件的缺陷情况,使用神经网络判定预测结果的真实性,从而得到精准的检测结果 .测试结果表明:基于混合神经网络的检测方法的误检率与漏检率较低,说明在混合神经网络的辅助下,缺陷检测的精度得到了显著提升,可以广泛应用于软件缺陷检测领域.
In the current process of mining software defect data in terminals of Internet of Things,a tolerant approach has been adopted for abnormal data,resulting in the poor detection results.Therefore,a software defect detection method for terminals of Internet of Things based on hybrid neural networks has been proposed.Based on the probability of mutation and abnormal terminal software operation data of Internet of Things being removed,fusing the data after removing anomalies with the probability distribution of data categories and attributes,the data features are extracted by predicting the fusion results of software operation data.The extracted terminal software operation data features are input into a hybrid neural network composed of a neural network and a memory network.The memory network is used to predict the defect situation of the software,and the neural network is used to determine the authenticity of the prediction results,thereby obtaining accurate detection results.The test results show that the detection method based on hybrid neural networks has a lower false detection rate and missed detection rate,indicating that with the assistance of hybrid neural networks,the accuracy of defect detection has been significantly improved and can be widely applied in the field of software defect detection.
作者
郑逸凡
黄瑞垠
WAQAR KHAN
ZHENG Yifan;HUANG Ruiyin;WAQAR KHAN(Fuzhou University of International Studies and Trade,Fuzhou 350202,China)
出处
《通化师范学院学报》
2025年第8期43-47,共5页
Journal of Tonghua Normal University
基金
福建省自然科学基金项目(2024J08242)
福州外语外贸学院校级教育教学研究项目(JF2023033)。
关键词
混合神经网络
物联网
终端软件
软件定位
缺陷检测
hybrid neural networks
Internet of Things
terminal software
software positioning
defect detection
